Speaker 1: 0:02

you have taste in a way that's meaningful to software people hello, I'm bill gates I would I would recommend uh typescript.

Speaker 2: 0:14

Yeah, it writes a lot of code for me and usually it's slightly wrong. I'm reminded, incidentally, of rust here, rust.

Speaker 1: 0:24

This almost makes me happy that I didn't become a supermod. Cooper and Netties.

Speaker 2: 0:31

Well, I'm sorry guys, I don't know what's going on.

Speaker 1: 0:34

Thank you for the opportunity to speak to you today about large neural networks. It's really an honor to be here.

Speaker 2: 0:39

Rust Data Topics. Welcome to the Data Topics podcast.

Speaker 1: 1:11

Hello and welcome to Data Topics Podcast there. Actually, I'm not sure if we're linked in, but um prefer to leave your comment or questions or send us an email via data topics at dataio. Today is the 15th of october of 2024. My name is, I'll be hosting you today and I'm joined by the one and only bart hi. Hey, uh, no guests. Today we also have a different setup. I don't know if actually people that are listening to us will be able to tell the difference, if it's going to be better quality or worse quality the audio?

Speaker 2: 1:29

I'm curious about that, yeah.

Speaker 1: 1:31

Indeed, but the people watching for sure don't notice the difference that we are recording remote. So something new for this second season, let's say of Data Topics Unplugged, but something that you'll be well. I think people can expect more of this as well.

Speaker 2: 1:46

But you'll maybe have to experiment, fine-tune and tweak a little bit going forward.

Speaker 1: 1:52

Indeed, indeed, indeed, indeed. So I guess you could say, Bart, that we are working from home.

Speaker 2: 1:58

Indeed.

Speaker 1: 2:00

Talking about working from home. I see here working from home. I see that was pretty smooth. I didn't plan this. Working from home is powering productivity. What is this up?

Speaker 2: 2:14

What's up with this part? It's a report from the IMF and it's about basically research into the effects on productivity due to working from home, and their findings is basically that they see a five-fold increase in remote work since the pandemic, which I think, I guess, not surprising, right? Yes, not surprising at all. But also that there is a significant change in productivity, one that has been that is very singular in recent years. There has not been such a shock to productivity in recent years than working from home has been. I don't know the exact numbers. Let me just quickly see if I can.

Speaker 1: 3:09

But then like so basically they're saying that working from home is more productive, or people are more productive working from home.

Speaker 2: 3:15

Exactly, yeah, yeah, it was a huge shock to productivity and a good sense that productivity went up. That is something that they haven't seen, such a big positive shock, in recent years. And they say that both hybrid and remote work increased workforce participation.

Speaker 1: 3:32

Also very Workforce participation.

Speaker 2: 3:36

It's that like what percentage of the potential workforce is actually working, and in that case that's even more than just productivity. It's also meaning like the people that actually can work. So that also has a positive effect on disabled individuals, caregiver, rural workers that were not able to participate in the labor force before. Cool, would you make sense, then? Which makes sense Also, apparently, a big shift in office space needs, where there are a bit of a transformation on city centers, freeing up real estate for residential use, these city trends I can't really pinpoint it to a practical example close by my door, but that's what the report says. This is a positive one. Well, I think in macroeconomics terms it's a very positive one, not sure if it's for the individual. Always positive, though.

Speaker 1: 4:32

What do you mean? Why wouldn't it be positive?

Speaker 2: 4:34

To me, working from home is also like it's very hard to a bit safeguard the boundaries between work and life. It's very easy to work a bit later. It's very easy to work a bit later. It's very easy to start a bit earlier. I mean it takes a bit of discipline to manage that. Well, yeah, and I think that also has an effect on productivity. I wonder if that is in a good sense, of course.

Speaker 1: 5:06

Yeah, yeah, yeah maybe it's not like, maybe it's like one workday, but maybe people are putting 10 hours on a workday now, right, so they're more, they produce more, but not because they are more efficient, it's just because they're actually stretching a bit. I felt that a bit the beginning of corona, actually, because it was just like it was a bit new, and I also feel like having a space for working makes a big difference. You know, I think when corona just started, like, yeah, I took the corner, the edge of my, my dining table, but then having the laptop always there present, nowadays I have an office right, so it's like you close the door, like it actually makes a big difference. So that's a good, cool, cool. And also, well, not cool. Well, maybe cool, not sure.

Speaker 1: 5:49

One thing that I do think is good is because now I feel like I have another argument. You know some people says, oh, yeah, you need to be at the office like four days a week. I was like, yeah, but we need to. You know, study show the productivity is actually high. And it's also interesting because, uh, somewhat recently, right, so I didn't prepare this, but maybe I just googled quickly an article that amazon tells staff to get back to the office. Here it says five days a week. I'm not sure if it's actually five days a week, so this is from the 17th of september. So amazon actually was telling people that they needed to go back to the office this year, right, which I think surprised a lot of people. I think you made some noise as well, like, is it Maybe? Well, I don't know what arguments would you have for telling people to go back to the office?

Speaker 2: 6:41

I think the blatant one is that you want to shrink your workforce but you don't want to pay a fee to fire people. I think that is fun. I think another one and it's also a fair one is that if you're not good at things like project management follow-up, if you don't have mature practices in place, a return to office is like an easy solution where it feels that you might become more productive. Whether or not it has long-term benefits is a big question, of course, but I have the feeling that a lot of these companies that are announced, they grew very, very quickly during pandemic years and they probably hired too many people, and this is a way to reduce your workforce but while not motivating people to find another job.

Speaker 1: 7:32

Exactly, yeah, yeah, yeah, yeah, I heard that argument. I also heard that it's easier to keep the culture if you're in person, which I kind of. Well, I think it's easier, I don't think it's impossible. And I also heard the argument of office spaces that they have like empty offices, and there's also possible that it's a mix of all the things, right, but yeah, I thought it was well interesting in the sense that it surprised me, but I wouldn't want to be in that situation. To be honest, I think, well, at least five days a week, I don't, I don't mind going to the office, but I also appreciate a bit of flexibility, especially if, like where I'm originally from, sao Paulo, there's a lot of traffic. So to go to the office also means you're spending like some hours of your day in traffic, right, which is, yeah, it's stressful as well. There's also the consequences, but, yeah, things can go wrong. Things could always be worse, right?

Speaker 2: 8:33

There's actually an interesting website. It's called hubblehqcom and they have an overview like a page I will add to the show. If you Google for companies, workplace strategies, it will HQ Like you have a list of a lot of a big list of companies and their stance on remote work.

Speaker 1: 8:56

It's not this?

Speaker 2: 8:57

No, it's not. That's on my website. It's in our private chat here and you can open it.

