happyhour-march19.mp3

[00:00:09] What's up, everybody, what's going on, everyone? Welcome, welcome, everybody, to the art of Data Science. Happy hour. Super excited to have all of you guys here. Thank you so much for taking time out of your schedules to come hang out with me today. Hope you guys are all doing well, man. So let's start off by saying, do you like this violence that's been going on against the Asian-American community over the last week? They should have fucked up. I've had many, many, many, many, many bad days of my life, but never once.

[00:00:38] I'm one of those bad days that are going to start shooting people. Let's just call it what it is, straight up hate crime, straight up domestic terrorism. That is not cool. Yeah, I just want to start off by saying that. But thank you guys for being here today and hanging out. Appreciate having you guys here. We got Joe in the house. We got Kate. Kate. How's it going, Eric? Jonathan is here. What's up, everybody? Super excited to have all of you guys here right on. And how are you guys doing? How's everybody's week been?

[00:01:06] My first office hours that it's been light outside since I started coming.

[00:01:10] So that's cool. That's the window. Oh, that's great, man. That's that's good lighting. You got there. I like it. I love it. Popcorn at the popcorn. Got a little bit of beer, you know. A little bit.

[00:01:28] Yeah. Yeah.

[00:01:29] I mean, it's happy hour, Joe. I mean, I'm about to have a beer in a second, so. Yeah. So this is a different time zone. So we have to you know, it's four thirty.

[00:01:38] I mean, so this is a local brewery here called Work Brewery and they are doing this variety pack and they don't name the beers at all. They just have different view of where these things called different suits from a deck of cards on there. Um, so it should be interesting, but super excited to have you guys here. We got the waiting room is packed. There's people waiting for me to let them in such that the men. And joining us now is Dr. Tom AIs. Right. Oh, man. How are you guys doing? What's everybody been up to this week? Man, I feel like this week has been going by super, super slow.

[00:02:10] We I just have to confirm, is that actually Kate me over there. That is popcorn, man. We hit the big time. Harpreet Oh yeah. Yes. We got the proper celebrities in the house today. Matt This is great.

[00:02:25] Well, I hope you guys had an opportunity to go and vote for your favorite Data community content creator. Um, the link is always in the show notes, so definitely go check that out and also help us spread the word. Help us, um, help us make this thing big man. I hope you guys could could share this on your own. LinkedIn tag your favorite content creators. And let's make this thing happen, man. Kate, you should also let, um, make sure you look at Eric's proposal for a presentation because he's got a great presentation idea. So, yeah, make sure make sure you get him on on to speak for for the next dedicated conference.

[00:03:03] We'll take a look. Yes. The next three weeks is all about going through the hundred and twenty speaker submissions and try to align it to industries and all that. So it's it's been fun.

[00:03:13] Yeah. Wow. Oh man. We've got Greg Coqui on the house. Nice man. Good to see you, Greg. Antonio Tor. Okay, Joe, what's up? Everybody else? If anybody has a question, feel free to go ahead and take the floor and go for it.

[00:03:27] I actually do have a question. I've been thinking about this all day today, so I just wanted to talk about some of the negatives of Data or like times when we tried to use Data for problems when it's really not appropriate. And maybe it's about ethical ness, maybe it's not. But like, for instance, I worked at a company where they really quantified everybody's performance and they would make decisions about like firing people or promoting people based just on those numbers. And I always really hated that and thought that was a really bad way to do business. So that's kind of like an example in my mind of like when, you know, we're using Data too much or it's not like really an appropriate problem to solve with Data. I just kind of was thinking about this idea of like we're in a Data revolution and we're trying to like, do the thing. We're like, if all you have is a hammer, then every problem is a nail kind of thing. So like what are examples of times when you think that Data is not an appropriate solution to the problem?

[00:04:36] So I get some clarification on that. So the situation you're talking about, is it like using metrics to kind of measure performance for people?

[00:04:44] Um, just in general, any time that, like Data is being used to solve a problem, that maybe it's not appropriate to solve the problem using Data and it could be about like ethical reasons or it could just be a company trying too hard to, like, shoehorn data usage into areas where it's not really like appropriate use.

[00:05:05] I guess I'd love to hear from Shantanu on this.

[00:05:08] I do have strong opinions about solving getting the right tools to solve the problem. So, yeah, just a high level for sure. If one is trying to force Data to solve a business problem where it doesn't belong, then that is not a great approach for for any problem solving. Anything like the way you start is by understanding the problem, defining it and sort of outlining what needs to be done. If it happens to be that one of the things that needs to be done is get get Data on this X, Y, Z, and then analyze it to validate or invalidate some hypotheses and so on and so forth. And that is the approach. If that's not part of the steps, then it's not.

[00:05:54] Yeah, I definitely agree with you on that. One tour's got some great comments here as well. You saying Data is never the solution. Can we support the solution? I agree with that as well. I guess it just depends on how you're using it. Like if you're like there's situations where maybe you don't need to go into, like doing super advanced analytics to solve just a trivial problem. That's just a waste of resources and time. But there's probably situations where you're using Data to make decisions that can marginalize groups, and I don't think that's ever right. Joe, what do you think?

[00:06:28] Well, I think you always got to remember the good saying way. Charlie Munger, Warren Buffett's business partner. Show me the inside and I'll show you the outcome. So you really need to start at the beginning. Like, why is it that a company would, you know, like or department feel the need to sort of like over Metro ties its business? That's just a made up word, but basically kind of mask every mask, everything behind Data like what's driving those decision makers to make that decision. Right. So you don't see to work backwards from from the root of like, why is it that these things are being done in the first place? So, yeah. Greg, what do you think?

[00:07:07] So for me is, is if you're in a room and you have a lot of anecdotes in a situation and everybody nobody needs convincing of the next steps, what the next step should be, and nobody needs convincing of whether these anecdotes are verified or not, then Data is probably not necessary there. Everybody in the room should agree that the next step should be X, Y, Z. So in this sense, you know, going after Data is kind of like a waste of time, especially, you know, business needs drive decision making. So these are big notes and everybody in the room aligned with what's already been verified. It's quite easy to make a decision with everybody in the room on board just to keep going. So no need to kind of dove Data in the sense that that's want to see it.

[00:07:58] Sartore had his hand up. Obviously, he's still here today, if you want to.

[00:08:03] I was just just a quick quick comment on that for me when I did many years ago. Well, 30 years ago, I guess close to it now, I was working in sales and Data was collected, of course, on how you perform and you had targets to meet, etc. It was never used to fire anybody, but it was more used to give you feedback on how you were doing, provide training, etc.. But ultimately, of course, if your numbers didn't meet your goals over X amount of period of time by by ideals, you were fired for that simple.

[00:08:44] Looks like Russell had some comments here in the chat that we're pretty good if you wanna go for a show, even though everyone has to see so many of you here.

[00:08:53] So my comment was Data is not the Data itself shouldn't lead to anything. The problem with it is it can be open to interpretation, mistake by the counterparts to it. So if a human has, say, a bias that they might push onto it or they just read it inaccurately, then it can lead to bias or objectivity. So objectivity in the in the outcome. But the Data itself is the ones and zeros at the end of the day. You can't have emotion that is completely enough.

[00:09:26] Yeah, definitely. Yeah, absolutely.

[00:09:30] So my team's doing a project right now related to predicting turnover in certain certain company. And really so as we're talking about it, you know, building, building the model, we've discussed that it's important that we use the data that we have ethically and maybe there are certain categories of data that we just don't use because sure, it's available, but just because and it's necessary. And then the other piece is, once the model is in existence now, is the model going to be used or applied elsewhere for other purposes? Kind of like, you know what it's like, is it? Created to track progress and then is it then turned to say, oh, well, we know that these kinds of people are going to turn over, therefore we're going to be biased in our hiring process against these kinds of people are going to you know, and it just really comes down to understanding, like we talked about, how do people did people agree to use for their data to be used in the way that it's going to be used to the people that the state is collected on, know how it's going to be used, if that's that's respect. And then I was thinking about this as if we were talking about it and say trust is a big piece. But the other piece is like, we can't always just trust everybody. Transparency is the other side. When you have a big when you have a big company, I don't trust every big company out there with my Data. So I want transparency. I want to know what they're doing and that it's that it's ethical and it's where somebody who's smarter than me can see it and be a watchdog on it.

