[00:00:06] What's up, everybody, what's going on? Welcome, welcome to the @ArtistsOfData Happy Hour. It is Friday, August 26th. My last day of work. Outprice industries can't believe it. Man Thank you so much, everybody at the team war and Dave Kyle, you guys are all awesome. Thank you so much for giving me this opportunity to work with you all. Really enjoyed my time there. I'm onto the next move. I'll be

[00:00:29] At Comet Emelle

[00:00:31] Working, uh, working on some really cool, interesting, fun stuff, uh, that's going to be happening in about a

[00:00:36] Month. Um, so

[00:00:38] I'm excited for that. Guys also shout out to the fact

[00:00:41] That as of today, we just crossed

[00:00:43] 75000. Download the podcast. That's wild to me. That's a huge metric for me. I'm really, really excited for for that.

[00:00:51] Thank you, everybody

[00:00:51] Who's been listening and helping to make that possible.

[00:00:54] Literally, I couldn't have done it without you

[00:00:56] Guys because it would be very difficult for me to listen to all 171 episodes.

[00:01:01] Um, but

[00:01:02] Whatever happened hundreds of times, I would have

[00:01:04] To watch each episode,

[00:01:06] Um, shout at everybody in

[00:01:07] The room. What's up, kanji in the building?

[00:01:09] David, what's going on? Uh, we got a guy. Monica, good to see you again. Monica has been quite a long time. Happy to have you back then. What's going on? Yeah, Ben Taylor, what's going on? Matt Sharp has become my friend, uh, Serge and everybody else super excited to have you guys here. Um, somebody else also had a

[00:01:27] Transition today is David

[00:01:29] Knickerbocker. David, let's go to him. Man, you got you got some comments here with people interested to hear about your new gig. I am, too, man. So how's it feel, man has been a Maccabee. For how long has it feel to, uh, to to leave after, you know, how many years it's been?

[00:01:44] It's definitely surreal. I've been a Macfie for six years and so I made it made a lot of friends, but most of my really close friends have already moved on from Macfie. But but I did have a core group that I was close with, so today was pretty difficult saying say my last goodbyes and whatnot. [00:02:00]

[00:02:00] Yeah. But you know, you got

[00:02:02] To do brave things

[00:02:03] Sometimes.

[00:02:04] So yeah, I take a risk man.

[00:02:07] If it is scary as hell though, you know, because there's no safety net anymore, you know, making a jump to start my own thing. And so we'll see

[00:02:14] How this is exciting. Man, I'm excited to hear about this. If you want to share any details, by all means, let us know what you think about. I'm sure we'd all be happy to be here, but. But if not, then like,

[00:02:25] It's

[00:02:25] Been my third job in three years. I remember leaving

[00:02:28] That first job back

[00:02:29] In the end of twenty eighteen. I was like a weight off my chest, like I just eat at that place and then and then my second job leaving bowl, I was like genuinely just bad because I loved my team and it's such a good group of friends, such a great team. Um and this is the first time I'm leaving and I'm just I'm feeling genuinely like I did, like, I'm super, super excited for for what's coming next, uh, as I'm sure you are as well.

[00:02:54] Um, I meant like, how about this. Yeah. Yeah.

[00:02:57] So it's it's a I'm not going to go into too much details yet because we're still working on a lot of stuff. But, but it's a disinformation mitigation company. And so we'll be mapping out that kind of malign disinformation, not just like people making mistakes on social media, but bigger than that. And so there's going to be a lot of network science involved. But NLP, it's all the stuff that you guys see me get all excited about. I get to build a tech stack in my image. And so it's going to be awesome. You know, it's just me and another guy right now. So it's just we're building the idea the way that we think that we should and we'll see how well it does.

[00:03:32] So that's cool, man. Sounds like a really, really

[00:03:34] Awesome, awesome project.

[00:03:36] And, um, shout out everybody watching on LinkedIn

[00:03:39] And on YouTube

[00:03:41] And Twitter. You guys got comments. Go ahead and drop them in the chat. We'd be happy to take your comments. I know a couple of people had some questions cued up.

[00:03:50] Abe, you had a

[00:03:51] Question that you'd hit me up about a couple of days ago when to go to your question didn't go to Russell's question. And then as soon as Mark gets here, we'll get to his question. And in the meantime, if you guys have questions, let [00:04:00] us know right there in the chat. And I will ask you,

[00:04:03] Did you just say my name Abia? Yeah. Yeah. Kind of cut out for saying yes. I had on first time speaking. So so I just started like my first SQL project at work. So I got my big boy pants on. You know, I started working on my first query and it's kind of getting some help. But I'm not asking everyone

[00:04:22] For, like, the answers.

[00:04:24] I'm just point me in the right direction. And I kind of did the step backwards. I like I started writing the query and then I didn't gather all the requirements first.

[00:04:33] So, like, what do you guys recommend

[00:04:35] For someone that's like doing SQL like for real at a job to like, you know, how to the look at the

[00:04:42] Problem or the question like

[00:04:43] Figure out what I need to do to, you know, get to the final results.

[00:04:47] Yeah, absolutely. Man.

[00:04:48] My approach would be to figure out the

[00:04:51] The schema as much as possible, see if there's a Data dictionary floating around, if I can get like a yard or something just so I can conceptualize

[00:04:59] What is in this abstract database,

[00:05:02] Which, um, the my experience that doesn't happen as often as. We go and ask a lot of questions. Connect with people and pick their brains, these local experts, and stitch together your own web of knowledge, but I think you might have been going through something similar. Right. Just a question about this topic a while ago. What have you learned since then?

[00:05:20] So. So I think

[00:05:23] Your main question was when you're gathering it, how do you just go about getting the right information so that you find you're not going to be barking up the wrong tree or. Right. Yeah, so there's a Data dictionary, but there's like thousands of tables because it's like I work in, like, the health

[00:05:39] Care like financial part

[00:05:40] Of my school. So we do the financials for all the clinics. There's like thousands of tables

[00:05:45] And there's not

[00:05:45] Vrd, but I'm only going to be working out of like ten tables. And then it's just like so it's just like I go to the Data dictionary, then I just try to figure out like, all right, this goes here, I guess. And I'm just like making like a cheat sheet for myself. [00:06:00] So like, all right, these are the ten tables that we're going to work with the most. And these are what I'm going to do to

[00:06:05] Kind of like the most. Yeah.

[00:06:07] So like a couple of things that kind of come to mind for me. One is, yeah, just ask questions because you're not dumb. If you ask the questions of how things are related to you are a blessed man to have a Data dictionary, that's fantastic because they don't just they don't grow on trees. And then the other thing is I have a like a little query that just pulls from the information schema. It says like and I just say, give me the table name and the column name where color name is like. And then I just type in like I'm looking for something that sounds like revenue. So I just search revenue and it brings up every instance of revenue that might be useful to me and that's been super helpful. And then the only other little thing I would add is tons of little tiny queries

[00:06:48] Like Show me that,

[00:06:49] Show me top ten. What does it look like? OK, and we top ten again. But I change something. Does is is it working for me. And just like iterating forward until until I'm getting somewhere bigger and then and then maybe try something a little bigger before finally doing a big query. It's going to take

[00:07:05] Longer and I avoid wasting time

[00:07:07] If I need that. Whatever you just said, I need that. I will drop it in the chat. I could do that. I don't know if you can copy it, though.

[00:07:15] Crap, I could share the text file. So let's hear from Ken on this. And by the way, if you guys did not get a chance to tune in to Ken and Ali Militar live on Instagram yesterday, you guys missed out, but you guys are running it back.

[00:07:30] So it got deleted on Instagram, which was a huge bummer. We had a great convo and then there were some software issues on the way out. But, yeah, I don't know when we're doing it, but we're definitely running it back again. I want to one hundred percent echo what Eric was saying about iterating on things, something especially when you're new to a company. I find this a lot with consulting where we have to

[00:07:54] Get up to speed really quickly is working and iterating

[00:07:57] In that fashion, but also keeping a kind [00:08:00] of running list of questions and asking in batches rather than constantly pinning someone on slack. Like when you run into things, it's this this list that you're creating. And sometimes you realize as you get further down the iteration process, you're like, oh, I answer that question myself. And so choosing specific points or maybe even setting like a 30 minute meeting or checking with someone

[00:08:21] Or you can ask a dedicated

[00:08:23] Questions can be really effective, sometimes even more effective than the ones that everyone wants to get rid of all the meetings on their calendar. But sometimes, again, in the early stages of getting together with a team or getting getting integrated with an organization, if you're using meetings very systematically and very purposefully, I think that they can they can create a lot of value for you and also building a relationship with the other

[00:08:46] People you work with as well.

[00:08:47] Thank you very much. Can anybody else want to comment on this? I saw some great comments right there in the chat. Anybody wants to speak on it, just go for it. If not, we can move on to the next question. Do you feel like your question was answered? Yeah. Yeah, yeah. Thank you. Appreciate everyone. Awesome. Awesome. Yeah.

[00:09:05] Cool.

[00:09:05] Guys, let's keep moving.

[00:09:07] And the Q right now is Russell Mark. And we got Eric

[00:09:10] And then Jacob. Oh, Russell. Go for it.

[00:09:14] Yeah. Thanks.

[00:09:15] Good evening, everybody. So I've got a

[00:09:17] Lot of questions related to Gan's etc. and it's kind of a bit of fun. But I think there's a there's a definite

[00:09:26] Ethical question behind gasoline

[00:09:29] Because go kind of.

[00:09:32] Yeah, I'd say

[00:09:33] It's a camera. So I think it's trying to pay me out of the picture. Maybe it doesn't like my face. I don't know. It was visibility again. Russell Yeah. So, yeah. So recently I've had the option to use creative mode on LinkedIn, so I haven't done it for a long time, but I've recently enabled it. And I know you can set the animated video for your profile picture. However, most of the ones that I've seen [00:10:00] just edit from the profile picture to a video and it's a it's a really kind of junk edit. It doesn't segre or transition particularly well. So it dawned on me that perhaps some of these great

[00:10:14] Gan's

[00:10:15] And deep like videos I saw Signorelli had a great one for Morgan Freeman that someone had to put out on LinkedIn, which you guys might have seen it. So I wondered if anyone is considered doing the same for that, their own picture.

[00:10:28] So to keep their

[00:10:29] Own profile picture and make their own profile picture and have that picture speak, whatever they're going through at the time, just as a kind of an interesting way to make the best use of this new feature in an LinkedIn. I mean, to me, it'd be good that you can probably see I've got a few gray hairs

[00:10:45] On my beard now.

[00:10:46] My profile

[00:10:46] Picture was maybe only

[00:10:48] Twenty four months ago. All of these gray has come during the covid. But yeah, it was a AIs to look a little younger if I was to do that. So I'm thirsty. Is anyone considered doing it? And then secondly, if it is possible to do that, what are the ethical implications? We're talking about a bit of a fun way of using this, but it can be done. What else can be done out in the in the wider media

[00:11:11] That could have

[00:11:13] Poor ethical repercussions

[00:11:15] And go for it?

[00:11:17] I like this question.

[00:11:18] It's a good

[00:11:19] Question, Russell. So I think I thought about doing it. So when you go do head shots like so I've got my headshot on LinkedIn and that was to go to a session, get a professional photographer. You have to get like a wardrobe, like it's a process. And so I'd much rather just generate like ten thousand frames from like a very high resolution, like for a camera. And then I would just have my account that I could then play with change the expressions. I could also modify the

[00:11:46] Attractiveness, like I just make

[00:11:48] Myself a little bit more attractive. So not enough that you would call foul, but enough that when you saw me in person you'd be like, man, he's having an off day. But actually I just like increased my attractiveness 20 percent.

[00:11:59] If the rest [00:12:00] of us the chance, then come on.

[00:12:04] Yeah, it's. Is it ethical? I don't know. I guess you have like models that get touched up today, like post-processing. Photoshop like they and so someone brought that up, there was a debate that's been brought up and they said

[00:12:15] That this this could be very

[00:12:16] Toxic because

[00:12:17] You can ehi making

[00:12:18] Humans look that much more

[00:12:19] Unrealistic in a way that's effortless. But that's already happening with Photoshop. So.

[00:12:25] So if you guys noticed that my eye color starts changing to something that's less than human, you might

[00:12:29] Notice attractants

[00:12:31] Going off

[00:12:31] The rails.

[00:12:32] I'm trying to go to, like, feature evolution.

[00:12:34] That's pretty interesting, right? Because, I mean, like

[00:12:37] In real life, like I did my

[00:12:39] Beard, it's quite patchy, white everywhere. And I turn the touch a filter on zoom a little bit to increase the the I'm confused.

[00:12:51] Why would anyone want to do white hair.

[00:12:53] Can someone explain this to me. It makes

[00:12:56] No sense to me if it's just starting to look better

[00:13:01] Than

[00:13:02] That once it's all evenly dispersed. I'll reveal it revealed to the Harp the way. Can you go for it or Harpreet the which will be, I don't know, whichever candles, it's cooler. So I think the

[00:13:14] The image

[00:13:15] Augmentation for for more attractiveness or whatever it might be. That's really interesting one, because I see a future

[00:13:21] Where all of us

[00:13:22] Are speaking through avatars that look similar to us as well. I mean, there's already that capability on iPhone. There's already those types of things. And I don't see, like touching up their appearance appearances

[00:13:32] Being that different from that. I don't see that

[00:13:35] As much of an ethical issue as this being used by someone to impersonate us online or something along those lines. Obviously, that's a lot bigger and a lot scarier thing. My belief

[00:13:46] Is if it if

[00:13:47] If you're doing this with

[00:13:48] Your own image or you're having someone's consent to

[00:13:51] Use their image and their likeness to create these things,

[00:13:54] That's totally fine.