Speaker 1: 9:04

Thank you, yeah, in our private chat here and you can open it.

Speaker 2: 9:08

Thank you, yeah, for example, morgan Stanley, the five days at the office, microsoft two to three days at the office, jp Morgan five days. Amazon, five days. Canva stands on the role, which also makes sense. Not all roles are suited to do remotely.

Speaker 1: 9:31

Yeah, that's very true, but it's interesting, I didn't know. This information was all public.

Speaker 2: 9:36

Twitter five days. Wow, Okay Cool. It's interesting to see how the space is evolving.

Speaker 1: 9:45

Indeed, indeed, all right, what else we got here? Zapata AI.

Speaker 2: 9:55

What is Zapata AI, bart? Well, to be very honest, I don't think that I really heard about them. They are on the stock market and they ceased their operation, I think a day or yesterday. In other words, they went bankrupt. They don't pay their people anymore, and Zapata AI is a software company that builds industrial AI solutions. That was more or less when I went to the archives of the website, because it's no longer online. Industrial AI solutions using quantum mechanics, quantum computing sorry, using quantum computing and I think, with that combination of things, to the surprise of absolutely no one, they went bankrupt. I don't think you can build like it's too early to build a successful quantum computing software company. I think hardware is something else. I mean, that's still a big bet, but software, we're not there yet, right?

Speaker 1: 11:03

Yeah, I think it was Especially for industrial applications for AI. Do you know how long they've?

Speaker 2: 11:09

existed.

Speaker 1: 11:10

No.

Speaker 2: 11:10

I don't know, I don't know actually. But it feels like they just threw a lot of hypeers together and then they hired actually very smart people and then the hype curve kind of died out and then they ran out of funding or something it's crazy that they were even on the stock market.

Speaker 1: 11:28

Maybe it was like Hail Mary, I think what they should have done.

Speaker 2: 11:32

To be a bit more sustainable is to also mention the block chain somewhere.

Speaker 1: 11:36

NFTs blockchain. Just like step by step there. See what sticks. Did you have any money on Zapata iBart? No, but maybe I should have.

Speaker 2: 11:44

I you have any money on Zapata AI Bart? No, but maybe I should have. I should have shorted it. Predict a downfall and there's a lot of smart people in the market there.

Speaker 1: 11:52

So you know, that's true, there's an opportunity for someone.

Speaker 2: 12:03

If anyone is looking for a quantum AI engineer.

Speaker 1: 12:04

This is your moment. All the people you know One, two, that's it. Cool, cool, Maybe some things that are more mature, I would say, maybe Python. Python is pretty mature, I would say, but they still have some gotchas. So well, maybe my personal. Why I thought of this this week is I was working with overloads in Python. Do you know overloads Like type hints, Bart?

Speaker 2: 12:37

Explain it to me.

Speaker 1: 12:39

So basically, yes, so basically type hints. It's as the name suggests. It's just a hint, it's not something that is enforced by the Python interpreter. It basically says if you have a function maybe I'll share the screen here If you have a function and you have X X can be anything in Python technically but basically say, well, actually this X is going to be, I'm expecting it to be a floating point number or an integer or a string or something, right? So it started as that, like just as a hint, just to kind of help the developers. And then more people kind of started saying, hey, we can actually use these things to do nice things, right? So by then it's a great example. And then you can also specify the return types and all these things. So overloads is Overloads, basically, if you have a function.

Speaker 2: 13:32

So the hints basically it's not a force, but it basically is a hint to the user or to the interpreter. You can probably expect this type of variable right, exactly exactly.

Speaker 1: 13:46

And again, that's kind of what it is now. And in Python, one thing you can do is having, like mypy or type checkers, which basically will look at all your code and say, well, this function returns a string and the output of this function goes in this other function, which is expects integer. So that will be an error, right, like you can expect something will go wrong, because if these functions are using the way that they intended, then that will be an issue. So it's also a way to catch some errors or bugs, et cetera, et cetera. So, in a ways like, if you're familiar with JavaScript and TypeScript, python with MyPy is a bit closer to TypeScript in a way, and Python in itself is just without type ints and without the type checker is more like JavaScript, let's say. So I suggest that's to say that if you're using, if you are, if you do want your Python code to be MyPy checked, you can also expect some headaches, right, like you need to make sure that everything fits, and one of the things is this overload, right. So the idea here is that you have and I have something on the screen for people just listening you have a function called process and then there's like an input argument, etc. Etc.

Speaker 1: 14:49

And one thing that is can be considered a code smell is if your function return one type or another, so it can be a string or an engineer. Right, because usually it's not like random. Usually this happens because the inputs are different, somehow, right? Right, so if X is an integer, the output is gonna be a string. If X is a floating point number, maybe it'll be an integer, right, something like that. So basically Python has this overload decorators which basically just kind of is supposed to cover all these cases.

Speaker 1: 15:18

Right, so say, overload process, if response is equal to none, then the result, the output of the function, will also be none. If response is an integer, then the output is going to be a tuple of integer and string, et cetera, et cetera. Right, so you have basically a whole bunch of empty function definitions just for the types with the at overload, and then you have the actual function implementation later. So I'm working with my Pychecked code and I spent, I think, hours just trying to figure out why the overload thing wasn't working. And in the end it's because if you have arguments with default values MyPy somehow, for some reason, kind of like it it accounts for that. So even if I have uh like a full argument that the default value is one, my pine doesn't always pick that up.

Speaker 1: 16:13

So I can still get errors in my default value of one without specifying the type, without specifying that with typing, yeah, but like, the thing is like, when I call the function, I didn't specify the full value and then, for some reason, my pi cannot infer the type or cannot infer that that argument, that new value that I am passing, is not full, it's not something else. And basically it's like, to be very honest, up to this day I'm not 100% sure why that happens or how to fix it, but I know how to avoid it. Basically, let's say, and the thing that I came across, well, the thing that made me think I was like yeah, what the fuck is this? Was this WTF Python? Have you ever heard of WTF Python? No, so basically you can kind of think of it as a book. It is translated in many other languages and it's kind of like a way to learn python with just like gotchas kind of.

Speaker 1: 17:08

I don't know if it's really to learn python, but basically I have a whole bunch of examples like uh, I don't know. Uh, you can, we can put this on the show notes as well, and people can kind of have a look if they, if they want to. So, for example, if you want to check your quality strings or if you want to get ID of strings, sometimes you get the same values or some, I mean and there are other things as well that you can do. There's a whole list here, right? So the example here is that you have, I don't know, a is equal to WTF, b is equal to WTF, and if A is B, you get that. That's true. But then if you repeat the process, you get that it's false. I make this a bit bigger for you. What the fuck?