[00:11:00] And just how honest I mean I mean the example I used to work well, I'm not going to name any names, but a product manager once brought up the idea that since we were doing Iot Data and he was half joking, but I kind of doubt it, actually.

[00:11:17] We can collect all this data on people that we can use. It was for workforce analytics, right? So the discussions were like we could just follow people around and find out where they are at any given time in an office and then come up with productivity metrics. And so what I what I ended up realizing, this also includes maybe trips to the bathroom, by the way, since that's kind of part of your office. And so you start taking it to its logical conclusion, you're like, well, now you have people that are tempted by the use of this new technology to start inventing new metrics for productivity. And that's why I think a slippery slope is it's like, are you doing enough to sufficiently do your job? But like, why do you really care where I'm at in the building? Like, why does that matter at all? Right. And so the thing that I'm going to find really interesting is when, you know, with remote work for the last year in the new tracking's of maybe certain employees or new technologies. And, you know, when you get back into this hybrid workforce workplace, like what that's all going to look like. I think to your point and it's like it just seems to be this temptation to throw metrics at everything. And I think it sort of reminds you of the quantified self movement where you just wear every fitness tracker known to man or woman all over your body and all over your house. And it's like, what are you trying to drive at the end? Like, what's the behavior trying to drive with these, you know, all these metrics? Right. I mean, have an Apple Watch. I mean, I collect all kinds of stuff, but like, what's the behavior trying to drive. And unfortunately, like I said earlier, it's like, what's the incentive and the outcomes behind these decisions coming from up above? That's what's scary. I don't think I don't take a lot of stuff to be measured just because you can do it.

[00:12:52] I think it's kind of dumb, actually, to that point. I'm sorry. I know I'm the one that brought up this question, so I'm going to answer my own question here. But it was making me think of how a lot of times there's companies that are trying to create some kind of like medical test or something to test for certain things like, say, there was there was one that I found one time that was like testing if you have allergies to a bunch of different things. But in the world of medical tests, it's really only as valuable as you can actually take action about it. So, like, if there's not really anything to be done with it, then what's really the point? You know, like I used to work at a genetics company and they would test for genetic variants that cause cancer. And it was like a whole thing of like, you know, when they first started of like, OK, that's great that we can, like, find out that the genetic variants cause cancer, but it's only going to be an approved test if that actually drives some result of like saving an insurance company money, because now people can take different paths for like how to treat their own cancer and stuff like that. So, you know, yeah, that's anyway, I just think that's also something that you see a lot of like tests that find certain things, but then like so what if I can happen for a sick Harp?

[00:14:12] I'm really sorry, first of all for answering. So generally at the beginning I think I missed part of the question because I was still logging off of work. But I stand by what I said, which is problem solving. Get the Data if you need to, and so on and so forth. But I do want to comment on this very particular thing as well.

[00:14:31] I mean, people try to use data to like answer all sorts of questions.

[00:14:38] And I totally think that there are a lot of the times they are mis motivated, poorly or motivated by the wrong things. And it's it's a very slippery slope. I'm not saying that one cannot absolutely cannot use data to say how a company should fill its workforce. But it's such a delicate question that I would want someone or a group of people to pay a lot of attention to the decisions that are made around it.

[00:15:08] To mitigate all sorts of biases, I wanted to build on what the example Eric pulled back, right. So you're you're performing the analysis on customer turns, right? If I understood well or employee churn and, you know, you have this six month long project, but if you know you have a poor culture in your organization, this is your company. You know, culture sucks. You know, people keep leaving. Are you going to wait for the end of that project to take action or are you going to use these anecdotes to start taking making decisions or are you going to wait for pulling data and things like that? So what can you do now, knowing that you have some clear, you know, Data like more clear events that's showing that people are leaving left and right for reason, X, Y, Z. And do you wait for that data to be analyzed before you can take action? That's that's what I wanted to convey.

[00:16:02] I don't thank you very much really to that answer your question. A lot of great responses there.

[00:16:06] Yeah. I just think that I was thinking about it and I just wanted to kind of have an interesting discussion about some different ideas. So, yeah, this is great.

[00:16:15] Thanks for the spark off. Some interesting discussion. So thank you very much for asking that. So I have Austin cued up next. Greg, did you want me to agree to the Q as well? Awesome. So I go to the Q and go ahead, Austin. Meanwhile, Austin is asking this question. If anybody else has a question, go ahead. Type that on the chat. I'll put you into the queue. Go for it.

[00:16:35] So my question is just around end to end Data projects and specifically how to approach planning one. My thought was initially, do you approach it more like a software project because it is kind of that whole pipeline? Or do you apply some kind of framework like the Chris DIAM or use like Khadra or cookie cutter? Where this is coming from is I do a lot of personal projects, but I'm trying to see like what would be a way to possibly introduce things like in like at my company head of that intrapreneurship trying to help, like, all right, let's just maybe do it one thing, start small and then kind of build from there. And so I was trying to figure out, like, what would make the most sense to kind of help show that business value and kind of get everyone on the same page.

[00:17:22] Yeah. So KEDO cookie cutter, that's great for like repository structure and just organizing the actual product project itself. But where I always start with any project is just a project analysis plan, and this is just a high level blueprint of what I have planned and what I plan to do and how I plan on going through this project. So that's kind of where I start with. But I mentioned this one over to Tom because he's got like this awesome framework that that he's actually done a few presentations on as well. And I really, really like Tom's pipeline for this. Tom, he shared that with us.

[00:17:57] Well, I was actually wondering if I should announce this, if y'all could just keep it here.

[00:18:03] Of course, I know it's recorded, but it's recorded and shared and that's OK. So Gaith Sancar and I just had our proposal for the same material approved to be a book with Packed. So I'm we're we yeah. We are beyond excited and we have some Megahed people that are going to help us with the book. And one of those I can't see his name, but his initials are Greg Coquille. And that's going to be nice to have his input because we know he's the top data scientist student in the world. So that's pretty cool. And then. Yeah, but so how do I put this? I woke up at 4:00 this morning to take my son to the airport and then I came back and went to sleep a little bit and I got up to be in Giovanna's workshop. So Harp I. What am I trying to say to you, brother? I'm half brain dead. Could you repeat the question just so I make sure I answer it really well?

[00:19:06] Yeah. So what Austin is asking is when we embark on a Data science project, how do we go about doing this? How do we go about planning our our route from data to decisions? And I was talking about how you've got this amazing framework, this pipeline that you have that I've seen presented a few times. And I thought that would be a great thing.

[00:19:28] No, thank you. That does help. And I thought that's exactly what you were getting at. But it's something that I developed more than a couple of years ago. And then I just thought it'd be nice when I was being asked to speak to to teach on this, because I sometimes see people, you know, not remain aware of all the the steps they should think through. But I mean, what can't you what can you not learn from watching all of Susan Walsh's great advertising? It basically comes down to treating the features with love. Well, actually, not always love because you do a lot of purging. It it's a set of best practices that we always want to remember to think through, and by the way, it's the spirit of when you look at great masters in any art form, what do they do? They get really, really, really good at the basics. And it's not to say that once you learn all these automated mechanisms or the pipeline, as some people call it, for developing a machine learning project that's going to go into production, it's not the seed that you would always use those. But by having gone through those basics, it's hard to go wrong even when you do something that's super creative and outside of that typical realm.

[00:20:48] So it it's just a spirit of remembering to go through all the steps of visualization always at every step you're trying to visualize things, but you you look at the kinds of Data problems you're encountering. Oh, I'm always going to have that in words. I've got to encode those. Oh, if I really want to get the insights I can always get from linear regression, whether I'm going to put that in production or not, I need to scale these features so that when I look at the weights on linear regression, I can get some insights about which features are most important. But I can back that up with principal component analysis PCA to see, hey, of all these features, which are the strongest, I can look for co linearity, by the way, I'm just throwing out examples. I can look for co linearity because I don't want to feature competing for the same position on the on the modeling soccer field, so to speak, that causes in some models it causes problems and others that doesn't care. But it's still good to only retain one of those. And which one do you retain. The one that's got the higher I can value again go back to PC and.