[00:13:55] But once we get into using things like the that person in the video store, [00:14:00] did they get consent from Morgan Freeman to use his face and use his voice in those sense? In those use cases? To me, that is that could be really scary. And we have some good fake detection, but it's a cat and mouse game. Like deep fakes are always going to be

[00:14:15] Like iterating and improving, I would

[00:14:18] Think, to a certain extent faster than we can keep

[00:14:20] Up with with tracking them, because the way that these models are

[00:14:24] Built is that like

[00:14:27] To generate an image

[00:14:28] Or like

[00:14:29] A deep fake, like

[00:14:30] You're using the

[00:14:31] Same tool to validate

[00:14:33] That it is, or to keep increasing its its efficacy or its accuracy of this fake model

[00:14:39] To train the model. The better the ID

[00:14:42] Models are, the better the fake stupe, because it is this feedback loop that's created. So to me, I actually like the idea of being able to have an avatar, have an off day. My my hair is like it was in the video the other day and I can just pop on my avatar. It looks very close to me. It doesn't have to be exactly me. But but that way, you know, like I'm in my PJs and my meeting, it doesn't matter, like looks great. I'm all trimmed up in a suit. I think that's a great

[00:15:07] Use case for this.

[00:15:09] Unfortunately, it does come with

[00:15:10] Those potentially

[00:15:12] Nefarious consequences. Attractiveness is actually such an interesting topic. I was going to sneak in real quick.

[00:15:18] The attractiveness of stuff for people are listening.

[00:15:21] I've actually done a lot of deep planning analysis on

[00:15:23] Attractiveness just

[00:15:25] Long time ago, like, I was really curious

[00:15:27] And interested in it.

[00:15:28] The attractiveness in their bio.

[00:15:31] Yeah, the attractiveness is not a global standard.

[00:15:34] And so what that means,

[00:15:35] If I know I'm speaking in Dublin, I could actually increase my attractiveness in Dublin versus South

[00:15:40] Africa. Like sometimes we

[00:15:41] Think it's a global

[00:15:41] Saturn standard

[00:15:43] Because like the Hollywood effect

[00:15:44] Or something. But it actually gets to do. I need to

[00:15:47] Look more familiar to the demographic where I'm traveling to. I'm going to go speak in Croatia or something like do I want to have hints of that in my avatar like that? That would be something that would it [00:16:00] wouldn't be science fiction. I'd be very, I think, be very straightforward to give hints of that ancestry in

[00:16:04] My face in a way

[00:16:06] That no

[00:16:06] One's they just

[00:16:08] Show up like a man

[00:16:09] Like you definitely look at. They're not going to let me not speak.

[00:16:12] But as far as my face, I don't know.

[00:16:14] I think it's it's funny what

[00:16:16] Technology can allow you to do in ethics. It becomes a very great place to play sometimes. Like sometimes it's obvious.

[00:16:22] Sometimes it's are we allowed

[00:16:23] To look more attractive

[00:16:24] Online? Like, probably

[00:16:27] If I just want to sneak in there fakes and at ethics, I think it was about a year ago I showed up at a company meeting with my CTO as an avatar, and I gave myself a promotion in the and it didn't stick. But like I said, some ethical boundaries say

[00:16:47] That's a good use case, too. I like that

[00:16:48] Next case that you show up as your CEO, angry at your CFO for a wire

[00:16:53] Transfer. That's the next that's the next scenario. I need this

[00:16:57] Wire transfer now. And you yell at them. Yes, that would be right.

[00:17:01] But Serge and he got to be.

[00:17:03] And yeah, that that sounds like

[00:17:07] A very,

[00:17:08] Very devious youthquake, but. I wonder where you're going to get enough

[00:17:13] Material of your

[00:17:14] Ceo talking, I guess, from conferences and things like that even matter to record him in meetings.

[00:17:23] And we do get all the training data from what

[00:17:25] I actually saw, there's there's one

[00:17:26] Company that with like

[00:17:28] 30 seconds of talking, they can produce a

[00:17:30] Pretty good voice augmentation.

[00:17:32] I can't remember that specific company. I was looking into this for a video a while ago, but I was astounded at that.

[00:17:39] The like like the variance in

[00:17:41] Pitch in a very

[00:17:42] Short snippet

[00:17:43] Of talking.

[00:17:43] It's pretty crazy.

[00:17:45] What if I could use all

[00:17:46] Of that of Ken's

[00:17:47] Youtube videos for training Data morph myself into kanji? Would you guys like me better if I was kanji as kanji?

[00:17:53] Yes, like me better. I don't know

[00:17:56] Then any any comments on this. If anybody else has any comments

[00:17:59] Or

[00:18:00] Anything [00:18:00] they want to add in here, definitely go for it.

[00:18:02] I think it's interesting when you start talking about the ethical ideas of this, what exactly

[00:18:09] What's wrong with

[00:18:11] You changing your appearance and playing with your appearance? And is there going to be a single definition

[00:18:17] Of wrong that everyone's

[00:18:18] Going to agree with? And I think that's the biggest problem that we

[00:18:21] Have with capabilities like

[00:18:23] Deep Fix is you can make a really compelling case for being a great thing and you can make a really compelling case for the exact same use case being a bad thing. And so we have these sort of conflicting ideas of ethics that have to be reconciled at some point. So I think, you know, when you say, is this OK, is this an appropriate use? You're not in one of

[00:18:47] Those absolute

[00:18:49] Red flashing lights or green landing lights areas.

[00:18:54] And there's so few

[00:18:56] Really ethical areas that are perfectly well understood and everyone agrees on. So I think that's the most interesting thing about what you

[00:19:04] Bring up is

[00:19:06] I would bet you the majority of people hearing you talk about that particular use case. But again, really there's I wouldn't have even thought about that as an ethical problem where there being an ethical problem. And I would be willing

[00:19:17] To bet that if we went through

[00:19:20] Everybody

[00:19:21] Livestream

[00:19:22] Here, everybody that's watching someone would have a compelling argument for why that wouldn't be an ethical use. And there's this reconciliation process that we need to have as a field for it's appropriate for eighty five percent or 90, 90 percent of people, but we have four percent who are heavily impacted in a negative way.

[00:19:45] And so what do you do then? And I think

[00:19:47] That's the that's the bigger implication it's interesting to think about is how do we create a framework

[00:19:53] To, number one,

[00:19:54] Accommodate the fact that the majority of people want to use this this way, but also at the same time protecting [00:20:00] a small minority

[00:20:01] Who could be hurt by

[00:20:03] That application, just the one that you're using somehow there is probably a small percentage of people that could be hurt by this. So that's the

[00:20:11] Interesting outcome

[00:20:13] Of conversations like this, is we need to have a framework for that. How do we allow users that are going to be, you know, just from the simplest greed perspective, very economically lucrative and in the vast majority of cases harmless. But we have to also balance every other group and one group will

[00:20:32] Potentially be impacted

[00:20:33] Negatively,

[00:20:34] Especially actually ties in nicely, because Mark had a question about

[00:20:38] Ethics as well.

[00:20:40] Russell, do you mind if we kind of pivot

[00:20:41] From the

[00:20:43] Topic itself and kind of take the flow of the conversation

[00:20:45] Towards this ethics

[00:20:47] Direction? Mark, go for it. You had a you had a question that you lined up.

[00:20:50] Yeah. So I recently posted an article about this company called Spot Shooter where essentially they use A.I. to pick up on gun shots and inform the police to help try to reduce crime. The article, a recent release, essentially was saying how the police pressured analysts at this company to basically change the data to help them help them with cases to get their desired results. In addition, they Data like a third party research kind of as a watchdog where essentially a lot of the gunshots, I think 90 percent of the time they were they Damián classify like gun crimes. Right. And so I beg this question. It's like, when does I for me, when does ethics start? Because it's not necessarily like the model specifically. There's like all these other components in the chain of like ideation and scoping all the way to delivering and where they're not even technologists, maybe like marketing or sales. Matt has some really great points actually agree with that. Change our perspective a little bit on it. But essentially it really is. The main question I have is like where [00:22:00] do we start the ethics process? There's some people who are really like on it, more so on the model building. Like, do you classify it correctly? As for me, I'm Rausseo thinking, you know, it's everything, every aspect, even the part that's not necessary. So I'm just curious what people think about that

[00:22:18] Last bit of a question to kind of cut out. But if anybody wants to comment.

[00:22:22] Um, sure, yeah, I, I'd say it starts with the Data generation process and throughout the entire process, I don't think it's like there's there's like one clear division between, you know, here's where the AAFES take over. I think I mean, it's just a question of having other stakeholders, other

[00:22:42] Professionals that are well

[00:22:44] Aware of the implications. Involved in the entire process, and that means like dealing with the Data

[00:22:50] Provenance, interfacing

[00:22:52] With stakeholders that could have troubling like, I don't know, like interest involving that problem, like you said, in this case, the police and then also at the same time informing, you know, a certain angle to domain knowledge, because there's there's a domain knowledge that needed to, of course, make the model perform it. But there's a different domain knowledge that's needed to make the model fair. And most case, in many cases, these things, they don't really they overlap, but they're not. Of course, one thing, if it's actually going to be countering the other, you know, you can make the model more performant and that's going to hurt the fairness and the fairness is going to hurt the performance. So I think it just it can't be disentangled. Like this is where you begin. And I end it just I think it's a messy process, which is why I advocate for having, like, you know, that a platform for A.I. for the future. I imagine it's like a Folan cockpit where there's all kinds of metrics being measured at the same time. It's not like [00:24:00] driven by a single metric as it is in many cases right now. And, you know, and there's different AIs on it.

[00:24:08] It's not just the machine learning

[00:24:09] Engineering, which is kind of the way it is, you know,

[00:24:12] Right now. Yeah, I'd love to hear from from people haven't heard from yet. So Monica, Vivian, Ben,

[00:24:20] Ian, Tom. Or do you

[00:24:22] Guys want to chime in on this. Please let me know. But while I leave you guys, I mean I think it starts just even before Data generation

[00:24:30] Process is just like, is this a

[00:24:32] Thing that we should do with this technology that we have available to us? Is this a problem worth solving? Is this a thing worth building? OK, if we do build this thing, what are some potential repercussions that are negative? Who who might it actually be bad for and think

[00:24:46] It through that way?

[00:24:47] That's kind of my perspective. You start from the beginning, like, is this something that we should actually be doing with this technology?

[00:24:55] It just like that? I think that's the key thing is you have to have a proactive discussion on potential. You haven't even started the project. Like what are the potential issues that could happen if you look in the media, most of the terrible things that have happened with a guy like

[00:25:08] A yet sexist

[00:25:10] Wrestling

[00:25:10] Model's racist

[00:25:12] Gender claspers, like all these things, they fell into the oops

[00:25:15] Category. They weren't

[00:25:15] Proactive. They didn't sit down and talk about what could go wrong, what

[00:25:19] Are the potential. So I like the is being proactive, reactive.

[00:25:22] Don't be reactive. Don't say oops.

[00:25:25] I guess another thing that keeps popping up because I agree with

[00:25:28] This, I was like, who?

[00:25:30] Like I, I personally feel like companies, they have a fiduciary duty just to make money. Right. And so that proactive fees, you can make an argument where it's like, yeah, you'll lose money if you get into the oops category. But with the fast iteration, that typically doesn't happen. So like, should we regulate what regulation would decrease innovation and some other countries may not and they may accelerate innovation. So after a messy balance, as I just I personally don't trust companies to to [00:26:00] make the right decision at at a system level.

[00:26:04] So, yeah, some things do.

[00:26:06] Oh, sorry. Someone else was.

[00:26:08] Ok, I'll jump in for a second. So like twenty years ago when I, when I got into software, we always had to, we were always told to think carefully about what Data that we actually needed to collect from people and only collect the data that you're actually going to use. And so it

[00:26:23] Feels like a while back we were much

[00:26:25] Stricter with ourselves on this. And maybe it's because we've become greedy in the era of big data where we can collect anything that we want and store as much data as we want. That kind of those requirements went away. Maybe we should get back to that. But I like the thought

[00:26:41] That before

[00:26:42] We build models, it's not very difficult to think whether that model is a good thing to be doing or whether the data is going to be difficult or whether there's ethical implications to even getting that data. So I hope that some of that strictness does come back to software

[00:26:56] Development in the age of big data is what I wanted to throw in there.

[00:27:00] Yes. Let's go to, uh, to Ben. And then and if anybody else wants to jump in here, please let me know. Um, raise your hand, uh, and I will add you the list.

[00:27:10] So some things do

[00:27:11] Have regulation

[00:27:12] And protections in place. Like I know for sure they like they've got the four fifths rule. They've got some different things. But the issue is it's not

[00:27:18] All it's not totally inclusive. Right.