Speaker 1: 17:46

Yeah, exactly, and then it's like basically a whole bunch of those things right, which can create like these cases where you're like debugging like a full day, Exactly yeah exactly, and it's really like if you look at the internals of Python, you can kind of understand why this happens and why that happens, but it's really not obvious at first, right, and then basically they kind of go over a whole bunch of examples. So actually, even if someone wants to contribute just by adding your translation to your language, here, for example, dutch, dutch is not here. So yeah, I think it could be also a fun way to dive in Python and also learn a bit about the internals.

Speaker 2: 18:25

Right, a lot of stuff here, cool, the things to stay away from maybe maybe just to come back very quickly to the overloading, like, like you, it allows you because I remember you last episode or maybe the episode before, like, keeping less in your brain is better. It doesn't. Overloading go the other way, because overloading allows you to define a method in tons of ways. You can have this method that has an integer as an input and an integer as a return, but you can also define the same method that does a string and gives back an integer, like whatever right, you can have all these definitions of the same method. Doesn't that make that very complex? Is it smart to do this? Why would you do this? What is a good use case for this?

Speaker 1: 19:17

It's a good. I thought about that as well. I thought also in the JavaScript versus TypeScript argument, right? So I think the first thing you can say is like, well, if you write it down, the compiler will not the compiler, but the interpreter will help you. Like the completion, right, you will know, like, if you call the function later and you have a variable, you can just hover over your ID and you know that argument the output value is going to be X or Y.

Speaker 2: 19:44

So in that sense, you keep less in your head.

Speaker 1: 19:50

But why don't you just make another method for it? You could, but then there's also a trade-off, right, like you can have. You have two methods that do almost the same thing but they just have expect different input arguments. Then maybe you kind of duplicate things and you have to keep more in your head in that sense, right, I think it's like complicated things and you have to keep more in your head in that sense, right. I think it's like, and I think even there's a I forgot the name of the guy, I think it's dhh the guy that criticized typescript. They said that typescript takes away the fun of javascript. Then then I remember in the article very vaguely yeah, he was talking about like type gymnastics that you kind of have to do to make sure that the but you very much needs to do this, yeah.

Speaker 1: 20:25

In.

Speaker 2: 20:25

TypeScript. You need to do this. Yeah, yeah, yeah.

Speaker 1: 20:27

In MyPy as well.

Speaker 2: 20:29

I think I would argue much less.

Speaker 1: 20:32

Ah, really yeah.

Speaker 2: 20:34

The problem with TypeScript is that you use all like you have tons and tons and tons of third-party libraries and they define their own types, and then you use something from this library and then you get a typing error because you're not defining the right type and then you need to dive into this code of this third-party library. Like what do they call this type exactly? And like it's very unintuitive, like at some point you're just doing it because it has to be done, otherwise you get a warning.

Speaker 1: 21:02

Yeah, yeah, yeah, I mean, yeah, it has to be done, otherwise you get a warning. Yeah, yeah, yeah, I mean, yeah, I think. But that last part, the sentence, I think, is the same in my part, like sometimes you just do it because, yeah, otherwise my part will keep complaining. But I think, and that's the thing, and that's why I also like the metric of keeping less things in your head, because I also feel like you give some leeway, for you know, like in the beginning, yeah, if you add some types, that's good because you're keeping less things, it's there. But when you start doing gymnastics to just satisfy the interpreter and no one understand what he does, then I think it also goes a bit in the other way. That is also not good, right, and I think the overload, I think, is maybe a bit more in between. I do see the the value in it, but, yeah, if you have like three or four different implementations, then, yeah, I also think it's not not good. But, yeah, that's a good question, good point.

Speaker 1: 21:49

Some other things that I came across. Well, I actually had come across this a while ago, but I hadn't had the opportunity to try, and I think you're familiar with this one. Why? Data profiling. You heard of this one right.

Speaker 2: 22:04

Yeah, I saw it passing by somewhere. I never used it.

Speaker 1: 22:08

Yeah, I saw it as well and I hadn't used it, but I thought I had an opportunity and that's why I wanted to bring it now.

Speaker 2: 22:15

Maybe, before we go into the data profiling, can you explain what data profiling is.

Speaker 1: 22:21

Yes, yes, yes. So well, maybe one thing that I used to do a lot well, I guess you could still do. You get a bunch of data and you want to kind of see what's there. You want to get a feel for what data you have. A lot of times in machine learning projects, that's the first step, the EDA right, exploratory Data Analysis. So you basically want to see what kind of columns you have, basically want to see what kind of columns you have, what kind of types they have, what are the max min? Do you have any imbalances? Do you have any correlations? So one of the things that's very common, you see, is the pandasdescribe right. So you basically get counts like if this is a, if you have M or F, for male, female, you can see what do you have more? How many counts do you have et right?

Speaker 2: 23:02

so if you have a bundles data frame, you call, describe on it like you said. Like you said, you get some statistics for every column. It's not a little bit like how does? This data look like right exactly.

Speaker 1: 23:11

Distribution of this data the max 50 medium average standard deviation, exactly, and that's what I usually think, at least. When you say like profiling, right, like you usually think of, like, yeah, you kind of get a sense, like you kind of get some more descriptives all over a table, basically, um, and this is a, a tool that makes it easier. So today, I guess, if I have a new data set and I'm gonna feel for it, the first thing I'm gonna do is probably run something like this um, so what this returns actually can be an HTML, it can be a widget if you're on notebooks or something, and basically it will do all these things that we're mentioning, but they will also do some nice plots. It will also look at correlation between columns. It would also do, yeah, it have warnings as well. So, like, this column has only unique values, this column has only constant values, this column has 100% correlation with this column, so it's a very like yeah, and it's interactive as well, right? So for people following the video, you can see here like a little GIF, right.

Speaker 2: 24:19

It's quite a like when you show them on the screen, like it's quite an extensive report you got right, indeed, indeed so it's quite an extensive report.

Speaker 1: 24:24

you got right Indeed, indeed. So it's way more than what you would see with just describe right, yeah, you would do a lot of steps in Pandas to get the same insights, exactly, exactly.

Speaker 1: 24:36

So, yeah, and it's really this is all you do. You have the Pandas data frame and then you have a profile report object and you just pass the data frame. It has to be a Pandas data frame, that's the only thing, and then you can play around with it. You can see are there correlations, are there any warnings or all these things. So if you're doing exploratory data analysis, I think that's a very good I mean, I had a good experience with it, let's say and something that I'll probably even go as far as to say that I wouldn't use any describe, or I would just probably spend the first hour just kind of looking and clicking through the different things that they have here.

Speaker 2: 25:14

I'm quickly going through the website and I think, like what is missing or maybe I haven't come to it yet is for people that are just like for experimentation, for non-configuration data like that. There is like a simple hosted service that just I upload a CSV or an Excel or whatever and you get this in your browser. I think that would be cool.