[00:22:03] But what are we trying to do. We're trying to get that feature list down to an adequate size. Again, cursive dimensionality. The smaller the model, the better. But I shouldn't say that the smaller the model that can do the job adequately, the better. And that gets into your are too adjusted metric. When you get down to you, you kind of build in a penalization for using too many variables. Why do you want are too many features? Why do you want to do that? Because if if you if you have even one too many features now, it won't generalize. Well, might overfit. I could go on and on. But then again, I never assume which model is going to work best. There's ways to loop through all the models, loop through the metrics for each model, etc. But as you can tell, I probably should write a book on this anyway. I hope that helps. But we can take it offline and I can I can dig out the link and put it in the chat so people can see the the talk that we gave and definitely love to hear from some other people too, because, I mean, everybody has their own way of kind of proceeding through a Data project.

[00:23:11] But I kind of have four distinct phases when I do it. And this is taken from my good friend Kyle McHugh, and it's like define, discover, develop, deploy. So define the problem really clearly. Create that project analysis plan and then do discovery. That's the EDA stuff and then develop a solution and then engineer that solution so that it is deployable. So let's hear from Ben on this.

[00:23:39] Ok, did you know I was talking to you telepathically to calling me? So I'm actually kicking off two projects right now, which I'm super excited about because I haven't done outside of marketing. I actually haven't done any real live coding since I joined in or about the last year. And so I'm kicking off two projects.

[00:23:54] And so I think the important thing is I love how timeline's I'm not telling people to procrastinate, but it's really good to have timelines. So if you have timelines, things just work.

[00:24:03] And and I like projects where they have to work or I'm swearing and I'm angry. So I guess the the only word of advice I offer is don't purrfect stuff that doesn't need to be perfected because that just slows you down. And so figure out with the least amount you can do is and then figure out a timeline for that and then hold yourself accountable to that timeline and escalate if you're having issues. I know that's not super helpful because most of my projects that I do are super hackie. Anyone that's more experience with actual projects that go live in the wild and people use are pretty hackie.

[00:24:35] When they first start, then I clean them all up and the better. So I thought there's some great comments here from from Angelo I don't see on my screen. Why are you still here? Yes, you are. Go it.

[00:24:47] So so my experience is slightly different. It's more from the corporate side and it's to get access to Data or to get access to the right daytop. It'll take probably more time than getting your personal project up and running and doing kind of the end to end. And what I was saying to to Austin. Is if you want a model, if you want a model of personal projects and end to end, I think the most difficult part is navigating through, you know, the business understanding and trying to define what you're actually solving, although there must be a project initiated within the firm, but no one knows how to assign someone to the first point of contact and they let him rot until he figures out the rest. And then they might be thinking there's a project going and they still haven't got the Data. They don't know where the Data are coming from. And people are asking for updates about the projects and how well the project is going. And you still you still haven't figured out what the data is. So it's kind of a good start much earlier in the process than actually getting into the into the modeling part. Yeah.

[00:25:55] So thank you. Thank you, Angela. Let's hear from Shantanu on this one.

[00:26:00] Yeah, I think everyone's really awesome stuff and I think that I just want to marry together with Angela just said. And what Ben and Harp also said, a process is important, be it a team process or a personal process. That's not to say like it has to be by the book. But one thing that I've I've really tried to work on in the past year or so is this idea of thinking like a product manager. It's a little bit clichéd, but it's true. If you can roadmap once you so Harp talked about coming up with a project analysis project plan. Once you have that, if you can really roadmap out the steps and what needs to be done, allowing the right amount of time for every stage that I love, you have to get yourself a generous amount of time to really explore the Data understand what it's saying to you. So, you know, as long as that's on the roadmap, then if you're spend five days going down this rabbit hole, you don't have to be lost because you had given yourself five days to do it and give it depending on the size of the project, obviously. And then it's a for me it becomes even more important actually in the next phase, the model sort of tuning phase, because there you can really sort of go off the rails, right? You don't you want to optimize everything you want to do great grid searches for everything. But at the end of the day, you know, time boxset because and then the road map is also going to help you get to that from end to end. Right. It's going to get let you have a barebones product project, whatever it is, and then you can iterate on it. That's fine. So you think like a product manager or roadmap your process.

[00:27:51] I'd even add when you're building your model out, even during the project analysis phase, just call out a baseline model that you intend on using just so you have something in place. I typically do this before I even do any type of exploration. I say I just, you know, having a intuition about this regression problem, classification problem, whatever. I just pick a baseline model and try to improve on that. And I'll start with maybe Corneau even before doing any like four or five candidate models that I choose to pick against this baseline and then go back and update my project analysis plan. If none of those things work out or if I gather gain new information through the exploration phase. I do actually. I'd love to hear from Antonie on this as well. So I didn't see that until you do.

[00:28:36] And I'm sure. So Austin was he was talking also about a little bit about like I think how do you how do you start at the corporate level? And then what I can add to that is I like to have kind of like low hanging fruit and then have some harder models. So, I mean, always depends how tolerable your business is. But when we start a couple of projects, some work on the strategy side and I work with a data scientist, I would like to define like, let's say three projects. One, it's like, all right, this is definitely going to succeed and it's going to be very easy win. So that way, if my harder projects fail, everybody is not totally demoralized. So I have like short, medium and long term projects and then kind of easy, medium and hard. So especially like you said, if you're if your organization is kind of new or they don't trust fully that data science, you don't want to tell them, well, I'm going to work on this thing. And then three to five years you're going to see results, because especially in today's world, everybody moves so fast. People just change different jobs, they get different promotions. And it just I've noticed that everybody doesn't have that kind of like long term vision. So kind of balancing that out and making people making people happy in the short short run is going to keep them motivated. But also it's going to get on even more exciting. Right. You kind of give them like a little bite and they're like, oh, this is this is good. Let's do more, you know? So you give them more and more and eventually you want to gain their trust and kind of. Then you become data driven, but it's not going to happen overnight.

[00:30:11] Man, thanks for taking us back into that corporate aspect of it. And we kind of went off the rails. There wasn't that answer your question. You kind of muted their microphones that not working us. And I quote, I'm just noticing that I work a shout out to some new faces I see here, like the do the coolest name ever, Mac Steel, I guess a cool name with the sense field T-shirt. I haven't I haven't heard that since late 2003. That is awesome. Some of the new faces, Nathaniel Wise, that chart, Brett Butler, you guys said treat the welcome. Eric Gitonga. Right. I'm happy. See guys here. If you guys have questions, go ahead in the chat or if you want to just queue up, then let me know. Semiprivate message and I'll I'll add you to the queue. Next up, we got Greg.

[00:30:57] So a quick question I wanted to ask just for for fun. What do you guys think? One day a post from LinkedIn will be sold through NAFTA. And what would be that post about? Is it going to be around EHI or something or someone of the right stuff in, you know, will it cross? Which one would cross a million dollars?

[00:31:22] I think the one of Ben wrapped in a blanket in the middle of the desert is definitely potential for a million and selling it now for ten bucks.

[00:31:32] Who wants it. Ten dollars. Then maybe it's yours.

[00:31:36] I just want to know, was there any was there any real time footage of the bin in the woods phase?

[00:31:43] Yeah, there's a channel to I'll post a link in the chat.

[00:31:47] Awesome.

[00:31:47] I've been wanting to ask so that so I was a paid extra on the TV series The Chosen. That's what Antonia's mentioning. One of our investors is an investor. They're two totally random because I'm the least likely person to be on like a Jesus said, I'll go grab that homeless link.

[00:32:02] I think we're just sharing something recently. And I just been in a chat where there's something about LinkedIn content creators getting paid. What was that all about?

[00:32:12] Yeah, I read an article and I don't know how legit this is, but apparently LinkedIn content creators are going to start to make money. I don't know when and how much, but it seems like, you know, how when you call it tick tock is paying their content creators after you hit a certain number of subscribers? Yeah, it sounds like LinkedIn is going to start doing something like that, but also you'll be able to get hired by brands to promote the content. So I think they're working on a new feature like that.