[00:27:21] There are plenty of biases out there that are not included in that. Then even with those regulations you run into, is this is this bias in the while discoverable? So if it's not discoverable, like you can have a big entity, a big company that

[00:27:34] Has in this

[00:27:36] Obviously happens today, like you have name brand companies that we've all heard of that have rampant bias internally. And sometimes that's unknown. Sometimes it's not malicious. It's not an. They're not aware of it, it's an auditing thing, so I think a big part of that is raising awareness. But when you're working with these companies, they don't want to share that data because it's a legal risk. So it's

[00:27:58] It's more about that

[00:27:59] Groups and [00:28:00] different people having that that awareness that I want to know what my company's doing like that. That's the question that everyone's asking. I want to know what my companies take, because if I come knocking to ask for your company

[00:28:08] Suing, your lawyers

[00:28:09] Aren't going to like that.

[00:28:10] Do that then and after. Then we'll go to a

[00:28:13] 10 mat then and then

[00:28:15] Anybody else wants to chime in. Please just hit the reaction button. Raise your hand. I love to hear from everybody,

[00:28:21] Then want to take this matter in five hundred miles forward from where we are right now because we're discussing all the easy stuff. We're discussing all the obvious stuff because companies do UPSs and that's yeah, that's forever a technology problem. But there are malicious actors, there are companies and countries. There are regions of the world that don't care, legitimately, have no

[00:28:46] Regulation, never will.

[00:28:47] And even the regulation that they may be putting in place is for some companies and some entities, but not for the government or for other companies. And so we're always going to have bad actors.

[00:28:58] They are going to build this stuff. It's going to get

[00:29:01] Out into the wild. People are going to be able to use it.

[00:29:04] Companies will buy it and it

[00:29:06] Will

[00:29:06] Propagate. There's no way to stop that.

[00:29:08] That's something. Regulation has failed time and time again at trying to reign in the impacts, especially negative impacts of technology. And so I think we have to start from there and just say, look, this is

[00:29:24] Going to get built, companies are going

[00:29:25] To use it, and now we have to create a framework because that's the reality. And I think the majority of regulations. The problem with regulation is it assumes that companies are inherently good. No one will build this if we tell them not to. And none of that's really true.

[00:29:42] So you can't

[00:29:43] You can't start with that. You basically have to assume we've got criminals out there just like we do with everything else. And this is cyber security is the same framework. Assume there are bad actors, assume that they're out there and we have to create a framework to protect people. That's really [00:30:00] where we need to go with this. And the biggest problem that we have with protecting people isn't so much we don't know what's going to happen because you can guess I mean, you look at hiring, you know, there are biased hiring systems out there. You look at credit risk scoring. You know, they're there. I mean, you know where the bad and the greatest impact

[00:30:20] Areas are right now.

[00:30:22] And so we we need to go to those areas and say, look, we have to get a framework that all of us can agree on. And this is the hardest part, because if you look at this from a even like a philosophical standpoint, we've been having this argument for three thousand years. I mean, Socrates wrote it down, is the Theros or Urethras dilemma, where literally he goes through exactly what we are about to go through with ethics. And Jay-Z said it way better. So I don't go through, you know, two hours. He said, Socrates ask is biased

[00:30:58] Because that's

[00:30:59] Where we're at, whose bias is OK and can we get consensus on the biases that we all have? That's going to be incredibly hard, especially in a country like America where our biases go from zero to one hundred and everything in between. We have to come to some sort of consensus.

[00:31:17] That's going to be the biggest piece

[00:31:19] Of the puzzle is until you have a framework that people agree on, you continually have bad actors at the fringes who are going to find ways to manipulate any system that we put in place. And so it's almost down to us as practitioners to be the ones who say, look,

[00:31:38] This this is how

[00:31:39] We're going to start policing ourselves to protect, not to prevent, because we can't do that, but to protect people from the impacts of unfair algorithms. And to Ben's point, we can't detect them. We're not going to know that they're out there, but we can pretty much guess

[00:31:56] Where they are. And so we can be

[00:31:58] Looking in those, quote unquote, [00:32:00] high crime locations like you keep an eye on your banks. Why? Because that's where bank thieves want to go get money. And so you want to keep an eye on the communities

[00:32:10] That have consistently

[00:32:11] Been marginalized, the communities and groups who have consistently been impacted by everything else, because, you know, that's where the impacts are in the Data. And so we have to, as a community, come together and say, look, we are going to keep an eye on these groups because we know where the problems are going to be. And so we're just going to try to protect them.

[00:32:31] When we talk about framework, I just as an example, can conceptualize that is that the framework would be something like when you talk about getting consensus on our biases that like Bill of Rights or the Constitution, if that's what you mean by framework

[00:32:44] Or is there some other example, it's going to have to be. Even more granular than that, because most of what happens online doesn't get covered by the Constitution and it isn't the Constitution is this wonderful document. I love it. I'm glad we built the country around it. But at the same time, it didn't anticipate the planet

[00:33:01] That we are on right now.

[00:33:02] And it's not a framework that could govern effectively in your world where at the end of

[00:33:10] Nations we really are

[00:33:12] We are coming to the end of nations. And so we have to

[00:33:15] Start

[00:33:16] Understanding that the frameworks we come to an agreement on this isn't just I mean, look, I said America is going to be hard. If you can get America right, you can pretty much get the rest of the world to work because whatever gets this diverse group of opinions to come together and agree will probably that process of consensus will probably play fairly well across the world. But that process of consensus

[00:33:41] Is beyond

[00:33:42] Our laws right now. It is beyond getting people on social media to follow hashtags. It is

[00:33:48] Far beyond where

[00:33:50] We have any sort of legal or ethical frameworks that cover when I talk about building a framework for

[00:33:57] What's OK and what isn't.

[00:33:59] It goes back to [00:34:00] what I said very little back in the last question. The legal framework has to allow for things that eighty five percent of the world wants to actually happen while protecting the maybe one percent or half a percent of people who will be impacted negatively by it. So we have to protect them while admitting to the fact that this thing that eighty five percent of the world wants is going to exist, it's going to show up. It will happen now. We have to protect people

[00:34:26] And we can't say, don't do

[00:34:28] It. I'll throw a billion dollar fine at you because Google will laugh at you. You can throw a billion dollars fine at Facebook. They won't care. It's really a OK, it's going to happen. Let's protect

[00:34:39] The people instead of trying

[00:34:41] To trying to do something that really isn't possible and trying to build a framework that's all

[00:34:46] Encompassing. I didn't think so

[00:34:47] Much has a lot to think about this. Excellent point. I guess that can. Let's hear from you then, if you can, Matt. You had your hand up and go. Let me know then after to go Tom shout to everybody else that just joined in. Abby, what's going on. What's going on. Greg, what's going on? My friend Ben, the Seattle Data guy. What's going on, man? If you guys want to chime in on this discussion,

[00:35:09] If you guys

[00:35:10] Want to say something, let me know. But just raise your hand. I'll add you to the queue. Can go for it.

[00:35:16] Yeah, that was a tough one to follow up. I agree with everything he said there, and you said it so eloquently. I think something that that I really want to highlight and it gets a little muddled in these conversations sometimes is that like the the models, the algorithms,

[00:35:31] They're not inherently

[00:35:32] Biased or

[00:35:33] Evil. Right?

[00:35:34] They're just math for the most part. Right. What happens is when we when the data goes

[00:35:38] Into them and when we choose

[00:35:40] A specific dependent variable to evaluate them based on, that's when essentially the bad stuff starts to happen.

[00:35:47] And so the data

[00:35:48] That goes in and what we're trying to predict, those can be looked at almost in isolation of the algorithm. We can say, are these things good to predict? Are these things acceptable to collect and analyze?

[00:35:59] And [00:36:00] this then becomes an ethics question more so than an ethics question. Like these are all ethics questions.

[00:36:07] I mean, because it is in that the machine learning AI domain, that does bring a couple of additional wrinkles.

[00:36:13] I won't completely

[00:36:14] Ignore that fact. But at a high level, even

[00:36:17] People that aren't

[00:36:17] Associated with this domain can understand the challenges that go into these models and come out of these models. And so

[00:36:25] I think that one that's a little bit

[00:36:27] Refreshing, like these things are tools where people can

[00:36:29] Misuse tools. They use them all the time.

[00:36:31] You can use a hammer to build something. You can use a hammer to knock them over the head. And if we're viewing him in that frame, although it might be slightly overly simplistic, it can also help us understand how we can set in place

[00:36:44] The frameworks that Ben was talking about.

[00:36:46] And so I would hope that

[00:36:48] That deconstructs these things for people that do have to make decisions around them. They do have to think about these ethical constraints on a daily basis that aren't Data scientists and engineers, AI researchers, whatever it is. Because to

[00:37:00] Me, that's that's something

[00:37:01] This is a problem we can all all start to tackle. And it seems like a lot of the time it's

[00:37:05] Just shoved on the engineers

[00:37:07] And the data scientist like you're responsible for making this ethical. No, that's that's a conversation for all of us to have. We might be building this and training it, but this is a lot broader conversation than just what we're doing independently.

[00:37:19] Yeah, absolutely. I think, like you said, the algorithms, they're just a series of steps. They're just mad. They're inherently without any judgment, bias or emotion or anything like that. It is the human factor at the front end of it that has the duty, the company question trying to answer the product and about the thing you're trying to do. Yeah, like I. Let's hear

[00:37:42] From Matt.

[00:37:44] You still hear. Yes, Matt. Let's go to Matt and therefore Matt, Tom and Monica, you got some comments in the chat. You want to chime in on that? I'd be happy to hear from you.

[00:37:54] Yes, I guess I was just going to say it was slightly different from

[00:37:59] What [00:38:00] was already said, but, uh, you know,

[00:38:01] Essentially kind of echoing.

[00:38:04] A little bit of what looks like from candidates, ultimately, ethics is really a human problem, right?

[00:38:11] Like a lot

[00:38:12] Of AIs, again, algorithms in math and in the Data you put in and other things like that. But obviously, humans still get in the mix.

[00:38:21] And a lot of those problems and was talking about then

[00:38:26] It's like I don't trust companies to do it. And it's like, OK, but I do

[00:38:30] Trust governments to do it. And ultimately, I think part of being at least American is lack

[00:38:36] Of trusting any authority

[00:38:38] To get it right. And so ultimately, if we do it,

[00:38:46] Then the advice like, hey, let's

[00:38:47] Look into our own company.

[00:38:49] And like, I think there is just something that everyone needs to realize. Like there are going to be times when

[00:38:56] You're at

[00:38:57] A algorithm's machine, learning algorithms are just going

[00:38:59] To get it wrong. And so I think a large part of the conversation is like, how do we build systems in place to fix these errors? I mean, this is a problem as eight year old, just because like, you know, like even before I like just software, you know, like, hey, what happens when Verizon charges you twice as much

[00:39:24] While you call them? I talk to a

[00:39:26] Customer service and get that fixed and hopefully only pay them once. And if if they don't fix the charge or pay you more for, like, fixing the charge and things like that, then I could get angry. Right. And so all of this, again, goes back to the ethics question and so kind of thing. And it's like, yeah, it's like this has to extend out to the whole human experience and how people experience how their interactions with machines.

[00:39:54] Yeah. I mean,

[00:39:55] What's the ethics round

[00:39:57] Like? I feel like ethics

[00:39:59] And morals [00:40:00] is the equivalent, the same problems. Is it different from what's the similarities overlapped with the ethics of guns and gun

[00:40:06] Ownership and things like that. Right.

[00:40:07] This one's a tangible thing and one is tangible. It's interesting stuff to hear from you and then whoever else might from the some great comments here in the chat.

[00:40:17] So this is a ocean

[00:40:20] Liner turn circling around back to something been said way at the beginning about the proactivity of it. And it reminded me of this friend

[00:40:29] That I

[00:40:30] Made on LinkedIn that reached out to me. We even had a call. I just entered your name in our chat, Divya, that she's a lawyer, but they go by a different term out of India. And she was proactively wanting to get involved in ethics. And I was so impressed by that, that she wanted to serve companies to get give them a leg up.

[00:40:51] I had to be more

[00:40:53] Proactive about the ethics with the other two things I wanted to point out is

[00:40:59] It seems like

[00:41:00] It's common for people to flail the arms very

[00:41:03] Quickly. If there's an

[00:41:04] Accusation that there's racism in classification within the job rhythm's, that can be unfair just because of our technology. At times it's quite advanced imaging stuff that can on

[00:41:20] The fly see

[00:41:23] Differences. I'm trying to use general terminology here. Let's say you've got an image and there's a certain

[00:41:31] Zone, but the shades

[00:41:33] That it's just hard to distinguish features in that zone. Now, there's some really great improvements, but they're not ubiquitous yet. And so that's a type of bias that's caused by technological limits. The other bias is something we could deal with more effectively in the Data realm. I think a lot of data scientist or I shouldn't put it on data scientist. A lot of Data organizations forget [00:42:00] to employ the central look at your data set. And then when you bring another new amount of data assets in and augment

[00:42:10] To that what is

[00:42:12] What is happening. Track to see if you're approaching the central limit theorem, using the tools for that and then you can

[00:42:19] At that, but you still have to

[00:42:21] Constantly challenge

[00:42:22] Yourself. Are we

[00:42:23] Broadly sampling enough

[00:42:24] From the population? In other words, what

[00:42:27] Sectors exist that we're not

[00:42:28] Sampling? And these are all ways we can deal with it at a technical level. But I think also

[00:42:36] It it shouldn't mean that we don't put some models into production just because we're not sure we've reached the central limit theorem or not. We just need to ask ourselves or we need to be honest that, hey, this model could have problems because we're still collecting data. Those are my thoughts on this.