Speaker 1: 25:40

This one. I think you can get on your browser, or at least you can save it like it's an HTML thing, because, yeah, it is an HTML report.

Speaker 2: 25:46

To be cool, there would be like a small hit of pages.

Speaker 1: 25:49

Yeah, I did I think that would be super, super, super easy to do. Actually, what's the automated behind? Yeah, there's quite a lot of stuff, so I tried it a bit. I think there's more features. I didn't try, but, uh, it's cool, cool, um, so that's one. And one other one that I learned not that long ago is this one called sweet biz. I never heard of this one. This one I hadn't heard either. Um, it's still well. They have an update of november 2023, so a year ago. So it's not. Yeah, maybe it's not the most uh up to date, but it's still maintaining that is, forget the terms.

Speaker 2: 26:29

That's old, like I also. Also. Always when I open up like a github repository of an open source, I always look a bit like what was the last yeah, the last commit and when you're down here like oh, it was a year ago and I think, oh, it must be dead, yeah that's the thing.

Speaker 1: 26:42

That's also what I usually like when they have. Also, the first note here is like oh yeah, maybe you should use this one. This is a maintenance mode or something, so I haven't tried this one yet, but it also looks cool. Let's see. So that one doesn't have a nice readme. But this is a bit how the the dashboard looks like. So it looks a bit sexier. I feel it looks a bit sexier. I feel it looks a bit sexier, I agree, but I don't know if it's, I don't know if actually it's nicer, right.

Speaker 1: 27:10

But again, it's another profiling tool, right, so you can have your data frame there and then you can kind of click through and then kind of compare. Also, they show here the training test. I guess, even if you have predictions on your machine learning model, I guess you could also throw it there to see if there are any correlations between the input and output. See, the balance on all these things, right, even for bias, right? Maybe your model is saying that people from a certain gender are more likely to have the outcome Y or whatever, right, so could be cool. I think it's a very I think it's also a good use case for these things, for these very interactive things.

Speaker 1: 27:49

You know, I just kind of quickly see what's there so I haven't used this one, but I think next time I have a use case for this, I will give it a try. What do you think, bart?

Speaker 2: 28:01

uh, I would go for the first one, though it's a bit more, more mature.

Speaker 1: 28:08

You know, I had the data books Maybe the first one is a good.

Speaker 2: 28:13

I haven't tried. Well, I did try it, but not for something actually useful a Python browser. There wasn't A cool project. Yeah, a cool project to see if you can build this fully in browser, fully client side. Yeah, that's true.

Speaker 1: 28:26

I don't know if we could do that. I'm not sure. I'm not sure how hard it would be Because also Pandas uses NumPy right and then NumPy is a C, and then can you bring that to the?

Speaker 2: 28:36

browser. I don't know if there is a Pandas implementation that is.

Speaker 1: 28:40

Yeah, yeah, another thing so this has to be pundits. So that's the other thing I didn't really like, let's say, because sometimes loading data in pundits takes more than like pollers, for example, and actually my I was working with pollers, which is written in Ross, and I had to actually convert from pollers to pundits just to use this tool. So I still and I had to actually convert from Polar to Pandas just to use this tool. So it's still like minimal step, but like I thought it would be nice to use Polar. And one thing that we also talked a while ago with the normals, which was like a Polar's API, that you have different backends.

Speaker 1: 29:15

When we talked about Ibis as well, and since then I listened to a podcast with the creator of normals and he was saying the difference between EBS and Narwhals is that EBS is more for consumers, so people that are manipulating, creating transformations on data frames, and Narwhals is more catered towards people that are building a tool or a package that expects to manipulate or simple transformations of data frames. So, for example, wide data profiling would be a good candidate to use narwhals. And by using narwhals, you automatically support pollers, funders and other things.

Speaker 2: 29:50

It would be interesting because they're both Ibis and narwhals. They're an API to your data or to your backend, basically, but it would be interesting to see what are the design choices that are different if you cater yourself to our end users, like data explorers versus solution builders. Yeah, interesting.

Speaker 1: 30:15

Yeah, because I mean I listened to all the podcasts, right, I wasn't on the discussion, but I listened. I was like, okay, that's interesting. But I was also wondering what are the practical implications, right, Like what kind of trade-offs they make on the design.

Speaker 2: 30:28

Exactly yeah.

Speaker 1: 30:29

Interesting and yeah like. Yeah, narwhals is a subset of Polaris as well, and like Polaris, is built in rust right. Yeah, I was going to say right, and apparently not everyone is a fan of rust. Not everyone is a fan of rust. Not everyone is a fan of Rust, not everyone. Oh, okay, because I saw here my negative views on Rust and I thought you meant yours Exactly. I love that. Can I see you're outsourcing this? I tend to agree.

Speaker 2: 31:12

It's actually an older article. It's a bit of an older article.

Speaker 1: 31:14

October a year ago I like how the design is, all of the. I don't even know if it's a blog post right, but I think it is. I think it's I like this.

Speaker 2: 31:24

I like it from blog post. Maybe, yeah, I could segue for next. Uh, anyways, the the. The blog post by uh chris, uh chris don is the website I'm not sure chris is called my negative views on and I think, well, he gives a lot of good points as well. I think the biggest pain points I tend to agree with is that. And why are we discussing this now? By the way, because this was written a year ago is because it was very much trending on lobsters.

Speaker 1: 32:02

Lobsteris.

Speaker 2: 32:04

Lobsteris, lobsteris, lobsteris.

Speaker 1: 32:09

It is also very popular for Rust projects, right yeah?

Speaker 2: 32:15

But one of the things that one of his gripes is that it's very much hyped, very much hyped, but at the same time it comes in at a complexity that is more or less comparable to c plus plus or haskell a bit.

Speaker 1: 32:30

What he's mentioning, um, definitely not python or golang, uh there's like this, definitely not python or golang, you mean in terms of, in terms of complexity complexity, yes, but quickly building something, I would argue, is way easier in something like Python or Golang. I mean I agree, but I'm also wondering. I mean it's also different use cases, right? I mean it's Python and Rails.

Speaker 2: 33:01

That's the problem. That is the problem. Everybody these days says Well, I think we were also there a bit past the hype curve, but a year ago, and especially two years ago, every new project had to be in Rust and he's also touching a bit on this because we were discussing Polos. He calls it the rewrite fallacy, where I built this in Rust, so it is x times faster than the original. Think about Rust, for example. Think about UD.

Speaker 2: 33:32

And the thing is the rewrite fallacy, how he's explaining it, and I agree there as well. If, let's say, you would have rewritten poetry I'm not sure what poetry is.

Speaker 1: 33:49

I'm not sure.