[00:32:41] That's pretty cool. I like that. Ichiko thinks it will be a post from Lex Friedman. I don't know, man. Every time I see Lex Freeman post, it's just a quote of some old dead person. I don't know. Can you see a little bit more original than that?

[00:32:53] I know, but look at how many likes and reactions he gets on that. Yeah, right.

[00:32:57] It's just it's just in quotes like of dead people all the time. Nobody likes my anyway. So I'm sorry. It's hard for you to hang out with Joe Rogan and do jujitsu and you'll be cool. Yeah.

[00:33:08] So we're also you got a comment there. I would love to hear your thoughts.

[00:33:12] What. That's about the NSA stuff. Yeah. Yeah.

[00:33:15] It's like you said, it'd be fun, but if the if the first LinkedIn post purchased by an entity was one of the early posts talking about NFC and putting it down, you know, just just saying this is this is crazy, you know, what was it? Sixty nine million for a piece of artwork, you know, what's going on? And then someone actually buy one of those posters. Justin Bieber is totally so Metzer.

[00:33:37] You know, I just happened really researching this blog and stuff since our conversation last week. It's really got me fascinated. So I'll be talking to I can't pronounce his last name, but it's right and tell something like that. But he's got a bunch, of course, on LinkedIn learning, so I'll be talking to him about block chain and stuff like that. But I've got this book, which I found interesting, which I want to go grab. But Bitcoin Data Analytics for Dummies, I think that could be an interesting intersection of analytics and and block chain have actually been picking up. Um, I've been getting deep into theory, um, this week and I actually sort of like learning solidity and smart contracts and stuff. It's fascinating stuff. Um, but yeah, hopefully, hopefully one of my one of my posts can be sort of NFTE starting at twenty five cents.

[00:34:24] I think one that should be sold is, is a André's post where he said if you have to, if you have to ask how to become a data scientist, you'll never become one. A lot of people are really upset by the Post, but I love that one. I try to take off like of the troll posts I made and post about the option to work.

[00:34:43] Oh, I was making fun of the work and people said I was a bigot, I was evil and I was sexist just because I said I don't like that. The work hashtag. Sorry, I thought that was funny.

[00:34:56] Yeah. Man bit of a straw man attack. That's that's not cool. Antoniou.

[00:34:59] Oh I think maybe a little something like somebody young now and they create like they become some billionaire like the. Next year, Basils, if you could somehow have their early college days LinkedIn profile, I think that would be pretty cool. And I was thinking about that because I was like, if I can kind of the way people trade, like trading cards, if I can buy like an early Jeff Bezos Amazon business card back from when he was like starting out, I mean, I think that will be worth millions of dollars right now or like Elon Musk's first resumé or something like that. I mean, that would be so sick I would buy it and hang it up and never spell it.

[00:35:41] What do you think about the very first iPhone? Because I have one of those LinkedIn. Do you think that would be worth something?

[00:35:47] I don't know. We have to do some Data digging and see how many are still in use because I work for a telecom and you'll be surprised how many people use ancient phones.

[00:35:56] Still keep still rocks beeper. Mikiko, once the once the mind is to release their hold on and Vidia GPS.

[00:36:06] And that's a good point because it would help all of us, all of us going dollar bill from Warren Buffett.

[00:36:15] It's cool. It's going to be an NFTE.

[00:36:18] I've got an idea for you. So what you do is you buy the previous Gen's GPS and you put them in a mining rig. Can now do you have the super simple trainer?

[00:36:31] That's basically what a paper space created and all those other guys did is that they bought up all the cheaper Jebus, got stuck with it and video came out with better ones. And now they're like, oh, we're going to just use it to build our free tiers and then, you know, get people locked in and then sell them on the more expensive ones once they have all their models living there. So that's basically what they did.

[00:36:52] Questions from anybody else. Shout out to Amy Smith in that room. Avery, good to see you here. Also, Robert Robinson, good to see you here as well, my friend. If anybody wants to take the floor with a question, go for it.

[00:37:03] I'll go for it. Absolutely. So I've been looking recently. I'm doing some research on the challenges, but that's across the spectrum, you know, whether you're working on data science, data analytics, data management or product management. So I guess my question to each one of you sort of sees it from their own personal kind of role in their workplace. What what is the biggest challenge that if you could solve with a magic wand or the biggest challenge that keeps you up at night would literally transform not necessarily your world, but your contribution would kind of make make problems go away and sort of you would see huge value add in what's happening and kind of about perspective. What's the one thing that if you could change that would transform completely the picture in what you do hear from a singer from from being on this one?

[00:38:00] I apologize. I was trying to answer a slack really quick before I forget. Yeah, you said the question and one or two sentences. I won't do that again.

[00:38:08] I'm sorry. Yeah, no, no, no problem. So Angela's asking if there's one thing that you can change about the culture environment, I guess, of your work to make your life easier or to be denied bail.

[00:38:20] Angelou or not not necessarily culture or anything could be kind of it. You couldn't get access to the Data or it could be a culture Data culture would be a tech for anything. The one thing that would really transform what you do and how you want to achieve things and maybe for your clients as well.

[00:38:37] Yeah, I think for me personally, the one thing I think about this a lot for my career is what are the activities I do that I'm I'm overqualified for? This is everyone on the call. Like, what do you do during the day that high school you could do? And if high school you can do it, why are you doing it? And so when you work for a bigger company, you can hire those resources out. So I think that's the constant thing I'm thinking about, is what are the activities that you are uniquely? But I guess that's a question for the group. Do you have a super power that you haven't been able to run with or do you have something that you're uniquely suited to help your business, but you're stuck doing some mundane things that you should not be doing? And so that's what I care about the most, is how do I stop doing things I shouldn't be doing? And a big part of that is hiring you hire the right people. You tell your boss you work with them. Does anyone else still with this stuff where they're like high school, make it do this?

[00:39:26] And I don't say anything necessarily. That's like high school meeting high school. Me just was an idiot.

[00:39:33] That's essentially what I'm saying. I saying, like an idiot can do this. It's an idiot can do this.

[00:39:37] And can we bring this into like junior talent or someone elsewhere?

[00:39:42] Well, that doesn't answer that question. But my thing is statistics and machine learning and classical machine learning, things like that. But I'm in a position where I have to help develop a Data strategy for this massive organization. And if I do, I've never done this before. So I'm like reading books and trying to figure out how I'm going to make this happen from nothing in my company is massive. It's not like it's a company made up of like. 13 or 14 other small companies and it trying to get all of that together and come up with a Data management strategy, Data strategy like this. It is difficult, man. Like it's not it's not easy. And if I can if I can somehow just make the Data just fall into order somehow so I can get back to the statistics on masculinity, like I'd be the happy skynyrd's. But I realize that this is a skill that is important to have that not a lot of data scientists actually get to work on. So it's like I'm able to clouds and that type of thing. I like doing this Data strategies like way up here, high level and the machine learning and statistics stuff is like the dirt getting getting dirty with the Data. So I don't know if that makes sense, but that's what I do.

[00:40:49] I bet there's people on this call that can help out where they've already like. Maybe I'm thinking of you, Antonio, or other people, where they've they've been there before. They've had to like revamp strategy or pull resources and fix silos. And like, I'm sure there's people on this call that have had to fix things. Yeah. Across the network. But I like what you're bringing up because that's something that high school you can't do. And so I, I hate when I get stuck with a high school me workload that distracts from the things I actually want to be doing this year from entering on this one.

[00:41:19] And then if anybody else wants to chime in after Antonio, just send me a message. I'll call you after that.

[00:41:25] Is this more about the Data strategy or more about the high school year?

[00:41:28] Well, let's let's try to answer Angela's original question. And then if Angela wants to hear about the strategy, I know I do think we can go into that.

[00:41:37] So, Angel, I was listening the whole time, but is is there a question around the high school you was about what is more about what you see in your environment and your kind of Data world, what you see the biggest talents? And what would you kind of if you if you removed one of these challenges, what would be the impact for your world?