[00:42:53] But this boy, I always

[00:42:54] Typically go silent when we start talking ethics because it's it's

[00:42:59] It's messy. Yeah, definitely.

[00:43:01] Tom, thank you so much, Monica, to chime in here.

[00:43:04] Yeah, it's definitely, definitely a spicy topic. I agree with you there, Tom, this whole thing I said in the chat, it reminds me of the health care industry with the different technologies that are being used now to analyze and predict certain diseases. So you need to collect data and pictures in order to train your models. And in some cases, that data might not be considered or even treated as Kii or ajai. So that could potentially lead to a future. Oops. And Russell had asked if there was any way to normalize that data, and I'm sure that there is. But then that could potentially harm your accuracy for your models because you need certain Data elements, though, for example, like your family history increases your chances to get certain cancers. You would need that to help you determine [00:44:00] or predict that information. So I think that's an interesting topic.

[00:44:04] Thank you very much, Monica. So a couple of comments coming in from LinkedIn. One is from Rodney Beard saying that a lot of old school, I had a much greater

[00:44:13] Engagement with ethical debates than the newer,

[00:44:17] Statistically based and afunction approximation type of stuff. Machine learning.

[00:44:21] A lot of work was

[00:44:23] Done exploring logic and ethics and

[00:44:25] Action theory. The newer machine

[00:44:27] Learning stuff would do well to look at the broader literature. I think as a boy and had a comment and perhaps a question here, it says Algorithms are a series of instructions, but the ultimate bias inherent in them is they're created by humans. And whether by error or intent,

[00:44:47] Biases come across.

[00:44:49] How do we keep improving and self-correcting and not just assume this is just hack? How do we ensure people's rights, dignity and privacy are upheld while we go gaga with Data? That is definitely interesting. I like that comment. I like that question as well. Anybody want to speak on that?

[00:45:08] Then they'll go for it. Yeah.

[00:45:10] And time flies like all this. Gray So I started to think how long it's been. So six years ago, seven years ago at HireVue, we had a team of PhD physicists.

[00:45:21] There were building

[00:45:22] Models that we the Data was biased

[00:45:24] There.

[00:45:24] Sexism, racism, ageism. It's guaranteed, it's unconscious bias. And we were coming up with techniques to actively remove adverse impact. So can we take a model to predict the performance output on the candidate, but also remove the bias that transfers racism, sexism, etc.?

[00:45:43] So seven years

[00:45:44] Ago, we were making a lot of progress, a ton of progress. So I presented to the EEOC. We went into detail about what we're doing. Biometrics was a big pioneer there as well. They're doing similar developments. And for the person that mentioned, like looking into classical algorithms and

[00:45:58] In techniques that were used, [00:46:00] we looked

[00:46:01] Into all those techniques. And I give you a sense of they've gone much they've gone much further, faster.

[00:46:06] So I think I've missed it on

[00:46:07] This on this group because one of the best ways to mitigate it is you build up you build models

[00:46:12] That compete. You build a model that

[00:46:14] Predicts the thing you wanted to predict. You build a model that predicts all the things you're the most worried about. And something magical happens as you figure out the features that are driving the bias transfer. And so the classic example of being a resumé name, fraternity, sorority, college or even school, like all that stuff gets nuked immediately. No human expert had to

[00:46:32] Come through and mark that up.

[00:46:33] I was very quickly able to figure out these are the features I'm killing because what was it?

[00:46:37] What is it doing?

[00:46:38] It's going to poison the model. It's able to transfer race. It's going to poison the mouth. That's able to transfer gender and biases. So so sometimes I get a little fresher when people

[00:46:47] Kind of talk about what are we going to

[00:46:49] Do or the world's ending. We can't make progress. We are making lots of

[00:46:52] Progress, but not everyone's making

[00:46:54] Progress. Like some people are still at stage zero.

[00:46:56] They don't need to be. The thing I mentioned,

[00:46:58] Anyone on this call can go do it, go build

[00:46:59] Two models and you'll kill

[00:47:01] The features on the one that

[00:47:02] Transfers race. Is that a

[00:47:04] Would that be big

[00:47:05] Again?

[00:47:05] But again,

[00:47:07] You could do with most models, like if you use some of the classic models, it's a little bit more straightforward. If you're using

[00:47:12] Some of the more exotic trip

[00:47:13] Based models, then really feature feature removals. The big thing, if I have ten features in a competency based model, I can't really throw features away. But if you're dealing with a resume, if

[00:47:23] You're dealing with these more exotic

[00:47:25] Datasets, it's easy to have ten thousand features. If I tell

[00:47:27] You I'm going to go throw a

[00:47:29] Thousand features

[00:47:29] Away, words and mentions

[00:47:31] And different

[00:47:31] Attributes, you can use any model you want.

[00:47:34] You pick your favorite model and I'm just throwing features away. And that's a

[00:47:38] Huge win for people that are concerned about bias. Actually, this

[00:47:41] This conversation is very relevant.

[00:47:42] There's a current thread

[00:47:44] On LinkedIn right

[00:47:45] Now where

[00:47:46] An article is called LinkedIn AIs job matching. I was biased. The company solution Maryi and some are from LinkedIn is actually commenting on this article, asking this individual who's just dumping on them, hey, we want to help.

[00:47:59] And I think

[00:47:59] That's the unfortunate [00:48:00] thing. The last thing I'll say is, like, sometimes companies make mistakes, like. There honestly trying, and they will make

[00:48:05] Mistakes, and it's

[00:48:06] Unfortunate that we burn them at the stake as fast as we

[00:48:08] Can, because here's this poor

[00:48:10] Linkedin engineer saying we would love to hear your opinions and and people say this is terrible. These algorithms aren't up for audit. They're not for audit because you burn them at the stake the moment you find

[00:48:20] Anything wrong

[00:48:21] Rather than us

[00:48:22] Working together. How great would

[00:48:23] Be if we're so transparent. This is what we're trying. What do you think?

[00:48:27] And so just like you messed up. Yeah, it's too bad. But I guess this

[00:48:31] Goes into like, why are we so politicized, divided. Like it's just a general human behavior beyond this. This is a bigger discussion on our follow up question for Ben for that. Yeah. Do you feel like, you know, the big

[00:48:44] Movers

[00:48:45] Like the the companies

[00:48:46] Who have way to

[00:48:48] Change the need? Or do you how do you feel like they have the will

[00:48:51] To to want to work,

[00:48:52] To want to fix this?

[00:48:54] Are they doing enough to say,

[00:48:56] Look, we know this is a problem, we're willing to fix it and create transparency

[00:49:02] To fix this? Do you see enough movement there?

[00:49:05] It's that's tricky

[00:49:06] Because I'm thinking of, like, the

[00:49:08] Whole Timna thing, like there that's super complicated.

[00:49:12] Like, it's it's just it's a mess

[00:49:14] Because if you're if you're talking about things that are antagonistic towards the moneymaking portion of the

[00:49:19] Business, that is problematic

[00:49:22] From a capitalistic perspective.

[00:49:23] So it's the bigger

[00:49:24] Companies have what it takes. I think they're doing useful research. I'm not familiar all the research that they're

[00:49:30] Doing, but I look at

[00:49:31] Companies like biometrics, like biometrics. There's been a thought leader in adverse impact mitigation. They're very they're CEOs, very outspoken about it. I must see the innovation coming from

[00:49:41] Not the bigger companies when it comes to ethics.

[00:49:44] But but, Greg, I. I worry sometimes that maybe the bigger companies, they have massive targets on them. Like how often do they get burned at the stake on a regular basis and how nice would it be as a community if a you guys screwed up? Here's a bunch of professors coming from MIT and [00:50:00]

[00:50:00] Stanford that specialize in

[00:50:01] This, would like to

[00:50:01] Work together.

[00:50:03] Yeah, it humans are so funny that way. Like, I'm the most negative things we can talk about with human behavior, like why are we acting like this? I could also say

[00:50:11] This is why we're so good at innovating. Like if Greg and

[00:50:13] I get in like this fierce competition on innovation, one

[00:50:15] Of us is going to win. That's not a very

[00:50:17] Nice teamwork thing to do. And so these things that are more

[00:50:20] Emotional and negative

[00:50:22] With human spirit,

[00:50:22] They also lead. There's other

[00:50:24] Flavors of them that lead to this competitive

[00:50:26] Innovation, like the hero's journey,

[00:50:29] Like why the hell would a human like climb over a mountain to die to see what's on the other? It doesn't make any sense. But here we are as humans, like, all right, I'm speaking too long. But that's true. It just real clear. Harp green. You and I are both lovers of the sticks. That's right. Yes. It's like organizations need to have a blue hat Data scientists full time blue hat data scientists watching for this kind of crap, not just this kind of crap, but also one that goes around and says, hey, what thirty Data problems that we have, we'll go try to fix them at the source is when you're in the trenches, you're working on your stuff. You don't have time always to get up out of your chair and go, hey, I'm worried about another beer. I'm worried about a dirty Data sources here, etc.. Be so nice. There was a blue hat data scientist in

[00:51:17] The blue hat of the six thinking hats. Definitely take that out of the thinking hat. Right. And that

[00:51:25] Thinking about the thinking, thinking the thing about what mode everyone's in

[00:51:29] Control had organized thinking itself, sets the focus, calls for the use of other hats, monitors and reflects on the thinking process used it had to wear them. Let's hear from you. And then if anybody else wants to speak on this, definitely let me know by reason head. But after Vigne, if nobody else wants to speak on this topic, would Eric's question with a question about sampling which ties back to something Tom was talking about or might even get in here, then Jacob has a question about Data engineering, and I know exactly who I'm going to call on that question then you, as [00:52:00] that has some questions as well. Also shout out to everybody else that you, Mikiko, what's going on right now. Matt Diamond, good to see you guys here.

[00:52:07] Then go for it. I'm just going to I've worked in the same space, the blended and the problem with many, many approaches.

[00:52:14] And I'm not burning

[00:52:15] Anybody at the stake. But this is really indicative

[00:52:18] Of, OK, when

[00:52:19] Do you stop? Because that's the biggest question with ethics is when when does your accountability stop? When you look at a resume, nothing in that resume is a causal feature for employee performance. Nothing, because it's a collection of words and it's an ugly collection of words. But we can make some best

[00:52:36] Guesses, right. We can all agree

[00:52:38] That if you read a resume, you can make a best guess on whether somebody is qualified and worth bringing in. But if you really go down the rabbit hole, you begin to realize

[00:52:45] That people who have

[00:52:46] Been impacted by discrimination their entire career resume

[00:52:50] Is differently. They don't highlight

[00:52:52] Their skills the same way as people

[00:52:55] Who have been supported their entire career.

[00:52:57] And so you get to the point where if you really

[00:52:59] Run down the rabbit hole, how far do you go?

[00:53:02] You build another classifier specifically for residues that are written differently. And this is where I talk about the four percent, you know, because eighty five percent of the people that model that the company's Ben's been pointing out, 85 percent of the people that model works amazingly for. But it continues to impact a small percentage of people who don't write

[00:53:23] Resumes as well, but

[00:53:25] Are as capable. Why don't they write resumes? Well, because the system that was there from time immemorial. But isn't the AIs fault. It's not the data scientists fault either. But that system has changed the way they write resumes, and it does. They don't write resumes the same way. They aren't as

[00:53:42] Confident about a particular

[00:53:44] Capability, even though their skill level with that capability could be exactly the same as someone else who's not experienced that level of discrimination. And so the bar for them to write confidently about it on their resume is higher because of the prior impact. [00:54:00] Now, where does our accountability knowing that because you can run down the rabbit hole and figure it out, knowing that we're as Data scientists, where does our accountability end? Do we have to mitigate that?

[00:54:13] Thank you very

[00:54:13] Much. It's been such a good discussion. Man, if you want to listen back on, doesn't look like anybody else has anything to say on this topic. If you do, please let me know.

[00:54:25] I was just in respond 20 seconds so that Ben brings up an excellent point that

[00:54:30] People write

[00:54:30] Resumes differently and that can have a negative impact that I think the resume should die and go away and it should be replaced with job simulations,

[00:54:37] Which is

[00:54:37] Problematic in itself. But we're all Data scientists. If you can put you through like a coding challenge that is practical, it's scored in real time or like even enter a VR

[00:54:47] Simulation, like we already do that for call centers for some

[00:54:50] Simple jobs. You can go through a job

[00:54:51] Simulation, but as we

[00:54:53] Do, that would be

[00:54:54] Like the Holy Grail, like can you do

[00:54:56] The job? And I don't

[00:54:56] Care about your background or you flunked out of

[00:54:58] School. Can you do the job like so many people could do? Like I work in marketing, like fantastic people could work in marketing.

[00:55:04] You've dropped out of college, but unfortunately, they

[00:55:06] Don't have that line in their resume right longer than

[00:55:09] Anybody else. A lot of good

[00:55:11] Stuff, a lot of good stuff going on in the in the chat as well. But I think we'll just keep a conversation moving just to get everyone's questions.

[00:55:17] Um, so let's

[00:55:18] Go to Eric's question about sampling, but great, great discussions in the chat. And you could figure out what's going on in the chat if you join us every week

[00:55:27] Already in the chat. So I probably have like the lamest question of the night, especially compared to all the previous discussion. Qu.