Speaker 2: 33:51

If you would have rewritten poetry with a focus on making it as performant as it can be, whatever backend language you use for it would have been more performant than poetry is now. If you do a full rewrite for something to be performant, youauder is now. If you do a full rewrite for something to be performant, you're going to end up there, right? So I don't think there's probably something like. You need to think about it less in Rust and it will be probably a bit more performant. But just saying that that is what you get out as the result is a bit like taking some shortcuts, right?

Speaker 1: 34:27

Yeah, yeah, yeah Also, yeah, there's the whole argument of yeah, it's faster to run, but also it's longer to write, so it's like there's also everything's a trade-off, right, if performance is not a bottleneck. Yeah, okay, you spend five years rewriting something and you gain half a second a day?

Speaker 2: 34:45

wow, and he mentions another, a number of other things, and I think a lot of it is that's because it's hyped, and is it? It's often chosen as the tool uh, while it's maybe not the best tool for the job that I think, that I agree, I that I agree. I think that I agree. I like how the quote-unquote friendly community.

Speaker 1: 35:06

I like this one. I don't know what this is about.

Speaker 2: 35:13

Yeah, I think there's a lot of. I think there is a very friendly community, a very active community. At the same time, it's a bit ups and downs. There's also a lot of discussion on how the whole project is governed, which I think tends to come with something that grows very rapidly, right? I don't think that necessarily Rust or the core group is just something that grows so quickly that no one really knew how to manage it.

Speaker 1: 35:42

So I guess you could say this is a hot take.

Speaker 2: 35:44

No, I think it's a hot take. Yeah, I think maybe we're a bit past it. I think it would have been a very hot take two years ago.

Speaker 1: 35:52

Yeah, I think I agree, but at the same time with the hype of UV in the Python world, that is true. So you guys are the hot. Hot, hot, hot hot.

Speaker 2: 36:08

We don't have our sound effects with us now.

Speaker 1: 36:10

Yeah, I did Something. We'll iterate next time, but yeah, cool. And talking about websites, blog domains, I guess it's the gift cities, or yeah, what was the second way again? Oh yeah, we were talking about blocks, right Say how, like how, how you like this blog design and maybe for people just listening, it's a. It really looks like you open your it's very minimal right. But I think even the font is the default. It's like times new Roman it's white with black text.

Speaker 2: 36:44

Yeah, it's like Times New Roman. It's white with black text.

Speaker 1: 36:46

Yeah, less is more. There's no header, there's a footer, but then it really just looks like he opened a new file, he saved it as a html like indexhtml, and then he just started writing stuff and that's it. No CSS, nothing.

Speaker 2: 36:59

I enjoy it. I enjoy it.

Speaker 1: 37:01

Yeah, you're minimal, but I think you like retro stuff. That's why you like this. You're minimal and retro and that's why you like it.

Speaker 2: 37:08

But I wouldn't call this retro necessarily, I would call it minimal.

Speaker 1: 37:12

But it's like yeah, I get what you're saying. It's not retro because the font is like modern, but that's the default one.

Speaker 2: 37:18

I guess it looks like it's the default one, at least yeah, is that your definition of retro, for like no moment, there's, there's not a gradient in there somewhere like it's retro?

Speaker 1: 37:32

no, not, not retro, but like it's on its way. No, it's like, if it's a spectrum, it is a bit more retro than you know the average website, wouldn't you say?

Speaker 2: 37:41

than the average website okay I'll follow you there.

Speaker 1: 37:45

What about websites?

Speaker 2: 37:51

Well, you have a blog, right. There is something on my name yes, you don't want too much attention Draw to it. How do you write the URL? No, but that's not the blog. Oh, you do, right I don't.

Speaker 1: 38:09

That's not the blog. Oh, you do right, I don't think that's the blog.

Speaker 2: 38:13

Morillio.

Speaker 1: 38:16

Morillio, yeah, it's your blog, right, it's not a blog.

Speaker 2: 38:21

It's like a more I would say maybe it's more of a bio, yeah, portfolio kind of.

Speaker 1: 38:28

It used to be actually a blog, right. I used to have, I mean, I think I don't even know if it's- more or less a block.

Speaker 2: 38:34

There's like no, no, they removed everything. They used to be articles, right, or am I completely?

Speaker 1: 38:41

no, it's something it was something separate, it was something I wrote one and I was like, yes, this is great, that's the one. I saw the one. Yeah, yeah, yeah. And then I was like, yeah, okay, okay, okay, okay, something different but what? What uh stack that you built is in this one, yeah, the marilio one, yeah, um, astro, astro with the yeah, nextjs and Tailwind and stuff.

Speaker 2: 39:09

It might actually look a bit like I was playing with the idea of creating my own personal space on the interwebs.

Speaker 1: 39:18

Like your personal domain, no, your professional domain. Ah, you have my professional domain.

Speaker 2: 39:23

It's something else, but actually I reserved a URL yesterday domain squatting no, no, it's with the goal to put something.

Speaker 1: 39:33

It's space that's a good one.

Speaker 2: 39:41

And then I thought about, because I had a blog back in the day. I had a blog that was visited quite a lot, quite a lot, and I not necessarily want to restart a blog, but at least, like, have something where I can sometimes add something to.

Speaker 1: 39:58

And the problem is hold on, don't move so quickly. How many people were in your blog before?

Speaker 2: 40:04

I don't know to be honest and is the blog still up.

Speaker 1: 40:07

Can I still check it out, what you're writing?

Speaker 2: 40:10

you can find it on the what is it called? The internet archive fishyoperationscom. It was very much focused on R programming back in the days yeah, big time R fishy what fishy operations, operations, but big time, mario, fishy, fishy.

Speaker 2: 40:28

what fishy operations operations? Wow, um, but what I? What? The discussion I would like to have with you, like, like there are so many stacks out there and I would like to go for something very simple, but like there is Yugo, there is Pelican, there is Astro Astro no, astro, astro, astro, yeah, which I think is too complex. I don't think you need Nextjs and stuff like this for something like a minimal blog. And there are tons of these static slides. What is the very famous one that GitHub pages use it? Jekyll, jekyll.

Speaker 1: 41:00

That's in Ruby right, Ruby on Rails or something.

Speaker 2: 41:02

Yeah, and I actually found another cool one. It's called Publi P-U-B-L-I-I, which really is like an application, a desktop application. You make it in the UI and it generates, like it syncs, the files to your GitHub or whatever you want, or an S3 bucket or whatever you want. So what stack should I go for, Marino? What are considerations you take when you go for a static site?

Speaker 1: 41:32

I have no idea if I'm looking for something, the only thing. A lot of times I'm doing this to try to learn, but the only thing is that I want to be able to write a markdown for me, exactly the same.