[00:41:59] I think this is all is communication. Honestly, the ninety five percent of the problems I see are related to the communication rather than technology. A lot of the projects I work on is across different departments and just getting people to agree on something, whether that is terminology or just I mean, sometimes it's the same thing as Sakine. Like, for example, I would figure out that two different groups or three different groups are all talking to the same vendor, a vendor, let's say they're all doing it. And separately, they're all being a separate price. They're all they're not getting that. If you if you all talk at the same time or you're going to have one solution, it's going to be a lot cheaper and it's just exponential. So much more cost for Ford, for the company. Another thing is we're trying to build a Data product, and this Data product is very similar needs for three or four different groups. And people would be like, well, who owns it or why does that person get to say how this column is named, for example, a lot of times because so my kind of group is new and a lot of these people have been working together for like ten years. Let's say there's a lot of history that goes behind it. Or sometimes like five years ago I gave this person Data and maybe I gave them access to the table and they never touch that table or they messed up something. And I'm like, this was five years ago, you know? So I it's just silly things like that. The like has been said, like even high school you should be able to do. But when you when you look at it, you like, oh my God, if people can just communicate like listen to each other and meet halfway, I think so many problems in Data and honestly and just in life too would be would be solved. And sorry that it's not like a technology answer, but I always go back to the communication.

[00:43:56] You know, that is interesting. The one I work for a big, big corporate. Yeah. Big corporation. That almost sounds like almost sounds like a procurement to two things. One, kind of siloed or siloed departments, organizations, and then not kind of group level procurement governance to actually make sure that if if a department is using a vendor for X, Y or Z reasons, then that's available to the other departments in the organization.

[00:44:23] Right. And a lot of these the big corporations who are older, they have they've been existing for like a hundred years. It's yeah, it's the silos. Silos is the is the problem. But honestly, it's in theory there's a lot of easier ways to solve it. But once you start digging into it, it's not that easy. But but it's fun.

[00:44:47] It's fun for thinking. Let's hear from Avery and then hopefully we will hear from Shantanu. Avery, we are unable to hear you. Right. We'll go to Shantanu, then we'll come back to Avery after he sorts out the tech.

[00:45:02] Sure. Yeah, yeah. I think a lot of the discussion is around. Setting up processes and delegating roles appropriately, which I fully agree with, that's extremely important. I think on top of that, for me, if I had a magic wand, I work at a small startup, less than one hundred people, I would get I would get documentation on all of our choices, all of the systems and services that people have put up. Because, I mean, even tech debt is you can deal with it as long as you understand where it comes from, what drove those choices at the time. So just absolutely not just documentation and code, but documentation. And then also, like a you know, the documentation has to be organized because it is a huge and this is, again, going back to, I guess, high school me. It's a huge waste of time to have to go through the Google suite and then the confluence of all of these things. So, yeah, that's not a Data science answer. But that's that's my one thing.

[00:46:07] If you can somehow just have all that information, like beamed into your brains, you don't have to go do all the digging yourself like that would be that would save so much time this year. Avery, how's your your setup gone?

[00:46:18] You tell me. Can you hear me now? Yes. OK, great. Sorry. Yeah, sorry about that. Yeah. I just as a data scientist I struggle with deploying things online. Like I always say, data science is kind of like, you know, in the needs like a solve world hunger. Tell nobody I feel like that's kind of data science. Sometimes it's like, oh, I made the world's best classification app or I can predict something super well. And everyone's like, oh, that's great. Like how do I get access to it? And it's like, oh, it's on a Jupiter notebook on my local machine. And it's like, oh, OK, that's not useful to anyone if you keep it to yourself. And so that's where you like you get into like machine learning apps or Data engineering. I need to deploy and share with other people and I still get that. I am no good at that. And so I was saying in the chat earlier, that's why people like like Joe exist and, you know, other people who are good at machine learning and or operations and Data engineering. And that's why software is exists like Data robot to make that interest more, more easy. But I just want, like, a magic wand that just makes that happen and then it's done.

[00:47:19] Just just put it on the block. The doctor, maybe you can have the kids do all the work. And I don't know, Mickey, about you. I saw you put them in the chat.

[00:47:27] Yeah. I think the three kind of top like issues. I think about our number one, like educating, communicating non Data people on like what what what does like machine learning at scale look like.

[00:47:41] I and then I think the other two people have mentioned comments, Mellops and documentation. I think that's already been covered. But in terms of like educating, communicating, I mean this is something that came up really sort of recently and serve ties into that roadmap question of like how do you come up with a strategic roadmap, right. For a longer term vision for companies? And to me, it's it's really been a struggle because I think so AIs Data or Data science or people, we sort of take a lot of this learnings for granted, you know, but for example, for an early start that I was working on a law, the struggle around, you know, setting up infrastructure, figuring out the tools, getting funding, figuring out how we allocate that funding and even figuring out like how do we allocate our time. It was very challenging because essentially, like, first you have to explain to someone, well, what is like a typical machine learning process look like ignoring the scale part. Right? Just ignoring it. Why do you need to do certain steps and then you have to layer on, OK, this is what we need to do at scale. And then you have to layer on other of conversations like around privacy, privacy and like ops or things like that.

[00:48:47] So there's so much data science machine learning material out there. A lot of times it's written from the perspective of either you're a practitioner, you're like a manager, and if it is targeted towards like executives or leaders, it's usually kind of selling hype and kind of doing some, like, high level defining. And I've kind of yet to find like a practical resource that combines the like. This is sort of like what you need to know, especially with regards to the practical information of like what tools you actually need and serve. What are some of the options like? I think full scale learning comes like the closest, but even then it's like really it's really kind of targeted still, too, like Michy, like practitioners and like managers or directors of technical teams. And that's really been my struggle in every company I've worked at, which is if you are trying to like, come up with a roadmap for like a year long process and you kind of have to like milestone it, or if if your business partner does not understand what that process is, that they don't understand what those milestones are, then you're pretty much going to get pressurization switch like basically like every quarter, like every quarter. I think we had like twenty five percent of our projects are there, had be pushed back or just were taken off the roadmap either because and part of it could be that they weren't.

[00:50:06] Scoped correctly, maybe they weren't, in fact, answering the key business question that they needed to, and that happens a lot, right? We all get really excited with our tools and our little like, you know, our little science projects in the corner. But I think a good chunk of it was that we, unlike the Data science or machine learning side, had assumed that our business partners kind of knew exactly what they were talking about, when we should have sort of thought about the like, you know, you don't know what you don't know. Right. So I think if there was some kind of tool or resource that wasn't one of those, like, really boring, like educational seminars you go through, like in corporate. Right. That was like practical was kind of like the bridge. I think it would help a lot. It would it would help on like even like the day to day, just not having to kind of re-explain, like, why do you do this, why you do that. So I mean, I think that's been one of my biggest challenges.

[00:50:53] I don't think you Mickey, let's hear from Toure and then wrestle. And if anybody else has questions, go ahead and let me know the chat and I'll ask you to the Q right now, the queue is empty.

[00:51:02] So, Gilpatrick, I'm just listening. It's quite interesting because technically one of the reasons why I'm joining in on this group is to kind of try and learn and get a better understanding of data science and Data programing and all of the things that Emelin Analytics and all of these things have been done. But for me, we're talking about somebody mentioning the silos that create problems to meet silos in themselves is not a problem. It's the communication between the silent silos that can basically refer to anything.

[00:51:38] I mean, it's like as a manager for one silo, you are not understanding the other silo and therefore you're not able to communicate language barriers, language barriers in the sense that the communication between you, if you don't speak a common language, then you have a greater problem for solving the problems you should. So for me, like a prize, that is a way of communicating between two sides technically, but communicating with management and communicating with the technical personnel, etc., these are very common problems, not just in data science, I'm sure. Personally, I've had the same thing working as a business controller. You have to communicate with engineering people, management you see from many different people and complaining and communicating with them required different languages. And this is really where I think the key is in an organization or a project is that you get to a common understanding before you actually start talking. And to do that, you are stupid questions. I call them stupid questions because very often people are not comfortable asking questions because they think they may look stupid or their high schoolers or whatever it is. So if you have it, I what I do on meetings, I ask the stupid questions.