[00:55:35] So so

[00:55:37] I just a little thing

[00:55:40] I'm looking at Data where

[00:55:41] I'm trying to see what proportion of

[00:55:45] This big old several

[00:55:46] Thousand rows of

[00:55:48] Data. So it's not unbelievably

[00:55:50] Huge, but it's too huge for me to like go through and check

[00:55:53] And see how many of these

[00:55:54] Might be, might not be the condition. And it's an

[00:55:56] Honor and it's just a proportionate thing. And I'm trying to think, [00:56:00]

[00:56:00] Ok, so if I have let's say

[00:56:02] Ten thousand, if I have

[00:56:03] Ten thousand and I'm looking for a proportion and I think the proportion is small,

[00:56:09] Maybe one percent.

[00:56:10] I don't know how many I need to look at before I can say I'm confident of that. And I've tried to Google it. And I

[00:56:15] Know like I've been able to

[00:56:17] Find something

[00:56:17] That I can think through

[00:56:19] With my Thursday or Friday adult brain

[00:56:22] To make sure I'm actually

[00:56:23] Considering the right stuff.

[00:56:25] So that's that's basically my question is like, what am I missing?

[00:56:28] It's good to Tom for this one because you're actually talking about this earlier.

[00:56:33] So, my dear buddy, I was talking, my dear white haired friend, then you it's actually dark here soon. That's turning it light. But would you repeat your question? Because I was like, OK,

[00:56:47] Yeah, I was just saying I have a I have a big set of Data and I need to find out what the proportion of them is that are well, just say failing instead of passing.

[00:56:57] And I think it's a

[00:56:58] Small percentage, like maybe one percentage

[00:57:00] Or one percent.

[00:57:01] And of think, how do I know what my sample size is? Data science one on one I have been stuck on.

[00:57:07] So with this.

[00:57:09] And is this a classification problem?

[00:57:12] No, it's just a good test to see what proportion of the. I'm trying to figure out what it is. Yeah.

[00:57:19] If, if I need to. So if I have and then let's

[00:57:23] Say if I have ten thousand rows of stuff and it's not easy, I can't just like do a pivot table to show me passes or

[00:57:31] Fails.

[00:57:33] And so I need to see what proportion of them are failures. And I don't want to look through all ten thousand of them to get my percentage. So how many do I need to look through before I can say fairly confidently it's one percent.

[00:57:46] How do you

[00:57:46] Qualify? Lots of failures, but I think we're missing. Uh, let's see here. Label value that it was. Is it is it your label column or what you think should be your label column that it says pass [00:58:00] or fail.

[00:58:01] So what? It is right now, as I'm looking at I'm looking at let's see here, I'm looking at like 90 names and then I'm trying and then I'm looking at some other related feel to that, like just another entity.

[00:58:16] But sometimes there

[00:58:16] Might be duplicates and sometimes they might not actually be a duplicate

[00:58:20] Like it might be to people

[00:58:22] Who like to people who both work for the same company. Or it might be to two different people who work for two different companies that have the same name. And I have to figure out,

[00:58:34] Like looking at them, this

[00:58:35] Is something I have to do. I have to eyeball it and make a judgment call as a human being based on the other data that I have,

[00:58:41] Which is why I can't just program an answer.

[00:58:45] Yeah, it sounds like it's a form of dirty Data. Yeah. Yeah, kind of. Yeah. And there's not enough information to discern if maybe these two identical names are two different people because you don't have enough information.

[00:59:01] I have, I have enough information. But like, like if if there's two Eric Simms's that both happen to work for two companies that have the same name and they very well could have the same name because they're talking about an entire country of three hundred thirty million people in 50 different states, there could be two people from a plus coin laundry that are that Eric Sims. Right. Be weird, but it's possible.

[00:59:27] It looks like Marc says he's worked on a problem that's similar to this. Let's hear from Marc. And then also, Rodney might have a problem that he's working on that similar to this as well. Sounds just like a caution test. You just think a few random samples and something like that and then have a null hypothesis and do that.

[00:59:43] But Mark AIs going to sound super scrappy and true startup fashion. But essentially I had a problem where working with our Data again, we had like massive big Data. So we're using like sparks. I could just go through everything. I was like millions of rows, but [01:00:00] essentially our Data sucks. And so when they were classifying the insurance type, they had numerous different types of insurance categories. This may be a little different as you're seeing names that may have a lot more variability. But essentially what I did was I negotiated with the state court. I said this is not going to be perfect, but what you have is utter or is horrible. So I can get you 80 percent there and you just have to accept that 20 percent is going to be wrong somehow and work with them just to say, like, let's just get the ball moving forward. We can figure out later you can start a PAC.

[01:00:37] So got the

[01:00:38] Negotiation down there. So that stakeholder is OK with that. And from there I just chose I just randomly chose five hundred rows and just self labeled it. And so I, I and that could have done something statistically significant. But again startup scrapie Jahshan before get processed out I just self labeled five hundred rows and self and it basically said like all right, cool, this is like a validation set. Right. And then I use regex and some business logic and just created some logic to classify and create some rolls. I ran it on, on all my Data. Right. And then I ran it on my validation set to classifying a fusion matrix saying like, this is what the insurance classifications are, you know, this is what it looks like. And from there I was able to go to the stakeholders, said, hey, I ran this on this validation stuff. I made my biases that I labeled it myself. And no one else is going to do it. If you go label if you want, but you don't wanna spend money on that. I reduced the Noles to less than one percent and I have accuracy f score. I want to do very biased,

[01:01:44] But like this I have

[01:01:46] And we have a baseline quantify that's better than before. And so again, it's not perfect, but you create a process to be wrong. You quantify how wrong you are. So now you have a process to be right further later on.

[01:01:58] But just very briefly, think [01:02:00] about, OK, these may be textures, but that doesn't mean you can't be engineer. You could put two textures together. I've done that many times to help me find dirty Data fix it or feature engineer text before I encode the numbers. We can take it offline like this said in the that I want your hear whoever Harp wants to hear from. But I'm also eager to hear from Rodney.

[01:02:27] Nobody else wants to chime in after Rodney. Let me know Erika LinkedIn to something that hopefully is helpful as just a quick little read on sample size. Fresh to me, my little proportion in various different methods for doing that. But it sounds like Rodney this year from you that Rodney will go to Jacob question on Data engineering,

[01:02:48] Which I

[01:02:49] Will hand that one over to. Uh, and Matt, there's no doubt.

[01:02:53] Bend over, Rodney. Yeah. So it's just like you said, it's just a test of proportions. Right. So.

[01:03:02] The question is, how large a

[01:03:03] Sample

[01:03:04] Size do you need

[01:03:06] In order to be able to test whether the proportion of successes or failures, whichever way you want

[01:03:13] To look at it, equals a

[01:03:15] Particular proportion. So there's basically two ways to do this. You can use the Z score formula for the test of proportions, which is basically

[01:03:25] The difference between some estimated

[01:03:27] Proportion and the hypothesis of proportion divided

[01:03:30] By the square root of the product of

[01:03:33] The hypothesized proportion times one minus

[01:03:36] The hypothesized proportion

[01:03:37] Divided by N, which is going to be a sample size. I love that you just like crank that it's was going to throw that intro, that formula at the top of my head. Right. Right. And then what? Then what you do is so that the numerator that is basically is basically error. Right. So what you do is you set a tolerance on the

[01:03:58] Error and then you [01:04:00] you rearrange that for a given

[01:04:02] Tolerance to back

[01:04:03] Out and write that

[01:04:05] The sample size. And so that's doing it. And then the second way you do it from a statistical perspective is

[01:04:11] As you basically look at the

[01:04:13] Power

[01:04:14] That you need to be

[01:04:15] Able to test the

[01:04:16] Hypothesis with, say,

[01:04:18] 80 percent power. And that's usually so I'm not sure about tools in, say, Python for this. I'm pretty sure there are, but I'm not familiar with them. But there's numerous tools for computing power available and something like that, for example.

[01:04:37] So it goes

[01:04:39] Down

[01:04:40] Relatively well known problem

[01:04:43] In statistics, I think. Yeah. And it's like as you're talking about, it's like, oh yes, of course. Like I've heard these things and learn them, but like, I don't have to do this particular thing on a daily basis at my job. And so I'm like, oh, good thing I got a question for Friday.

[01:04:59] So just to clarify, it sounds like you have like a small set of Data and you're just trying to make this inference look like the larger data that is out there that have yet to look at. I think that the proportion of whatever it is that I'm looking at is less than a threshold that I like.

[01:05:14] Trying to guess how big a sample size, you know, which is precisely, I think, what Eric A.. So, yeah. Yeah. And that's and that's probably the other thing is, I mean, I could

[01:05:24] Like, you know, I mean, just business

[01:05:27] At the time I had like, I can I don't necessarily have to go through and calculate the right sample size. Do every every single little thing like good enough is good enough for it. But it just made me think like, well what if I was dealing with ten million rows rather than ten thousand and what am I what am I going to do? And just trying to think through that because, you know, next week I could be dealing with that instead and just trying to be proactive about it. Yeah.

[01:05:50] So here, I'll give you a link to this thing. It makes this easy. You just plug and chug. It's really easy. Um, so you don't have to do exactly that. Oh, [01:06:00] cool. Let's keep it moving. Data got a question about Data engineering. Well, it's a good thing that Data engineers

[01:06:07] In the building

[01:06:09] Never all trying to learn them Data engineering. So I saw Google cloud course and I'd be confused. I don't lose to take. And I also want to know if it's necessary to have

[01:06:24] Like a drink, remember. Oh, OK.

[01:06:31] So that has to do with the

[01:06:33] Software

[01:06:33] Development lifecycle if it's necessary to

[01:06:36] Help some it

[01:06:37] It's engineer.

[01:06:38] So those are my questions.

[01:06:39] Yeah, definitely.

[01:06:41] Uh, over my friend. Good to have you back and good to hear from you man. Yeah.

[01:06:46] No, I know I just drop off like that, but yeah. Just to give you kind of my perspective on it, I think to start out with when you're looking for like whether you go to high school and so forth, I think usually I always say focus more general first, like in terms of like understand like if you are if you don't understand, like Data engineering as a concept, I focus more on on those processes first, like understanding, like streaming battles, things of that nature, data warehousing, cloud, data warehousing. Kind of the difference is first because specializing I think can maybe rap like one pigeonhole you a little bit, but also like any of those things like Google CERT, I've looked at it. It's very much geared towards Google, right. Like it's like that's what it is. You're not going to learn, I think, how to be a Data engineer. You're going to learn how to work with Google products. And I haven't looked at I don't think I've seen databases. I've seen like Azure. But I think it's a similar story and I think it's more important to be general. And then hopefully you can apply those skills with the different tools that they provide because they all have kind of similar tools across their space. But at the end of the day, I think it's better to kind of learn the general skills, the skill sets, utils, Data pipelines, eltis [01:08:00] things of that nature and then apply that to. Specific tools,

[01:08:04] That's just generally

[01:08:05] Like how I feel about that and then what was the second part of that question? There's one other part now based on

[01:08:12] I'm talking about the detention software. Oh, I'm life-cycle about software development.

[01:08:22] Yeah, no, I would

[01:08:23] Say, you know, there's a lot of aspects that I think, you know, applying the same similar principles like software development lifecycle to is valuable and data engineering. There's a lot of programing involved in in Data engineering. And even if you're not doing a lot of programing, there's often a lot of like psychologic that, you know, I think benefits from similar like mentalities and principles. So I would say, yeah, you're very much I think you need to apply a lot of similar concepts. In fact, even when it comes to, like, Dashboard, I think like even there has some some benefits that you can apply and treat it more like software and not just like like an ad hoc thing. I think that's the big thing about Data engineering is like we're building infrastructure that's meant to last for hopefully a long time and not just doing an analysis that will disappear after it, after it's mainly done. So it is much more like software in that way. Even if you're doing more like, I think low code solutions, I think it's still benefit from having similar testing, similar lifecycle process, because, again, it needs to be around for a long time. It's not just meant to be built and then destroyed or built and looked at and then not looked at again. So I think it is important to it's an important aspect of the work that we do.

[01:09:31] Thank you

[01:09:33] Very much, Ben

[01:09:34] Harp.

[01:09:35] Yeah, so I guess for the first point, you know, whenever anyone asks, you know, like what tools you should use, usually my recommendation is you should figure out what company you want to work for. Figure out what rule or tools they use and learn

[01:09:50] Those just because

[01:09:53] It doesn't serve

[01:09:55] Google or Azure

[01:09:58] Or whatever, they'll pretty much [01:10:00] have similar ecosystems. And the code you end up writing, it's all very similar, but just slightly different. And so just learning one of them will make all the skills transferable.

[01:10:12] It's similar like

[01:10:14] Learning a programing language. Should you learn Python

[01:10:16] Or Ruby, C++, etc.. Like, you know, usually the best answer is if you know where you want to go.

[01:10:25] Choose the right

[01:10:25] Tools that'll make you stand

[01:10:27] Out.

[01:10:27] Go work for that sort company. And I guess the

[01:10:31] Second part of that question is like, I think it's a really good idea to learn the software cycle from my experience, I think for Data engineering. So as far as I know, there's no there's no colleges that offer a Data engineering degree, like it's just such a new field. I think that engineers only really came around as the

[01:10:53] Title seven years ago. I mean, really, it's

[01:10:58] Newer than data science. But I think from my perspective, often it's it's easier to take like a software engineer and teach them Data than it is to take a data scientist and teach them engineering. So, I mean, generally, you're going to do one of the two to create a Data engineer. But I just think a lot of the engineering aspects are a little bit more rigorous and maybe a little less fun and so on.