Speaker 2: 41:46

And then you often end up like the static site generators, which basically means like you have markdown. That's really the only yeah. Well, for me, exactly the same, for me, exactly. And then you often end up like the static site generators, which basically means like you have markdown, markdown pages and that can either be articles or like full pages on your website, and you have a cli tool that, from those markdown pages, generates a static site. Yes, the static site generator comes from um and basically like a set of HTML files that you can then upload to wherever to host it. So, very minimal, all a bit with our ups and downs, and I think I will go the Hugo way, which is a Go-based static site generator.

Speaker 2: 42:24

I'll keep you posted on that.

Speaker 1: 42:26

And Hugo is also. Hugo is not the sexiest and the hottest.

Speaker 2: 42:31

No, he's not the sexiest. That is a bit the thing. I always fall for what's the sexiest today, and then I regret it later, and I think Hugo has become the safe choice. Hugo used to be sexy, but he's no longer sexy yeah, but Hugo, how old is Hugo?

Speaker 1: 42:46

I would I'd like to say 8 years. Hugo used to be sexy, but he's no longer sexy, yeah but Hugo, how old is Hugo?

Speaker 2: 42:48

I would like to say eight years.

Speaker 1: 42:53

Ten years. Maybe that's super old. Huh, that's ancient for a.

Speaker 2: 42:58

I'm sure I'm wise.

Speaker 1: 43:01

Oh yeah, so you're going with Hugo.

Speaker 2: 43:07

Something else, because I was looking a bit into today and what I noticed. I went to my like an asset where I get some gifs, gifs, gifs a staple of the internet in other words gifs cities. It's called gifs cities gifscities, it's called.

Speaker 2: 43:28

GiveCities GiveCitiesorg it's actually like it is a place that hosts tons of GIFs from the old GeoCities pages For the GeoCities back in the day, back in the day GeoCities. Back in the day. Back in the day, geocities it was very easy to create your own page and you could customize the hell out of it and everybody had these retro. This was retro. These days it's retro like old school GIFs and stuff, and GIF Studies still holds hosts tons of these GIFs, and I noticed today. Today it's down Like you can press search and then you never get a result. So now I wonder who we need to notify about this, because this is painful, right.

Speaker 1: 44:21

Well, for me it's loading, but I guess it shouldn't be loading. It keeps loading.

Speaker 2: 44:24

Forever. Something is down. Whoa, there's no kid oh that's sad we need to uh notify the overlords of the the webs. That is down. Yeah, oh, hold on.

Speaker 1: 44:42

But like look, maybe you have your service availability. Wayback machine other internet archivist is temporarily offline. Please check your official account. There you go, bart. This found who you have to talk to.

Speaker 2: 44:53

We need to notify them on master clone. I'll do that, yeah.

Speaker 1: 44:58

I mean it's on the podcast now, so for sure they'll listen. That's true, For sure you'll get there, right yeah.

Speaker 2: 45:11

But you want to start a blog again. But um no, not necessarily a blog just a bit of a personal space. I'm also a bit intrigued by, uh, micro blogging what is my like?

Speaker 1: 45:20

it's a little bit like it sounds like twitter blogging is a it sounds like blogging is a drug, and then you just micro-do it.

Speaker 2: 45:27

No, but you have just a thought or a picture or whatever and it ends up on your blog. But you own it a bit more and it's not based on likes and upvotes. I see you have some of these pages I think it's actually called microblog which is probably the most famous one, and you can automatically cross-post to Mastodon and Twitter and LinkedIn. But you own and you can automatically cross post to Mastodon and Twitter and LinkedIn you own your page which basically holds all your things. I think it's cool. I'll keep you posted on this, how this goes.

Speaker 1: 46:00

There we go Also here. Microblogging for the people that are, personal blogging that makes it easy to be social, interesting, interesting.

Speaker 2: 46:11

Interesting concept.

Speaker 1: 46:12

I guess, like the Today I Learned as well, can also be going there. You're also talking about component libraries. I saw this on a YouTube video in Fireship ShedCN. Have you heard of this? I've used it before. Yeah, I used it. So it's like well, do you want to explain what it is?

Speaker 2: 46:33

I'm not sure if I can give a good explanation. To be honest, it allows you to basically very quickly generate code for a bit of a predefined components that you can then use in, for example, nextjs or something, and it's AI generated Based on. Is it, though? I don't think everything. No, I thought it was.

Speaker 2: 46:59

No, I don't think so, to be honest. Maybe now it is. I mean, you can't do anything without something being AI generated, right, and I think it's a heavily built on tailwind, which is like season's framework, and it, uh, it generates code for you that you can then including your uh, your next projects, react projects or maybe even other types I've used in the context of next years I see, yeah, I think maybe maybe they and it saves you.

Speaker 1: 47:31

It saves you a lot of time yeah, yeah, I think this is maybe, maybe the because v0 is ai generated v0 is a ui component generator.

Speaker 2: 47:40

Uh ai generator from vercel yeah, yeah, probably leverage this there as well.

Speaker 1: 47:47

yeah, yeah, in documentation. Yeah, don't take it for it.

Speaker 2: 47:49

It's their core focus.

Speaker 1: 47:51

Yeah, yeah, but here it says open components in E0 for customization. So I guess you can still go there and customize it. Cool. And what I saw is that you can use the CLI, so you have a unique identifier for the component that you created and and then you can just kind of use your CLI and then that component will get slapped in your. But that's not the future for you then, because you tried it, because when the video really sounded like that was the future.

Speaker 2: 48:19

Well, how do you mean? It's not the future for next years or components of next years, components, I guess? I think a lot of people use it these?

Speaker 1: 48:26

days I saw this and I was like maybe I could be a content. I was looking at this like maybe I can be a front-end guy now I think you can be everything that you want to be really wow you heard it and I probably leave today with a smile in my face wow see, that's where we need the clapping voice.

Speaker 1: 48:45

You know the clapping sounds just going with a smile on my face. That's when we need the clapping voice. The clapping sounds All right, cool. You know one thing that every blog needs and every website needs Bart Domains Domains, in fact, murilloio is a very popular domain for tech. In fact, datatopicsio.

Speaker 2: 49:10

Are you not worried about your personal space on the interwebs? Morilioio.

Speaker 1: 49:17

On morilio.

Speaker 2: 49:19

Morilio.

Speaker 1: 49:23

I'm not sure. Why would you be?

Speaker 2: 49:26

worried this week about the future of your io domain.

Speaker 1: 49:32

Because apparently there was a change in the geopolitical space. So the io, so actually a lot I don't think all, but I think a lot of these domains they actually represent countries. So in Belgium it's be, in Brazil it's com pr, in UK it's couk. So basically a lot of these domains they are attached to a country. Right, and io is no exception. Right, io is from the Indian Ocean.