[00:53:07] Even if I know the answer, I'll still ask those questions so that I can benefit, or maybe somebody else can benefit from it. All these terms and terminologies that are being thrown around PDF and quite often at all because of my having joined this now over a month or two, you know, I go and I start reading about Python. I start now. I'm starting to understand what you're talking about. But in the beginning it was Greek. I had absolutely no idea what you guys were talking about. There's still a lot of so and recommendation I have here. If you use a term iPad and or some other term, add on and explain what it is, because this is how you educate people, get used to it. When you use the three letter term, then add what it is because it will people will not ask. They think that they're stupid or they don't know. So this would be a good step. And that is, in my mind, a magic wand for training, educating people around you and your team and your organization.

[00:54:13] Absolutely love that. I usually preface all my questions with that. Are dumb questions like this is a dumb question, but I'm going to ask it anyways. And then just I've got nothing to complain about because I warned the most dumb Russell. I made comments in the chat here, so I have to wrestle. If anybody wants to take the floor, the question go for it.

[00:54:33] That's done quite a lot of the comments in any one particular, you were looking for more elaboration.

[00:54:38] Well, just to just to unlock the comments from the chat so that they're on the idea frantically scrolling back up to see what I've said now.

[00:54:49] So, yeah, I, I think the main one from from my experience, one of the biggest challenges in in large organizations is the conflation of what an individual wants and what needs to be done. So trying to separate the vanity of the person's opinion of what should be done versus what's the best thing to be done is and that can be as straightforward as a binary decision this should be in or it shouldn't be in or as onerous as, you know, change that TULLEN by a couple of shades of hue. So that's a better continuity on the page, etc.. Those types of things I find can be real kind of time sinks trying to to get to the bottom of those decisions if people are not able to separate their own perspective from from the product that we produce. So yeah, I'd waste a lot of time on it. Yeah. And I've said a couple of things more recently, but I'm sure there's been far more comments. So, so happy to listen to others in the meantime. Thank you.

[00:55:52] Thank you very much. RESTfully and everybody listening on the audio. Don't worry, I will link to the chat because there's a lot of good stuff happening in the chat. So if anybody has a question, um, definitely go for it to the floor. Uh, Jennifer, what's up with you now?

[00:56:07] I'm just happy to have a great Friday with you guys. Yes, absolutely. Hadija does anybody Miss Dave lingers just as much as I do. I haven't seen a guy in like almost a month. Man What does he been?

[00:56:18] I have a theory about him. You know, how he looks so much like Sean Bean. And every time Sean Bean appears in a movie, you know, he's going to die. I'm beginning to think that David Languor died. I can't. Oh, he didn't die. He's still posting.

[00:56:33] Yeah, it could have been crazy. I don't know. He's fine.

[00:56:38] Yeah, he's actually next week. I think so. The week after. Yeah. Yeah.

[00:56:44] Nobody has mentioned the evil empire or the free lunch there in weeks. And I miss that too. Yes that's true.

[00:56:51] Have a question. Yeah. Go for it. So we, we hear about a lot of hype in email. And Tom you can, you can say you Transformers. You talk about that and I here are important D.R, what do you guys think would be the next breakthrough? You know, something like the hype of, I don't know, when your network came in in people's minds or whatever it is, became a Harp. What do you think is the next wave? Is it going to be existing Tuesday, going to pile up kind of in a creative way and then make it look like the big next thing that people will start talking about the next five years?

[00:57:31] Or what are your thoughts, Ethereum and Blockin for like this is obviously the future. And if we can find an intersection like while the while this thing is still emerging, if we can find an intersection to take our data science skills and combine it with whatever blocking the theory, there could be some unique opportunities for not only just personal research, but just opportunities that we could see for ourselves and an entrepreneurial type of space, or just doing cool stuff for our companies because I'm all in on this Ethereum thing. And, you know, I mean, obviously got a book for dummies. I'm not a dummy, but this is going to be interesting. I'm really excited to dig into this. I'd love to hear what everybody else has to say.

[00:58:10] I'll go. I think you hit it on the head, Greg, but I'm not surprised you do better research into space than all the other data scientists put together. But, yeah, I think it's going to be something cool, like ensemble transformers. But I would guess the Transformers are going to be smaller, easily trainable on a single GPU or not, but it's going to be the person that figures out how to ensemble train them. Well, that's that's to me, that'll be a significant breakthrough. But not the last breakthrough stuck in step was then it was no in.

[00:58:45] But, you know, some cool stuff about causal. I think he's pretty bullish on it. The more I start thinking about it, I think that's that's a that's an interesting field, especially from the operational aspect of it. How are you going to production AIs causal models? It's a bit different. So stay on that.

[00:59:03] Can you break that down for us? Causal Emelle.

[00:59:05] How much about it man. Like I wish, I wish I knew more but I'm not going to claim to be an expert yet. Give me a couple of days or Kurzawa, I'm all for dummies and then I'll know all about it.

[00:59:16] Just doesn't have anything to do with active learning does it. Because I'm here now.

[00:59:21] Active learning is something people are starting to talk about now. So it's kind of like training on the fly with live Data. And in turn what you end up doing is being able to train a model with very little amount of Data versus, you know, a big batch that you kind of, you know, push through. So active learning is starting to be the conversation nowadays. So I'm curious to see how this might change things. So I don't know if causal model might have to do anything with that.

[00:59:52] So would probably go through and read all the girls books on this and and probably become more confused about everything in the universe.

[01:00:01] But when you say Judea Pearl's books, is that any anyone in particular?

[01:00:07] I mean, he has a popular in the book of why, but I know he has some others, a causal inference and few others that are a bit more mathie, which I tend to run into something. I tend to go that direction just to nerd out more. Yeah, stay tuned. Because because I don't have enough other stuff going on. Yeah.

[01:00:23] Yes. See Harpreet Sahota what to do encourage you that. I think how do I put a cerium is awesome by the way, because you can use it for so much more than just another cryptocurrency. I'm really glad you're going after that one.

[01:00:36] Yeah. This idea of smart contracts and and just the block chain where you can just run any programing language and I think that's awesome. So definitely been enjoying learning solidity. It's been quite fun. Vivian, I saw you unmetered. So if you want to chime in, definitely go for it. Yeah. Then after Vivian, let's hear from a listener from Shantanu Mikiko.

[01:00:56] Well, I didn't look too much into this, but just earlier this week I saw the news that Facebook like which training some algorithm, using photos that it like labeled the Data like for them. And I thought that was a cool idea. Again, I didn't look that much into it like how they did that.

[01:01:14] But talking about the self supervise the self supervise model.

[01:01:18] Yeah. They like collected like a billion images or something. And they were trying to with the goal of trying to get an algorithm to label the images for them so that then they could like do classification. So the goal of like trying to find some way to get computers to label the Data for us instead of having to have humans do it. And I thought that was a really cool idea.

[01:01:39] The way I understood this one is kind of like the way I read it. Somebody might tell me differently this Savi's learning that Facebook put out this kind of like a model that kind of do an internal check, kind of like this is a table. I recognize this is a table. Can I confirm for myself that this is a table kind of thing, kind of like the idea of being self supervised learning. Maybe somebody can help me understand the idea behind.

[01:02:02] Anybody know anything about supervised learning? What was the question?

[01:02:06] Yeah, so they talked about I think the acronym is Seares that Facebook just released. They trained like billions of pictures and they labeled them. But that model is kind of like based on self supervised learning and the way and that is kind of like a model that kind of teach itself how to classify pictures. And to me, it's kind of like, you know, I don't really understand a self supervised learning concept. Is it one of those, you know, hey, I'm going to learn what a table looks like. And I can confirm by myself that it's a table without asking people that it's a table.

[01:02:44] I'll just pick up just kind of a beginning here. If you look at what the models can do, I would imagine and by the way, please don't be afraid to see bullshit Tom on this, but I think because of the way they've trained some of these transformer's, they might be able to, at least in the language processing realm, begin to create legal training Data. Well, it makes me wonder, Transformer's can be applied to vision problems, too. Why couldn't they begin to label vision for classification? Data is just a thought experiment here, but it seems like that was one of the many reasons for IGP. Three was such a huge breakthrough because it's not it's not a supervised form of learning and yet a complete thought. It can complete assignments. So if you can kind of take an extrapolation from that point and think, yeah, I don't think we must be too far from being able to do self supervised well or machines supervised. How would we put I don't really like the way that's called. I think what we're meaning to call it is a machine labeled machine labeling of Data. Would that be more accurate?