[01:11:27] And so it's just a matter of kind of learning that software

[01:11:31] Engineering cycle is very useful,

[01:11:33] Very much.

[01:11:34] But anybody else

[01:11:35] Want to chime in on this, Mikiko? And you might have some thoughts with you in the chat here, but I'm not sure if it's a related question or

[01:11:42] I'm going to agree. As a scientist that went over to the engineering side, it is a very, very hard hill to climb very hard. But I think it's worthwhile just depending on your incentives. Like I worked as a data center for years and I got to the point where I was basically had a fork in the road. I could [01:12:00] continue being like a very mediocre data

[01:12:02] Scientist if I would be

[01:12:03] Really honest

[01:12:04] Or

[01:12:05] And mediocre because I just didn't have that love of innovation and to go research questions and kind of be OK with, like, OK, well, you know, 80 percent of research projects failing. Like, I just wasn't down with that. So I chose the engineering side. But I'm going to agree with Matt. It's is a lot more rigorous. So, for example, just taking a very, very simple workload, like we want to create a python package and like, get it production AIs then secure. Right. You have to deal with everything from like, oh, like cloud storage bucket. Do you create and destroy like the IAM provisions, like all this other stuff.

[01:12:40] So like security

[01:12:41] Permissions pipelining robustness.

[01:12:43] What happens if, like predictions go down

[01:12:45] All this nonsense. So I think that's just like something to consider. Yeah, it's definitely like learn principles and concepts and then sort of I would say like don't go after if you do kind of like Chase after one of the solution providers. And that's the kind of keyword is like managed services for everything. I would say pick one and then sort of stick with it and kind of like learn the concepts just because, like you said, you have some kind of like similar offerings. And a lot of employers, they'll say, like you, you know, if you look at the job description, they'll say like familiarity with one of us, Azure Data. But essentially for a lot of company companies, once they adopt a provider, they will just kind of stick with it because it is very, very expensive to, like, move in between DP and Azure. So that's just kind of like I think I think Mårten Benjamin being the experts in this area, I think they totally nailed all those points. And just personal experience wise, it is a very, very, very, very hard move which, you know, make sure you're very informed as to

[01:13:49] Kind of what are the

[01:13:50] Skills and the experiences you need to have and make sure you appropriately kind of set up your learning and upskilling path, because there's a lot of rabbit holes that you could just kind of like get [01:14:00] lost in. And you don't really want that.

[01:14:03] So awesome. Mikiko, thank you so much. Appreciate that. Let's go to the

[01:14:07] Question that

[01:14:09] You have yet.

[01:14:10] Hopefully you guys can hear me.

[01:14:12] Yeah, absolutely clear. Man. Good to see you.

[01:14:13] How are you doing? Good.

[01:14:15] Um, my question is tailored specifically for Can Win and you happy and it's centered around content creation because I believe you guys have skin in the game and have a significant experience like

[01:14:29] Building and on.

[01:14:31] To stand in your brand and just a couple of questions in terms of content creation, I see this this uptick, this upward trend in

[01:14:41] Viewership and engagement

[01:14:42] For social media platforms

[01:14:44] Just in terms of covid. And I was

[01:14:46] Wondering, what

[01:14:47] Do you guys think if this

[01:14:48] Trend is going to continue for the long

[01:14:51] Term? Or do you think like as you go back

[01:14:53] To

[01:14:54] Normalcy and AIs spaces,

[01:14:55] Public spaces and offices sort of open

[01:14:57] Up,

[01:14:58] Are we going to see like a natural decline or decrease

[01:15:01] In this industry?

[01:15:02] Or for someone like who's contemplating coming up, creating their own content and coming up with their own thing? I was curious, like, what do you guys think

[01:15:11] Of this trend? So that's

[01:15:13] Number one. And just went really quickly.

[01:15:16] I was also

[01:15:17] Wondering and I'm curious to know, like for all four of you, like, what

[01:15:21] Is what is this

[01:15:22] Process like on like a day to day basis?

[01:15:25] Like, do you plan

[01:15:26] And

[01:15:27] Schedule your content a week or a month ahead? Or is it like a day to day

[01:15:33] Engagement where like it's driven

[01:15:34] By what

[01:15:36] Problems are, what questions people are coming up, coming to? Or just like what do you think about the trend

[01:15:41] And the process of content creation?

[01:15:44] Yeah, definitely. I'm excited to hear what everybody has to say on this.

[01:15:48] I'll just say

[01:15:48] That I think that when it comes to content creation, you have to think about the effect of the long

[01:15:53] Tail. Right.

[01:15:54] And the more that you can reach down and really hone in on the things that [01:16:00] you are interested in, that that you can truly be authentic, speaking on and and creating content for there's always going to be an audience out there for you. Will it be 10 million people? Will it be one hundred? And how many? 40 something thousand. Fifty something thousand YouTube subscribers or will it be seventy five thousand downloads. Podcasts had hundreds episodes. I don't know. Right. But the Internet makes this possible that there is always going to be some

[01:16:26] Niche that you can serve that will still

[01:16:29] Allow you to create content that you love while being authentic to yourself and doing something unique and original that the Internet just makes that possible. Right. The long pillars of being. At another point, they'll probably come to

[01:16:41] Me later and I'll interject if that's

[01:16:43] What I want to add a third question to that. What do you guys

[01:16:46] Think is like the top

[01:16:47] Five or the deviously talked about topics or in terms of content creation, what are the

[01:16:54] Topics that you feel hasn't

[01:16:56] Been given that much attention, their scope

[01:16:59] In this in a space of this engineer?

[01:17:02] Yeah, that's a good question. I don't have a readily available answer to that. And I think one thing for sure maybe is we just had an hour long discussion on ethics. I don't see many people posting about that or talking about that. And I feel like maybe maybe something I'm not actively seeking out at the moment, but I just don't see any of that. My feet up and up and things like that. Right.

[01:17:22] Yeah, thanks for that. But I mean,

[01:17:24] Mellops and you know this intersection. Right. But I mean, the thing is, like, instead of addressing the five things that you you know, why don't you go and say, OK, what are the things that people aren't talking about that I could talk about, that I can say something valuable and contribute something to the discussion and then let me go talk about it.

[01:17:40] I know that list sounds like a great blog post to me. The five things that people are talking about and the Data science space, like it's thinking about those things, because if you're having those questions, other people are also going to be having those questions.

[01:17:52] And something that that I've always lived

[01:17:56] By is the things that are interesting that I find interesting [01:18:00] are the things that I want to create content about. If I'm finding these interesting, if I'm reading all of the other things I'm listening to or following all the incredible creators that are in the space in this even specific Zoome, there's going to be something that I want to dove into further, that I want to add more context that I want to give my experience on.

[01:18:18] And that's going to be something

[01:18:20] That inevitably someone I would imagine I'm not the only and any scenario would also be interested in in terms of growth and in terms of of the landscape. There's so many forces at play.

[01:18:34] I look at

[01:18:36] Trends in terms Data science machine learning. I they fluctuate almost on a day to day basis. And my viewership, whatever it is, it does correlate pretty happily with those things. And so to me, it's like as long as this field is continuing to grow the after effects that people maybe like doing a lot of learning at home or whatever it might be that might that might be a part of it. But I expect that there's going to be a continued growth in content in the space regardless. I mean, you look at medium, for example, there are some really good articles, but

[01:19:09] It's like overflowing

[01:19:10] With so much content. Now, just be great. And I really think that if you're producing good things and you're creating value and you're sharing it with community and you're engaging in the community, that's one of the key things. That's what helps to establish you are to help

[01:19:27] You to grow if you want

[01:19:28] To go down more of a content creator. Right to me, there's so much value and I like sharing what other people are doing. I find it fascinating. Why wouldn't I want to share what my friends are working on or whatever it might be? And that's how you grow. That's it's not about what you create or that or the other things. It's how you how you find your place in the broader community and the value you can you can provide and provide across the whole thing.

[01:19:54] I don't know exactly

[01:19:55] Where I was going out, but that

[01:19:57] Doesn't mean in can

[01:20:00] Think [01:20:00] of like

[01:20:02] In terms of motivation as well. Like what, what is like what are the primary motivations that you guys have.

[01:20:07] Is it. Is it like

[01:20:08] Financial independence or like community

[01:20:10] Building or cloud

[01:20:12] Or is like a mixture of all those things, like in terms of what lives, you

[01:20:17] Know. So I think I've been pretty vocal about

[01:20:21] This is that,

[01:20:23] For example, I started making YouTube

[01:20:24] Videos because I enjoyed making it, that I really

[01:20:27] Enjoyed that process. And I would have never thought that so many people would have watched them by now.

[01:20:32] Like that blows my

[01:20:33] Mind every day. And I did that because I love that process. And it helped me to learn. It helped me selfishly to improve my ability to speak and to convey information. I think that that's something that's really important. And selfishly, I have the podcast which is continuing to grow, and I talk to incredibly intelligent people like biweekly.

[01:20:53] And I get to ask

[01:20:54] Them questions that I'm fascinated and I get to learn and I get to push my learning

[01:20:58] Forward through these mediums.

[01:21:00] And the beautiful thing is I get to share the learnings that I have with everyone else

[01:21:04] Who's tuning it. So this selfish

[01:21:05] Pursuit are somewhat selfish. Pursuit of information on my end leads to broader information and sharing throughout a larger community.

[01:21:13] And I think that that I don't know if

[01:21:15] That's the motivation for everyone. I think obviously I've made some reasonable income. I've gained more financial independence. I've gained a following. But to me, it's always been about that pursuit of of of knowledge and interests.

[01:21:29] And like the more

[01:21:30] People who are following

[01:21:31] You, the more interest

[01:21:33] You have, the more like different questions you get, the

[01:21:36] More information

[01:21:37] You have, the process, the more different directions your brain can go to to uncover new questions that you might not have previously thought of on your own. So it's this it's the development of this beautiful

[01:21:47] Ecosystem that,

[01:21:49] At least for me, makes me keep wanting to do it and keep diving

[01:21:51] In further. Absolute love

[01:21:53] That. I mean, I also like pecans point. The best way to learn something is by teaching it. If there's something that I [01:22:00] don't understand, OK, I'm just

[01:22:01] Going to teach it right. Like, OK,

[01:22:02] I kind of need to get great deep learning, especially the new job and I don't know much about it. So let me twenty one days of deep learning

[01:22:10] And just do that

[01:22:11] And hold myself accountable to the entire public that I'm going to learn this thing and share that journey with you guys. And I mean in terms of motivation, why do this dude like, uh, we just see how far I could push the boundaries. How big could I get this thing? How, how

[01:22:25] Who can I get on that? I you know, people

[01:22:27] Say that I can't go going after that guy. Let me just try it. Right. Because I'm trying to be like I'm trying to be Joe Rogan. That's my competition. Like, I mean, just not these other. That's a good question. I mean, John Crohn, my home, a good friend, but I compete with my friend every day. Sounds great podcast. But Joe Rogan. Bill, you James out here, you know that. That's what I'm envisioning. I'm just seeing if I get that big and I appreciate that. Can I do something that massive? Um, let's go to Benjamin.

[01:22:55] Yeah, no, I just I guess to poke poking a little bit as I mean somewhat of a content creator at least on YouTube more recently, but on medium now it feels like forever. I think it's like three or four years. I think what's what's like interesting. Like as you're trying to like nesh down and things of that nature. I think one thing, because I think Ken brought this up earlier, like, you know, there's like so many articles, I don't know how many articles are posted per day on medium, but it's, you know, ridiculous. No, I think medium has done something similar in terms of like how the like very like what we call digital cameras on phones have also taught us something very important, which is we all suck at taking pictures. And if you're able to take very good pictures and which is a very hard skill to get you, you know, it's a very valuable skill in the same way, there's a ton of people producing content, but producing content is such a hard skill to really get good at and like to find a nation that if you can do it well, I think you're going to stick out. I think that's always something I always think about.

[01:23:51] It's like, okay, how do I just, you know, if I just keep putting out good content, I'll stick out and all the other content that maybe is not as good will just make my content look better. And [01:24:00] I put out terrible content to all the time. YouTube loves letting me know that when it's like, oh, this video is doing nine out of ten compared to your last videos, it's like, okay, well just going to stop making videos. No, but yeah, that's just generally I think like my, my view of it as far as planning, I know I should be like do a much better job, but like content planning. I like try to like make these plans like oh I'm make these videos and then, you know, between working full time and consulting, it's like, OK, well now I got to did. These videos and focus on that, but, yeah, I think I think content planning would be very helpful just in terms like long term strategy. I think that's definitely something I would like to pick up more. That's like doing things like the 21 days of deep learning, things like that would be, I think, a good, good habit to start building.

[01:24:46] So, yeah, I missed that. The second part of that, in terms of content planning, I guess they could talk about contemplating two ways and planning in terms of planning what I'm posting so I can pay for that. And I'll just post on on Kenber and like I mean, I'll post and stuff like that. In terms of actual ideas.

[01:25:02] I've got an idea right here.