Speaker 2: 50:11

Well, it's actually the British Indian Ocean Territory. Yes, yes, yes, yes.

Speaker 1: 50:18

And then here, you see here on the TLDR. But when the British government announced last week that it was transferring sovereignty of an island in the Indian Ocean to the country of Mauritius, gareth immediately realized its own implications the end of the io domain suffix, because this island was the last island that was still part of the British Indian Ocean territory that holds the io top domain.

Speaker 2: 50:46

Top domain is like the last part of the domain name, like io in this case, and with this geographic territory stopping to exist, the question becomes a bit what happens to the top level domain name? And the big challenge of this, uh, that's uh, this is here. Say what I'm gonna. I'm not sure how correct this is, but what I understand is that uh top level domains that are based on uh geographic regions, countries, whatever, um, they are the only ones, they are the only ones allowed to have two-letter top-level domain names. So let's say io be, because the easy solution to this is that it becomes a commercial top-level domain. Like you have app dev space. They are commercial top-level domains, but commercial top-level domains are not allowed to uh have two two letter uh top level domain names. Um, so I think for it to happen it will probably happen, because there's a lot of value entangled in the dot io domain name there needs to be an exception to this, to this standard yeah, there's a lot of tech companies that are on the io domain.

Speaker 2: 52:09

Well, and if we're at the risk of losing Merrillio, I mean, people will realize that we need an exception, right?

Speaker 1: 52:16

Well, again, I'm sure they're listening to this podcast, right? So I'm sure they're going to make the right decision.

Speaker 2: 52:21

So that's interesting because this I think no one, when choosing their domain name, at some point said yeah, but what is Great Britain going to do with this island?

Speaker 1: 52:32

I think a lot of people don't even think of oh yeah, this is from an island.

Speaker 2: 52:38

Yeah, that's true.

Speaker 1: 52:41

They're probably just like oh, that sounds cool, it's like input-output and stuff, right, maybe another thing I thought I had opened. That sounds cool. It's like input, output and stuff, Right, yeah, maybe another thing I'm trying to. I thought I had opened this. It's not. Let's see here. This is an article from November of last year, so about a year ago from now. It's also about domains from the mit technology review. So, um, it's about tokalaos tk domain. Have you heard of this?

Speaker 2: 53:16

uh, yeah, very much, oh yeah very much I think if you would have been my age but when was this actually? Uh, so why I know very much about it is that from the moment that I don't know, I think when I was 16 years old or something that is a very long time ago I was starting to play around a bit with hosting and stuff like this. I had very little money but tk domains you could get for free.

Speaker 1: 53:47

So everybody got a to get tk domain exactly, and that's a bit the the, that's a bit the article is also talking about, you know, if you have the guy that. So tokalow is like a very small uh country, so it's just 1400 uh people there in the population, but there was like millions of domains yeah, it was super easy to register.

Speaker 2: 54:10

I probably still is, I don't know, uh as was completely for free, so why would you not do it right?

Speaker 1: 54:15

exactly it's taking 25 million, um. So basically like, yeah, and it's also a business right, like if you can host stuff and all these things. And that was actually a dutch guy, I want to say from amsterdam actually. So you know, you know one of your people, bart. He was like oh yeah, I'll manage the domain for you, you know, I'll figure it out, just give me a bit of the profits and all these things. I think he wanted to do ads, I'm not sure. Yeah, so I need to refresh on the article. But basically he thought it would be a good idea for creating this domain for free and give it to people, and good idea for creating this domain for free and give it to people.

Speaker 1: 54:49

And he said I'll manage it for you, but then of course a lot of people started creating these domains, but also a lot of cyber criminals started doing this for phishing and all these things, and that's how this tiny island, the pacific island, this tiny pacific island, became the global capital of cybercrime. And um, yeah and yeah so what?

Speaker 2: 55:08

how did they misuse the?

Speaker 1: 55:11

dimension like spam here, so I'm just used probably to spy spam phishing in other.

Speaker 2: 55:17

Very easy to get an identity on the internet, maybe exactly.

Speaker 1: 55:20

Yeah, yeah, I think matter of there's a with. I think I don't know exactly and I think you know this better than I do but how this also ties with, like in email addresses and all these exactly, and I think you know this better than I do but how this also ties with email addresses and all these things.

Speaker 1: 55:29

Yeah, that's true. So I think, basically a lot of people. And again, because it's free, if you're spamming, if you're phishing, it's also a numbers game If 1% of the people actually react to your attempts. But then you just increase by a lot, right, like the amount of tries and the different types of domains. And because this was free, people just kind of went bananas, right, and there was no. You can scale up Exactly.

Speaker 2: 55:53

So if you Is the summary of all this that like if you get an email from princejerry at nigeriatk like you need to maybe be, a bit worried, we're always like Twice before you set your old checkings account right.

Speaker 1: 56:10

Indeed. Yeah, if you see tk you should raise an eyebrow.

Speaker 2: 56:15

If it's not barttk, then maybe it makes me wonder, if these are still active, that they would have expired to the ones that I created.

Speaker 1: 56:24

Yeah, that's true, I'm not sure, I'm not sure.

Speaker 1: 56:28

Yeah, but basically the the guy also. There was a bunch of drama because the guy disappeared as well. So the guy said he would manage. He set a lot of stuff up, but maybe he got like bored or something, so he just kind of left and people from tokalow they also kind of he also had a bad reputation for the island, right. So yeah, the article kind of goes through again all the different ins and outs and what happened, and actually there's an audio version of this as well. Here's also. There's a. Just looking here, I just forgot, but there's a guy named Vitaly here. I just wanted to highlight that in case Vitaly is listening as well.

Speaker 1: 57:08

One of our dear colleagues. Yes, but yeah, again, it's one of those examples of how domains are actually attached to countries, you know, to regions, and I don't think people think about that a lot. I didn't think. I mean, I still don't think a lot about that, right, but it's funny how a lot of things are rabbit holes when you start looking more into it and how, like in this case, a geopolitical change in the world kind of could affect the everyday developer, right. So something to bear in mind. What else do we have? What else do we have. What else do we have? Maybe one last thing maybe Ducks. You like ducks, right, bart, I like ducks. Yeah, data Topics also has a.

Speaker 2: 57:56

A duck logo.

Speaker 1: 57:57

A duck logo Some mascots right.

Speaker 2: 57:59

A duck is a mascot.

Speaker 1: 58:02

So this is not something serious, but one thing that someone put. There is also some Gen AI stuff, because I don't think we can really go through a whole episode without talking about Gen AI.

Speaker 2: 58:14

I have one more Gen AI stuff to end with after this.