[01:04:04] Yeah, see, I looked it up and interestingly enough, it says that on tax, Lincolnwood is self supervised learning. According to John, the soon a computer scientist known its impressive work in the field. The closest we have to self supervised learning systems are these so-called transformer's which tomate.

[01:04:24] Yeah, Yambuku. And by the way, the CNN creator. Yeah, yes.

[01:04:28] Yonaguni. Um, so hopefully answer your question. We've got a question actually cued up from Jeff Cheen. And so, Jeff, I think you are still here. Let's go to your question and then for the respondents for this Mulgoa to Shantanu and Mikiko.

[01:04:44] Hey, sorry, I didn't I didn't really have a question. I was just commenting for the gentleman that was talking earlier about the stupid questions.

[01:04:52] You know, I'm I've been trying to get into Data science for two years now, three quarters through a master's degree. And I sit here and I listen to this show, this podcast, and I just like I feel like an idiot eighty percent of the time. And I feel like it's just such an incredibly broad field that sometimes, like, it would benefit from some I know some better stratification of, you know, levels of technical. I think it was three weeks ago you guys were talking about technical ability and the word technical ability has lost all meaning to me because I've learned to write some python, I've learned tabla and I've learned some other things. So in my mind, I'm now technical, but I listen to you all talk and it's like, no, I'm still I'm still high school, you know? So there's not really a question there. I guess it's more just a comment on the field of data science and how broad it is and how difficult it is to understand from the outside looking in.

[01:05:40] Yeah, that's still I mean, 80 percent of the time you do better than me. I feel like an idiot. Ninety six percent of the time sitting here listening to people talk. So you better than me. But I think the the thing is, when it comes to like technical ability is just a willingness and just being unafraid to want to go and learn the stuff. Um, I think that is the core of technical ability for you. It's just like, yeah, I don't care if I break this thing, let me test it out, see what happens. But let's hear from I want to definitely hear from Shantanu and Mikiko on this and then we'll go to Greg after that.

[01:06:13] Yeah, I think technical ability is so vague and broad and as mechanical points out in the chat, a very context specific as an example. So we were talking earlier about what are the things that you enjoy doing or consider part of your job description and what are some of the things that you have to do anyway? A lot of the times I'll get requests from account delivery or whatever for to pull some sort of analytics. Right. It might be a quick one. It might be complicated and it's not part of my job description. But the person that I pass it to is senior director at the company and they're the director of Data Insights. I happen to be a machine learning engineer within within engineering. Right. And. That person is definitely a couple of grades above me, so the point here is that, you know, it's not like beneath me or something. It's just like a different field and someone else's is better suited for it. So, yeah, I totally agree that Data sciences itself is very broad. And just figuring out where you want to focus and usually that decision will help you make that decision is where your existing skills are. So just mapping back to something you've already done and some experience you already have, and then try to use that to hone in on which part of this world of Data science you want to specialize in.

[01:07:41] That's an important point because there's not one particular skill you can point to and say that that is a that's that's the data science is right there. Some meta skill, right. It's a combination of various different skills. So definitely. Definitely that response to you, Shantanu. Let's go to Mikiko and then Greg Soraia, which question? Why am I speaking to you and to Jeff's remark about.

[01:08:06] Oh, yeah, yeah. I mean, like, I'm a shit programmer, I'm OK admitting it. I'm still learning, you know, but when it comes to technical ability, like, first off, it's kind of like it is very context specific, you know. So if I were if I were to try to like, interview into like a back end software engineering team, they would be like, oh, man, you're terrible. You're awful like you. You don't deserve to have a job in this field. If I were to interview into like a heavy, hard core research team where it's all like, you know, postdocs who did really smart super mathie stuff, they'd probably also go, man, you're an idiot, you're a clown, your code is passable, but bad culture fit, you know? So I think it's just it's kind of like dating, really, you know, like, you know, first off. Right. Like the right opportunity has to be right for both of you have to be right for the business and it has to be right for you at that moment.

[01:09:01] And that's usually not it's not a statement as to your value and core as a person and also what your value add is to the field and the team. So I would separate that out. And I think the other party is like technical capability, like it's something that can be worked on. So if we look to sort of like, ah, you know, like our brother and sister cousin adjacent fields like software engineering is someone is a front end web developer. Are they a terrible web developer because they don't know, like no jobs or react? Probably not. But in that area, it's certainly acceptable for, first off, people to have multiple sort of skill sets, technical skill sets, and to also display kind of different levels of capability. And because some people are early in their careers, some people are later. So, I mean, that's the one thing is I would sort of like keep in mind that, like, technical ability is a context specific. Right. Secondly, you know, where you are now is not where you are going to be like even a year from now. And also that something that does help, though, in science is to be really kind of crystal clear on what is the next thing you want to work on and develop, because it is kind of like you could boil the ocean.

[01:10:12] I think it's rather kind of going like, OK, here are all the skills I need to learn. It's always a question of sort of like what is the next bottleneck that is preventing you from, like sort of going further in the in the field, you know? So, for example, if your issue is like modeling, then you probably don't need to be working on your SQL skills. You probably should. And you don't need to be learning multiple languages. You should just pick one language, whether it's, ah, Python, Java, C++, and you should like build models in that area. If it's deploying, then you can kind of like focus on that. You probably don't need to go back all the way to like your idea. Right. As long as you're working on like, you know, what is this bottleneck that is preventing me from doing X work? And you're sort of crystal clear on what the work you want to be doing is, then it gets easier. But definitely like if you get rejected because of, quote, technical capability. Right. I understand it's a skill you can keep growing. It is fine. Right. We're all in different journeys to keep going after that bridge you go.

[01:11:07] So that's interesting. You say that these skills are context specific. I was talking to Andy Hunt, author of The Pragmatic Programmer. You interviewed him earlier this week. And we're talking about the drive, this model, which is a spectrum of competence and expertize. And you can be extremely expert at one thing, but not expert about another thing. Sounds quite obvious, but the way he broke it down, the interview was awesome. And he also answered my question, which which was do I have to be an expert programmer to think like an expert programmer? And he said, no, you don't have to. Um, so that was comforting. So definitely great. Great, great insights there. Thank you very much. Greg, I know you wanted to say something.

[01:11:48] If it was just a quick comment for Jeff. I thought I heard you said that, you know, Python or something like that feels like you're already comfortable manipulating data and using a tool that not everybody.

[01:12:02] They can do and learn from that, I'm sensing a little bit of what you call a part of imposter syndrome here that we all feel at some point. And I can tell you that you're probably the best place to stay ahead of the pack of people who call themselves or feel like they're technical.

[01:12:22] And what I can tell you is if you focus on gaining industry knowledge, you will be so comfortable with tackling what needs to be done from the technical side to solve these business problems. So the more business savvy you are, the better you can communicate with business folks, identify their problems, then you can work backwards to figure out you need technical skills that you need to solve them. And with passion, you will be able to do that for sure. I've sat in rooms with folks who grilled me on the business side grill. The other one's on business questions and things like that. And I can tell you, the data scientist, the SD software development engineers who stayed ahead of the pack, who were those who could absorb those business logics so fast and could figure out on the fly what kind of technical solution they could bring to the table so we can all align quickly to tackle a problem. So keep, keep keep up with the business knowledge, understandings, and you'll be good.

[01:13:25] They have you got this? So Avery says specialized with a bunch of exclamation points. I don't know man. Like I'm all about the generalist. Like I say, generalize to stay a future proof and specialize on demand.