[01:25:04] I just write ideas of things I want to create and just do that. And then in terms of when it happens, like I'll just if I'm just inspired in the moment, like I was to do this and just

[01:25:14] Create a ton of content, like I've

[01:25:15] Got more content than then I know how to release these little infographics and things like that. Like, I've got so much stuff. It's ridiculous. Um, so I mean, my new strategy is going to be OK. I'm just pushing only, um, stuff from the podcast, my, my podcast page instead of my personal pages because I'm going to do a hundred and sixty hours of content. That's what I've got planned. That's what I've got to schedule out this weekend is one piece of content every hour for one hundred sixty hours, which is number of hours in a week and do that entirely through my @TheArtistsOfDataScience page. Um, let's go to Ken and Mark and Vivian.

[01:25:51] Yeah, I

[01:25:51] Just wanted to highlight

[01:25:52] Something that that Ben said. I you know,

[01:25:55] That I don't think we actually met in person or like face to face, but we've exchanged [01:26:00] some emails, excited to talk more to you.

[01:26:01] But you don't necessarily

[01:26:03] Have to stand out

[01:26:05] Because of the

[01:26:06] Topic area. You can stand out

[01:26:08] By having a different take.

[01:26:10] Let's say it's, you know, one of the one of the ways I differentiated my YouTube channel from the onset is that I made

[01:26:17] What I thought to be more

[01:26:18] Cinematic videos around Data science concepts. They're incredible creators

[01:26:23] Like Chris Rock,

[01:26:26] Code Basics and quite a few others that are doing similar things. And they're competing in different ways. So Chris produces a video every single day. I knew I

[01:26:34] Could not do that. I knew that I

[01:26:35] Could could

[01:26:36] Could not create a niche for

[01:26:37] Myself based on volume.

[01:26:39] And I thought, OK, if if

[01:26:42] If it was pure subject

[01:26:43] Area knowledge, like

[01:26:44] I think I'm a decent data scientist, but it's not like I'm in like an absurd expert in any of these specific niches. So why not leverage some creativity? Why not use like some stupid comedy or what I believe to be funny, to make this interesting or or go on some of these different themes? And then the last thing is another guy who who I really like up to, a good friend of mine,

[01:27:06] The Data professor, he creates

[01:27:08] These beautiful Hendron infographics and he shares them. Right. That's not something I've really seen too many people do. And that's helped him, aside from his channel, to really grow and expand beyond what he was doing before. And so I think just framing the information in a

[01:27:22] Different way, making it

[01:27:24] More digestible, even if it's the same stuff, and giving credit to people who up by then who are who are doing similar

[01:27:32] Things like

[01:27:33] That also creates

[01:27:34] Value. And if you can bring

[01:27:35] In another additional

[01:27:37] Skill set, great.

[01:27:38] If you can look at it or shared in a different way, I have nothing to add on the organization I like find my podcast episodes. I would just like make videos. I've just got over. I do one a week on this huge list that I go through and I do some polling. I need to get better at the system as it is.

[01:27:55] Yeah, yeah. Yeah definitely.

[01:28:00] Yeah [01:28:00] definitely. Yeah I, I'm by no means no big name like like Ken Harpreet but definitely you.

[01:28:09] I follow,

[01:28:10] I follow you as well. I follow most of the people if not all on this panel. They're not.

[01:28:17] I still feel small and I like, I like, like the small mentality but I think it's like I'm growing and and like I like that aspect. Like it's the hustle component for me. The key driver for for doing this is like, is it fun? Am I creating content? That's fun for me because I recently did a whole analysis on my on my LinkedIn post and it's gaming. And I actually tried a whole like specific content strategy and it bonds completely and it showed up in my Data like all my views engagement. Shrock is just the strategy didn't work, but I kept waiting because I think at that point I just gave up because, like, this isn't working for me more. I don't have the same views, but because it was so fun just to create content, it's I kept on going and now my numbers are going back up. And so I think it's key things like can you create content? That's fine because they were trying to write for other people and it doesn't allow it being fun. It's going to show up in your content to suck on all ends. Right? I think that's what happened to my my campaign. I was writing. I think that was having fun with the other. The other aspect is consistency. So I, I when I first started, I was going like. Twice a week, just doing it consistently, and now I approach every single weekday and now it's a habit. If I don't post in the day, something feels off and it's taking a thought out there.

[01:29:42] And then the other aspect I really care about now, as I told this page from Gary V, who I think is just really good at being a content creator, Hayama Love, and I personally love them, but essentially you create key content. So like so like you have your podcasts or your videos. For me, [01:30:00] I'm trying to make my key content being like Python tutorials for like how how to you like analysis and data science. And I essentially like break my content, you know, like a burnt I burn up to that. So like I'll post about my proccess and then released finally it builds up momentum around it or I'll do like a podcast or live stream of someone and I have that content. Ask them if I have permission to use it and now I can break that up in different pieces and share that later on. And so you get these large pieces of content and you break into smaller ones and just have consistency day in and day out. And what I've noticed is that now with the consistency earlier, I was getting maybe like ten dollars a week, fifty. And now I get like one hundred fifty dollars a week and it's going to grow hopefully. And so that's, that's the main thing. How fine do it. Consistently cool projects or main piece of content. Break it apart to get around. Got to just

[01:30:51] Consistently posting things that you love and can talk about

[01:30:55] In some fashion. Yeah.

[01:30:56] And then oh that's another thing is like also find a niche. So I think, I think and I'm curious I think but I don't think it's right by finding the right topic to talk about is that you create your community around a topic they care about. And so you choose a topic that you find instead of finding the right topic, you find the right community to go the route that gotta

[01:31:21] Go for it. I guess that I was thinking about how I get these questions from people. A lot of like and we've all talked about this like, how do I get started? Like, how did you, like, do this? And so I'm not some big content creator, but I feel like this applies to this kind of principle. Applies to anything is like that.

[01:31:41] You just start like and

[01:31:42] It's OK to just start and like make kind of crappy things at first and like that's fine. Like you just do it like that. I'm I'm convinced that the biggest difference between, like, me being employed as a data scientist and somebody who's just starting out is that I just it's time like that. I just like put in a [01:32:00] certain amount of time and like.

[01:32:01] Sure.

[01:32:02] Like natural ability and talent and things like that can come into it. But like 80 percent of it is just like putting in time and like learning step by step. And this idea of like finding your nation and the way that you plan content creation like those are all like great things. Like I'm not trying to dismiss anybody saying that like they're great ideas and great things. But if you're not, like, first just putting in time and like just doing it, like, that's going to be what gets you 80 percent there and then trying to, like, do these things to like find a niche or find, you know, how many times should it be posting? Like, if I include certain keywords, if I do this or that, like that's like the twenty percent of like tweaking to try to like increase your performance up at the top, you know. But really what will get you 80 percent there is just to do it, to start to put

[01:32:51] In that time and like

[01:32:52] The rest of the path will

[01:32:54] Reveal itself like as

[01:32:56] You go on.

[01:32:57] Yeah. Thanks for that point. I usually like overthink a lot to the point of, like, perfection before even posting anything that you're going to

[01:33:06] Do like this. I just I, I think yeah.

[01:33:09] Maybe I'll have to change that really doesn't matter as long as you're consistently posting. Awesome.

[01:33:16] And maybe, maybe I

[01:33:17] Can try to abstractly connect Vivian's idea with something you can put in here. I mean, Benesch of one.

[01:33:24] Right.

[01:33:24] Those that are trying to say find your niche, just be your own mind issues, just being Harpreet Sahota because nobody can compete with me on being me. Impossible for anyone to try to do, which is related to what can I say here? It's more stable to build a niche around how you talk about topics compared to specific topics. And if I can quote and I've already gone here when he's talking about becoming the best in the world and specific knowledge, he says when you're searching for what to do and in this case, the what to do here is content creation. You have two different foci you want to keep in mind at all points in time. One is going to be the best at what you do.

[01:33:58] And next is what you [01:34:00] do is flexible

[01:34:01] So that you become the best at it. Right. And then you get to a place where you're comfortable at a very comfortable place where you're like, yes, this is something amazing that still be authentic to who I am. And it's not like an overnight discovery or anything. You it's a duty you keep creating. And I mean, I still haven't figured my shit out yet. I don't I don't, you know, like I mean, he's got his

[01:34:23] Like like

[01:34:24] Who has the game figured out yet.

[01:34:28] You guys have

[01:34:30] One. One last thing I'd like to say is from the content creation part of my brain, not my full time work or any of the other stuff, like what is the goal

[01:34:38] Of of creating information online? It's to

[01:34:41] Create value,

[01:34:42] But it's also to get to a place where people

[01:34:45] Listen because it's

[01:34:46] Using it, not necessarily because of what you're

[01:34:50] Saying. And you get to that point because what you're saying creates value. But I would love to be able to a place to be in a place where one day I could make any video that I wanted that was as interesting as possible to me. And people would be interested

[01:35:02] In it because I

[01:35:03] Created it.

[01:35:04] Right. That that is like the the personal branding Mecca. And I think that that but there are some challenges with it. It's hard to differentiate

[01:35:12] Your life, your personal life, with your public life and some of those senses. But but like if I was

[01:35:17] Getting paid to do whatever

[01:35:19] I wanted to do and create whatever I wanted to create every single day, that to me seems like a pretty incredible

[01:35:23] Existence because you're like

[01:35:25] Living what you what you believe and you want to create. And so I would start thinking

[01:35:30] About it is like, what would I like to

[01:35:31] Be making

[01:35:31] Every day? What would I like

[01:35:33] To be producing every day? And if you start

[01:35:35] Doing that and you

[01:35:36] Start growing because of it, you're going to be really motivated to do that, to continue to make it. And you're also going to be able to create a beautiful following that

[01:35:44] Likes you for your value,

[01:35:45] But also like you

[01:35:46] For the

[01:35:47] Very unique and individual things that you bring,

[01:35:50] Like you for for being

[01:35:51] Bring Harpreet. They like me for being candid and the little weird idiosyncrasies that we that we both can occasionally breath.

[01:35:58] Yeah. Can I look for biases of [01:36:00]

[01:36:00] Speaking of papayas, somebody commented here on LinkedIn. I enjoyed Tannen's Papy Q&A. None of the other Data tubers that I follow do things like that as being authentic to them. So you had your you had your hand up as well. Great LinkedIn you haven't really heard from my friend.

[01:36:15] Even though I was going to say to Persia at the end of the day,

[01:36:19] It's about you

[01:36:20] Projecting yourself into the future and kind of kind of looking back, asking yourself how how did I get there? How do I want to get there? Right. So if you see yourself as someone who is knowledgeable in a certain subject, whatever the subject is, what the Internet has allowed us to do, as Ken was saying, is to talk about it and to showcase our journey or talk about our journey. I think people gain followers because they can relate. They can relate somehow, whether they're also experiencing the same thing or they would like to adventure in the same thing. So that's what I describe as a follower. Right. If you think about the biggest soccer superstar right now, AIs followers, not because they think they can be soccer superstars, too, which is Ronaldo, by the way, from Portugal, is because they relate to where he comes from. Right. So they so Ronaldo from Portugal, Portugal, Mikiko is probably the most followed soccer star and on Instagram and you know, it it's all about creating transparency about your journey. And that's what makes the Internet so, so beautiful. And you're you're being vulnerable by sharing what you learned along your path and you're consistently doing so. And with time it builds up. It creates this flywheel where more and more people will take

[01:37:57] Notice into your

[01:37:59] Progress. [01:38:00] And even though people may not relate to what you're trying to achieve, but they relate to you making progress that will that may give them motivation to do their own thing or to hop on their own journey. Right. So there are always there's always a benefit to to to that. So the most important thing is and I think people said it already is to start, as Vivian said, start somewhere and then keep going. If I could give a confession here, I don't consider myself a content creator on LinkedIn. To me, a content creator is the Harp that can the top of the world. We have like this this this mechanism, like whether it's on YouTube or this platform, the podcast and things like that, for me, I'm a vehicle and I do it with a purpose. I like to connect with people who know

[01:38:52] Stuff, who know things,

[01:38:54] And with that I like to bring my special set of skills or knowledge to make sure the I help them achieve something right. So I'm a connector. I've always been good at that. And by tapping into a quality pipeline of things that is that are going on, I just selected one subject, which is my realm, which is huge, and I just talk about things and my way of understanding it. And with that, people come to me to ask for ideas and I help them along the way. So my end point is to talk to a lot of startup founders and help them along that journey. Another point that I have personally is to enhance my brand where I can feel the freedom to put a price on my head for either consultation from a business, from a person who's from the business, or put a price on my head in terms of getting hired by anyone. Right. So, again, this is going to be depending on what [01:40:00] you want to do or to what level you want to become a content creator. Somebody will post on a daily basis like myself on LinkedIn. You want to call the content creator? Yeah, I can't I don't know. I don't see myself as content creator.

[01:40:12] I'm just a vehicle, a spread of news or a spread of news in my own way of understanding it. But I know it does generate value. I'm not saying I'm not creating value. I know it does. Like some people find it valuable. But but it's up to you to to to Deep Dove. Do you want to be the the person who does it well through medium go forward if you want to do it for you to go for it. But at the end of the day, you have to have the will to keep going. You have to have the will to to know that it doesn't you don't find success at day one. Right. You have to keep going. So your first YouTube videos will suck or your your next one hundred videos will suck. And then next thing you know, that one trigger will send you to the moon and people will constantly go back to the older videos because they know, hey, this guy has been producing for so long. So at the end of the day, it's what you do on a daily basis that will pile up to that that compounding effect. So so, yes, there's no good way, bad way of doing it. So just start.