Speaker 1: 58:17

Okay, perfect. So basically, this is a Duck Impostor game. So this is, I think, powered by MotherDuck at Coalesce. So MotherDuck is the people from, people that basically is like a managed DuckDB by MotherDuck at Coalesce. So MotherDuck is the people from, yeah, people that basically is like a managed DuckDB right on the cloud and stuff, and DuckDB is the single machine, all-app engine, right, super performant, that you can also use it in Python and whatnot. And their logo. So MotherDuck, their logo, is also Duck-related and DuckDB as well, right? So they created this MotherDuck logo. No, motherduck game, their logo is also duck related and that could be as well, right? So they created this, uh, mother duck logo. Yeah, mother ducking. Sorry, duck imposter game. It's late. So basically you have like two images here and you just kind of say what is gen ai and what is not oh, wow yeah, yeah, and I don't think I've ever passed the, the third level.

Speaker 1: 59:01

Let's say um yeah, so see, I got one right it's that.

Speaker 2: 59:07

It's just super difficult. Yeah, you're showing the images impossible, so basically yeah for what does this tell us about society, if we can't keep?

Speaker 1: 59:16

exactly right. So maybe for people listening it's like it's just two images the ducks right and then you just kind of have to say what's the the imposter here? You have to click right or left, and if you click the right one you have another round, so you increase your score by one, and if you get it wrong, then your game is over. Um, yeah, maybe this also makes me think of a game that I was shown when I was younger. Um, it was japanese or chinese, and it was basically two people. You have to guess which one is chinese and which one is Japanese but yeah, it's interesting, it's so difficult.

Speaker 2: 59:56

Right, the ducks are not Chinese, japanese, but the ducks are. The ducks are super difficult when you show it on the screen wow, you're really deflecting the second topic yeah, I am Maybe also to give some more context as well.

Speaker 1: 1:00:13

I'm half-blood Japanese and I think that's why growing up a lot of people in Brazil that's why you weren't allowed to discuss these things. That's why I'm allowed. I can even show you real quick here. You can see, you have the two countries. If the person is japanese, pretty good at it. Um, I'll send this link to you, barter, later and you can play and you can tell me afterwards. We'll review the results next episode.

Speaker 2: 1:00:39

You know uh, ducks, yeah. So if that's now on the downfall of society when they narrate it, it will go like the first thing is, we couldn't distinguish ducks from real ones anymore, fake from real ducks anymore well, ducks, I think, is getting really good yeah, if the ducks goes, shit hits the fan if the ducks goes, I'm not sure I'd follow you if the ducks go, if we can't distinguish them from real or fake anymore?

Speaker 1: 1:01:11

What is real? What is real Abarth.

Speaker 2: 1:01:14

And then linked to that. My last point is that Adobe Firefly, which is Adobe's Gen AI engine, which generates, amongst other things, ducks, right, but still images they used to generate. They also generated vector art, actually, which is also quite interesting because it's something that's a bit more difficult, and they now also launched uh, or announced you can join a waiting list an ai video generator and the. The videos are quite impressive, like, in my eyes, sora level. Impressive that we saw some videos from OpenAI earlier this year that has still, by the way, not gone public, oh really.

Speaker 1: 1:01:57

It's still secret. Yeah, so I put some here on the screen for us. They're short.

Speaker 2: 1:02:05

Right, they're shorter than the Sora videos. What about what you can see? Looks, uh, it looks impressive I think the nice thing about adobe is that you, they have like a whole workflow around this. These things like, uh, they're this case probably with adobe premiere, um, so you can take it much further than just the image or just the video.

Speaker 1: 1:02:30

This is really cool. It looks very well polished as well in the style Actually, yeah, so this is still the way it is.

Speaker 2: 1:02:39

They very much picked the nice videos for their website, right.

Speaker 1: 1:02:42

I was going to say that Curious to see where it goes. Yeah, and when is this going to come out? Do you know?

Speaker 2: 1:02:48

No, I will sign up to join the waiting list.

Speaker 1: 1:02:53

Wow, but this is impressive. You see the sand movement as well. That is something super hard to simulate.

Speaker 2: 1:03:02

And trained on content that they actually have the license for to do.

Speaker 1: 1:03:09

They explicitly mentioned that. Yeah, because we saw like I think I don't know, some while ago. We also talked about this.

Speaker 2: 1:03:16

I think they didn't give a single F about anything when it comes to copyright. I think when you ask Zora to generate the whole Simpsons, series and episodes Exactly and everything step by step.

Speaker 1: 1:03:35

It gives you a whole episode, right, yeah, yeah, yeah, okay, that's very cool. I'm also curious how this there's one example here add a little extra to a scene, so generate atmospheric elements like smoke, a lens, flare or water to add to a scene, so I guess it's like a video and then you can. It's a bit like inpainting.

Speaker 2: 1:03:58

when it comes to images, Like you say, like okay, this area of the image, in this case video, I want to add something to it.

Speaker 1: 1:04:05

Yeah, but it Cool huh, yeah, definitely. I'm really curious how it will work. Very impressive, Cool. I'm curious to hear. Maybe we can produce something or demo on the pod whenever this is available. Yeah, that's good. Very curious, Very curious. Alrighty and I think that's it. That's a pod for today. That's a pod. Anything else you want to share any words of inspiration?

Speaker 2: 1:04:44

are you putting me a bit on the spot? That's what I do live free. You're putting me a bit on the spot, that's what. I do Live free.

Speaker 1: 1:04:52

Wow, actually you mentioned it. I'm a good one, you can be whatever you want. No, is that what you said? What did you?

Speaker 2: 1:04:58

say I don't know, we lost the magic. I think it's good to end on this.

Speaker 1: 1:05:03

Okay, all right, we'll stop here.

Speaker 2: 1:05:08

Alex may have to do some magic Now we're going to sit still for 30 seconds so we can add it in the outro sound.

Speaker 1: 1:05:17

Yes, all right Cool.

Speaker 2: 1:05:20

Bye everyone.

Speaker 1: 1:05:21

You have taste.

Speaker 2: 1:05:23

In a way that's meaningful to software people. Hello, I'm Bill Gates.

Speaker 1: 1:05:31

I would recommend.

Speaker 2: 1:05:32

TypeScript yeah, it writes a lot of code for me and usually it's slightly wrong. I'm reminded, incidentally, of Rust here, rust.

Speaker 1: 1:05:43

This almost makes me happy that I didn't become a supermod. Cooper and Nettie.

Speaker 2: 1:05:49

Well, I'm sorry guys, I don't know what's going on.

Speaker 1: 1:05:53

Thank you for the opportunity to speak to you today about large neural networks. It's really an honor to be here.

Speaker 2: 1:05:58

Rust Data topics. Welcome to the Data Topics Podcast.