[01:13:42] Yeah, I'm with yeah I'm, I'm a generalist as well. But I do think if you're, you know, you're trying to break into the space and you're nervous about finding the right role or you want to like a really high paying job, I think specializing can help you get there, because if someone if someone spent, like it sounds like already Data scientist or close to Data scientists and they spent like, you know, two or three or six months, like mastering, deploying and Data engineering, like they can get a really specialized job and data science that's that's pretty high paying. And and you'd surpass me as a data scientist. Like, I cannot do that stuff. So, yeah, I totally agree that it's I think it's really fun to be a generalist because I like to try to get my hands on everything. And I think people think that the you know, like we talked about Data science is really wide. There's that notion of the full stack Data scientist. That's the unicorn. Right. And no one really is that. But it's kind of fun to try to be that. But I often think, man, if I just got really good at one thing, that would be cool, too. I'd be like the master at that one thing. So it definitely is a trade off. There's pros and cons.

[01:14:47] Yeah. Like one thing I'm all about is just becoming the type of person that you cannot go to school to become. So finding these unique intersections of skill sets and getting good at that and then really good at that, try to move on to another unique intersection and just try to get there before other people do.

[01:15:05] Just yeah, I really like what you just said. Did not notice that. Yeah, I'm always here. I'm just in the closet. I think I really like what you just said. That really resonated with most of the value of my career has come from things that school never taught me. I think that's a superpower for all of us. What are the things that school can't teach you? Shop, right?

[01:15:26] I mean, seriously, though, I think that experience caught you on a path then for, you know, into more of like a kind of a technical career. Right. So, yeah, just all these skills. Any sense at the time?

[01:15:39] Yeah, I just I was just going to add, this is something who in my friends that are young enough to be my kids, almost my grandkids ask me, you know, what should I do? This should do that. I said, listen, learn the basics, but really master understanding the concepts because you of forget most of it, but you can hold on mentally to the concepts. And then once you have the concepts, you can relearn the specifics very fast. And then once you've amassed understanding several concepts, it's not too hard to go off and learn new concepts. But I remember teaching control system design five different semesters at the university level. I had to review it each time before I taught it, but each time I reviewed it, it went faster and faster and faster and I could leverage from old things. But boy, if you focus on the concepts and you get a few concepts mastered and the world's your oyster, and then you guys know to once you've gotten pretty good at one programing language, super easy to pick up a new one.

[01:16:46] Fundamental, I agree. When I hire, I tend to wait on like, you know, or center on your ability to learn how to learn, like overweighed one skill of anybody I work with, hire, whatever. Are you good at continuously picking things up? And getting really good at it, and that's a skill in itself, so I think some really good resources time on the mental models aspect, which it kind of touching on, like Charlie Munger mentioned, for he's got a really there's a really good book by my Desert Island book is actually for Charlie's Almanac, which is a book of his talks. And he talks about mental models, the ability to the needs, like basically maybe 90 different mental models out there that you need to master, which is a lot. But once you have those and you become Yoda ish, I guess the absolute number and another good book is Super Thinking.

[01:17:33] That's that's a book. Good book. Learning How to Learn is a huge skill. You're absolutely right. Which is why I interviewed Dr. Barbara Oakley and why the conversation I had with Andy Hunt was heavily geared towards his book, Pragmatic Thinking and Learning. And like to Tom's point, the fundamentals, the basics. Do I go back to that shit all the time? And just here is proof of how they plan on spending my summer. I absolutely hate textbooks. Like I don't do textbooks, but I'll fuck with comic books all day. So I have the Cartoon Guide to Microeconomics, Cartoon Guide to Philosophy, Cartoon Guide to Physics. And this is all summer reading. And then the same thing with calculus, statistics and various other concepts, just revisiting the fundamentals and making sure I understand them intuitively and don't lose touch with them and just try to make it fun for myself.

[01:18:22] So, yeah, I think I think the only area where like like, like there's a balance between generals and specialists, I think my and for everyone that's they have to determine kind of like at their particular point in time. Right. Like do they diversify versus you know, like drill down on one thing. I think my my number one vice that's been consistent though is never stay married to like one tool because like I've been on like a bunch of these. I mean, you have to balance. Right, like your individual sort of fiduciary responsibility to yourself with the needs of the industry. But like I've seen some post right. Where like at least like for some of the software developer forums I've been a part of where people are like, oh, you know, I I'm like, I have these and these skills and I can't get a job. And it's like and it's like, OK, well, you know, like they've just been married to that, like, tech stack for like the past 15, 20 years. And then it becomes it becomes very hard, you know, for them to like sort of do a complete 180. And so on the one hand, like, employers are going to say they want like X, Y, Z, you know. But I think everyone has a fiduciary responsibility himself to no one, like, still kind of keep up to date with what the market wants. And if you identify that the market is kind of shifting in terms of skills, you know, as sort of your responsibility to kind of like be open to that, but also understanding that, like, you know, a lot of times when employers put those skills on the list, they're not saying you need to have everything. They're more like saying like you should have one of like each group, like one of each category. Right. Like I have a friend who she's an animal researcher over at Apple.

[01:19:55] She had never in her life written a single code or a single line of like C++ in Java. She had other languages, like in her toolbelt. And like, you know, they were like they hired her and they basically paid for her to go to, like, do you like a Java C++ course?

[01:20:11] You know, she's been there for like three or four years. Right. So, like, there's a lot of companies for whom you might get stressed out looking at those job descriptions, job descriptions. But what they're really asking is like, do you have experience solving these problems and and working with these mental paradigms? Right. Like if you've worked with C++, it's probably you can carry over to some of the other, like JVM languages relatively easier than if you've only worked with SQL and same thing. Or I like if you've worked with like Python. Well, I was gonna say you care to. Ah, but you know, I tried doing both. It was a rough transition. I know Avery is like shaking his head like. No, that's right. But that's really the most important part is like getting just enough of the skills that you can kind of like, go deep on them, learn the concepts, but then also making sure you're still staying open to like what the market is saying in terms of like what we need. Because there is a reason. Right. There's a reason why you would use a GM language versus like a non GM language, you know, so I think that's what they always say is like don't like stay absolutely married to a skillset and don't let, like, the market just, you know, pass you by like 20 years. Yeah, that was yeah. I've seen a lot posts like that recently, unfortunately.

[01:21:18] Data covid so and such great, great advice. Great topics. Anybody have any last minute questions. Man it's been pretty amazing. Amazing Friday with you guys. Any last minute question before we wrap it up Tom. I mean it.

[01:21:31] Oh no I yeah. No I'm sorry. I'm kind of brainless at the moment, but I did love the session.

[01:21:39] Yeah, no, this is great. Thank you so much for asking some amazing questions and, you know, really appreciate having you guys here. If you guys have a interview coming up and you absolutely, positively want to tune into the episode I released on the podcast today with Evan Pilet, we talk about the science of successful interviewing. Unfortunately, my audio was messed up during that. But that's OK because. Don't say anything important. He does so check that interview out. A lot of great insights and and tips on how to progress through the interview process. Don't forget the Data Community Content Creators Award. Please go vote, please. Just like it on LinkedIn. I'd really appreciate if you guys help us do that. Just share that link, spread the word. And don't forget Sunday I have an office hour sponsored by Comet Emelle. Comet Emelle is awesome. Gideon is awesome. He gave away pretty much had a GPU on the clouds. Mikiko, if you want to get you on Google Cloud, check out something I shared earlier. I'm thinking was yesterday or the day before. I can't remember.

[01:22:44] But yeah, he's hooking up some good cloud Computershare to match what a must watch video and the chat is this young crazy guy living out in the woods when he was an undergrad is that young crazy guy you come, oh, no, no, no, I'm not a super human like this guy will have to go watch it to figure it out.

[01:23:02] And Taylor and Taylor, homeless college student, lives in a tent in the snow. That is what this is called. All right. This will be shared on the show as well for all. It's my favorite then, by the way.

[01:23:18] Sorry, I'm so weird, everyone. I was born this way.

[01:23:23] I love it. This is great, guys. Thank you so much for hanging out. See you guys next week. Same time, same place. And for anybody listening, that's in Europe. I don't know if you guys do like the time change thing like we do here in North America, but the time had changed. I think we went forward by an hour or so. Um, sorry if I messed up. You got a schedule, but come an hour earlier or. Yeah. Is it an hour earlier than what you normally used to. Yes. Coming out earlier than what you normally used to. I guys take care, have a good rest of the weekend. Remember you've got one life on this planet. Why not try to do something big? Cheers, everyone. Thanks for hosting.