[01:41:19] So Greg claims to not be content creator. LinkedIn AIs spotlight. Everyone here look on the bottom right. That's the LinkedIn top voice award.

[01:41:31] So I don't know. You aspire to be content creators,

[01:41:34] Though, do you aspire to it? If they're not, do

[01:41:37] I do I had a I think of a gig. You did invite me to a podcast Friday and we talked about that. I don't know if you remember came to but. Oh, I remember this one. You're right. And I know you're waiting for me. I need to go to you for that. I haven't had the courage to start. So I told Ken during that broadcast. Right. That I want to start this YouTube channel with that one where I invite [01:42:00] startups to discuss. So to answer your question, long story short, is yes, I want to do so. Time is my biggest enemy. It's a matter of really understanding what I want to do and also doing something that nobody else is doing out there. So I want to make sure I understand what what it means. Will every startup founders want to

[01:42:21] Be

[01:42:22] Present in my podcast to talk about this? A good source? I highly doubt it. Right, because of the the landscape. Right. They have to protect their their secret source. So I'm still munching on that. So I still have you in my mind, Ken. So I think that for that to he's my go to mentor for starting a good YouTube channel.

[01:42:39] So let's hear

[01:42:40] From Matt Diamond here, that after Matt Damon, Cam's got an interesting question and then Africanness question. We will wrap it up,

[01:42:46] Go for it. He it's just

[01:42:47] A question along

[01:42:49] The lines of what Gregersen? Is there

[01:42:50] Any motivation

[01:42:52] That that you

[01:42:53] Change when you're putting content out there? Like what's is there a line between authenticity and how will you change how something is described to make it resonate with a wider audience? Or is that not even a consideration with

[01:43:09] Putting a video or a post out there? Or am I just overcomplicating this? Absolutely. You're right. So so my purpose is to simplify it in my own layman's terms, is to take something that may have taken me an hour to read and to quickly summarize it in the first three lines on the LinkedIn post that would make somebody want to know more or at least make a short summary on the LinkedIn post that somebody may not need to open the appended dock to Deep Dove or if they're interested more. They have a pretty good idea of what's in there based on my understanding of the content. And when we with that, some people may come in and say, your summary sucks

[01:43:53] Or your summary helped. Regardless, I

[01:43:55] Learn from that and I get better. At the end of the day, I help myself [01:44:00] to better communicate, to read faster, I help myself to better write summaries, etc., etc. because I'm aware that people don't have time and there's so much content out there that you need an optimal way of consuming these things and even filtering out the things that are not necessary. So you're absolutely right. I do it with a purpose of putting my own spin. If it's not my own spin, I always make sure I quote it because it helps me a lot. And I. Or ahead of time, I will ask somebody here like that, do what can I do me give me permission to share that with the audience and I know them that we a couple of people in this audience right now are content that I'm just passing on to the bigger audience.

[01:44:45] Yeah, that's that's super helpful.

[01:44:46] I do when

[01:44:47] I follow Data people, not anybody here, I mean that I feel like there's just a bunch of little dopamine nuggets

[01:44:54] That that are nice in the moment, but they can't

[01:44:58] Be contextualized for the problem or the issue that I'm really deeply engaged with at the time. So great stuff like that. Super helpful. Awesome.

[01:45:06] Read on, guys. Excellent.

[01:45:08] Excellent discussion. Depositary. Thank you very much. Hopefully, hopefully this gets you going man. Let's see some stuff. You can go for it.

[01:45:16] Yes. I have one last question to end on.

[01:45:20] And so that is it's completely hypothetical.

[01:45:25] But if a data

[01:45:26] Scientist were to be a

[01:45:27] Protagonist in a movie, what genre would that movie have

[01:45:31] To be in order for you to

[01:45:33] Be willing to watch it? When we have drama, we have Mr. Holmes, we have horror thriller, whatever it is.

[01:45:38] So show us.

[01:45:39] So that's a mystery genre mystery.

[01:45:43] Ichigo, go for I'll play the lead.

[01:45:45] I will make you go, go, go. You have the art hand

[01:45:48] Charts also fine for this one.

[01:45:49] A Sorry guys, I was eating and also I was listening to all you because I don't like listening myself sometimes actually though the longer so my

[01:45:59] Earliest [01:46:00] like kind of

[01:46:01] Connecting with the science was actually the show numbers. So it's, it was a crime murder mystery and it was on CBS, I think for like seven seasons. And what was great was essentially it was like this older brother is a police officer. The younger brother is like, you know, the smarty pants who I like age eight or ten or whatever, guys PhD and some long lines. And he would provide all these different little obviously show was like solving a crime based off of like a principle. But then Wolfram Mathematica at the time had provided like a website that would show you demos of each of the simulations. So I'm like and I feel like if I see a data scientist and something like, I don't know, like the Hallmark Channel or something on a romance, I'd be horribly disappointed because let's be honest with them, it would be them coming up with a spreadsheet going like, OK, let me see if I can input, like, all my sort of wants and and score them. And then coming up with a bunch of fake profiles and then like cat fishing people to see how they could better, like, tailor the content, which someone actually did write a book on. So that happened in real life. So, yeah, crime murder mysteries. That would be my jam. What book was that? Oh, my God, I don't remember. But it was basically it was pink and it was like love in numbers or something. And it was by this like author who is a professor of presumably either political science or something like that. So basically she was like, oh, I'll just use numbers to solve this because she's like, realistically, how many eligible? Also, do you like Jewish men of a certain age

[01:47:31] That are single

[01:47:32] And willing to mingle, but at the same time are in it for the long haul in New York? So I think it was something along those lines. If you look up Data numbers, dating spreadsheet, cat fishing. Yeah, she did touch on it, too.

[01:47:44] So that was really cool. So there's that. I don't know if he has seen this TV show. It's Jack Ryan, Tom Clancy's Jack Ryan. Jim from the office is in there. Jim Kazinsky, a nice nymphet was actual real name is, but I don't think John Krasinski. [01:48:00] So I'm pretty sure he's a data scientist because there are some episodes where he was talking about writing custom SQL queries and making

[01:48:05] A financial analyst.

[01:48:07] Yep, right. Yeah, I saw that.

[01:48:09] And so, I mean, before I get to work, I

[01:48:12] Just want to chime in on this idea

[01:48:13] Was on top of my head. I would make it like a romantic comedy. There would be something like a Data scientist who who works out of like Match.com but gets disenfranchized by the the corporate thing and then goes off and just starts matchmaking himself and doing some interesting stuff.

[01:48:29] Like he'd be like

[01:48:30] Rich, but like a data scientist would be a rom com. He's wrong. Comes across.

[01:48:34] I love I was thinking that too. Like a like a bachelor bachelorette and then having like the data scientist trying to explain what he does and just for everybody to know what he likes to do in your spare time, create spreadsheets.

[01:48:50] There's actually not that I watch The Bachelor,

[01:48:53] But my wife used to make me watch

[01:48:55] It. But there's one season where there's actually a data scientist who's describing his job was like, oh, he's a scientist.

[01:49:01] It's funny how

[01:49:02] This go to you. And then let's hear from from Mark.

[01:49:05] I was just going to say quick, one of my wife's friends was trying to explain to others

[01:49:12] What I do.

[01:49:13] It's he's like Chandler from friends. You can't really know what happened to home. That was the most hilarious explanation.

[01:49:21] What Chandler was a data scientist. His job was statistical analysis and data reconfiguration. Using these markets go. You and then Greg,

[01:49:31] I was about to say, I think the two options for me is one is a manga, manga, whatever it is, and specifically that character is like I know this exact calculation for this reason why I stood right here, and that's why I'll defeat you, like all the calculations in the head, can just defeat you with just logic by itself or like Chickamaw from Naruto. If you're not here to fan, the other one on the more grittier side would be a cyberpunk thriller. So essentially [01:50:00] corporations have taken over the Facebook, the Googles, and it's up to our fearless Data scientists to crack the code and understand the Data to stop them from ruining the world.

[01:50:11] So, so good stuff, man. So Greg is an actual science fiction writer or you wanted to be a writer. So this opportunity man.

[01:50:20] But I'd like to be a one one day. Right. And and and by the way, I did talk to a couple of people who who actually want to automate that, write the script, writing, using your AIs. So I've been pretty excited about that, talking to these folks and what they're doing in this industry. So I'm a big fan of sci fi. So one thing I always think about is this. This world where companies or I've been taken over by a guy right there, they're telling us they're creating the balance between supply and demand in terms of who needs a job, who doesn't, controlling whether you're fit for joining a company which are, again, the board of directors or just entities who are kind of telling you what to do and things like that. And it's scoring you whether you qualify. So there's no such thing as an interview or whatever. So so then you have a group of of of people who are trying to fight this system. So I you know, whether they're data scientists or trying to understand what kind of logic this these entities are controlling the the big things are doing to us and how that affects society. So I always think about about those things. If you follow Prime, there's this show called Electric Dreams. I am hoping they can do season two. But Electric Dreams is a series of each episode isn't connected to the other. So there's a different actor set of actors for each episode, their own Data each episode is its own story. And it's [01:52:00] kind of like a twisted short story, but is super set in the future. And they're exploring human societal issues where, you know, being a therapist might be seen as a disease or being racist might be seen as a disease and how to surface that, et cetera, et cetera. So it explores a lot. I like those kind of dark, gritty future sci fi

[01:52:28] Settings right there. Sounds like black mirror. Yeah.

[01:52:31] Yes, exactly. That's what I

[01:52:33] Hear. Yeah. Yeah.

[01:52:34] Like LinkedIn says, it's the anthology type. Mikiko says, you can argue that any movie

[01:52:40] Is that any movie

[01:52:41] About surveillance is basically a fictional simulation. I think there's a movie like Eagly with Qalibaf, that's all. Or something like that. Oh yeah. That was good. Yeah. Armonica. What kind of movie would you uh what kind of movie to envision

[01:52:53] With the data scientist. Oh, I was just,

[01:52:54] I was just thinking about the, the romance, the like the reality TV. Yeah. So um I think that would be so hilarious.

[01:53:05] Wrestle anybody else

[01:53:06] Now. Thank you for entertaining me.

[01:53:08] I think I'm

[01:53:09] Planning on making the cities.

[01:53:10] Yeah. So that's what comics that you

[01:53:14] Know, I

[01:53:14] Just, I just think it's fun to have these high level, high level conversations.

[01:53:19] I, I personally like

[01:53:20] The sci fi angle, like aliens visit and we have to decode all their signals and whatever it might be. That could be really a really interesting thing and like a perfect use case for like where Data scientists I think would probably

[01:53:33] Stand out

[01:53:34] Like a cross between. You could bring in linguistics, you could bring in like movement patterns. And if they are anomaly detection, a lot of really interesting

[01:53:42] Stuff on this. You like arrival then? I did quite

[01:53:46] A bit

[01:53:47] For that. Yeah. With the linguistic stuff. I don't know how she translated that alien language, but it was a cool movie.

[01:53:53] But that's OK. Yeah, I agree. Like I agree with a Data scientist

[01:53:57] Creates like some type of glasses that allows them to see different [01:54:00] possible outcomes from every decision

[01:54:02] That he makes. Right. So he sees

[01:54:03] Like probability distribution of the future and is able to act accordingly, but then makes a wrong turn and things go crazy.

[01:54:09] I don't know. I'm just making up my head.

[01:54:13] That'll be interesting to look like there are no other questions, no other movies that people want to see. That is that that's great. Great wrap up, guys. Thank you so much for joining in over two hours. This awesome man hopefully has a chance to actually tune in to the podcast or release an episode today with somebody I think that we all know. Jonathan Petyr, he comes by. Gives us his testimony, Jonathan Tensorflow, you guys probably know from LinkedIn he's at Kuga, really,

[01:54:35] Really, really enjoyed

[01:54:36] Speaking to him. So definitely tune into that. Um, yeah, that's pretty much it. See? Oh, yeah.

[01:54:42] I just loved this whole meeting. And then you were on fire and I just wanted to echo will. All of you were. But when you're posting anything to share because you care fill holes. I love that. I think we were all saying basically that. But I think you can't go wrong. And the more you post, the more traction you're going to get. If you really are sharing because you care and you're trying to fill holes in the community in a caring way, you're

[01:55:08] Going to win. And didn't go for it.

[01:55:10] I just wanted to share my good news that I got a job as a scientist at get this Facebook believe for AIs.

[01:55:19] Congrats, dude. That's awesome.

[01:55:21] Thank you. I can hardly believe you, but I'm super excited about it.

[01:55:25] So, so excited man. Vacillations. Why did you wait so long

[01:55:29] To drop in here? Because I didn't want to interrupt all these great ideas and all the things that were happening.

[01:55:35] But Vivian, that's so awesome. Thank you.

[01:55:39] Um, awesome that the guys. Thanks again for tuning in. Great, great discussion today.

[01:55:45] Hope you guys join us next week. And don't forget to tune in to the podcast. Give a listen. Also, I'm trying to think of my game on Instagram. The follow me is AIs Harp. I'm here to push forward content.

[01:55:57] So let me

[01:55:58] Ask you if people guys [01:56:00] take care of the rest of the weekend. Remember, my friends, you've got one life on this planet. Why not try to do some big.