UX for AI

EP. 87 - Creativity in the Age of AI: Language, Technology, and Innovation w/ Bob Van Luijt

• Bonanza Studios

Share episode

Send us a text

In this thought-provoking episode of UX for AI, host Behrad engages with Bob Van Luijt in a philosophical exploration of creativity, language, and technology in the AI era. Bob, the founder of Weaviate (a vector database company), shares insights that blend his unique background in both music and technology.

The conversation begins with Bob reflecting on his identity not as a musician or coder specifically, but as someone who "likes to make stuff" - emphasizing his passion for the creative process across different domains. He discusses his TED talk about language's fascinating ability to express infinite new ideas with a finite alphabet, drawing parallels between literature and software as systems of formal language.

A central theme emerges around formal languages and their evolution alongside technology. Bob explains how spelling itself emerged from the practical needs of the printing press, drawing an analogy to how AI prompting languages are evolving today. As technology advances, we naturally develop more structured ways to communicate with it - a pattern repeating throughout history.

The discussion turns to anxieties within creative communities about AI replacing human creativity. Bob dismisses these concerns, noting that similar fears have accompanied every technological advancement throughout history. He articulates a nuanced perspective: machine creativity and human creativity are fundamentally different, with humans uniquely able to assign aesthetic value to creative works. He challenges the notion that people in traditional "creative" roles have exclusive ownership of creativity, pointing out that creativity exists across all professions and endeavors.

In the latter part of the conversation, Bob explains vector embeddings - the technology behind his company Weaviate. He describes how these multidimensional representations of information allow AI systems to understand relationships between concepts without explicit programming. By converting unstructured data (like text, images, and audio) into vector embeddings, organizations can unlock new capabilities for information retrieval and generation.

The episode concludes with Bob's perspective on enterprise AI adoption, offering unexpected advice: companies should focus on bringing in talented Gen Z employees who intuitively understand emerging technologies. He suggests that the tension between experienced professionals and youth with "antennas" for what's new creates a productive dynamic for innovation.

Throughout the conversation, both host and guest express excitement about the current moment as an unprecedented opportunity for creation and innovation. Their infectious enthusiasm culminates in a call to action for listeners to stop theorizing and start building, as the barriers to creation have never been lower in the age of AI.

Interested in joining the podcast? DM Behrad on LinkedIn:
https://www.linkedin.com/in/behradmirafshar/

This podcast is made by Bonanza Studios, Germany’s Premier Digital Design Studio:
https://www.bonanza-studios.com/

00:00:02:08 - 00:00:05:09
Narrator
Welcome to UX for AI.

00:00:05:11 - 00:00:10:23
Behrad
So, Bob, are you a musician or are you a coder? Let's just start from here.

00:00:11:00 - 00:00:32:20
Bob Van Luijt
Oh, I'm neither, so I'm, I just I like to make stuff and, you know, and there's there are many things that you can use to make stuff. But some of the things I'm just not really good at and, and when it comes to, to software and music, that was something that I, I guess you would call that was gifted in.

00:00:32:22 - 00:00:48:16
Bob Van Luijt
So, but I mean, I love to cook, but I'm just not a really great cook, so I would have to open a restaurant, but if I would have been a great cook, then I might have be a maker in, in food. So it's a it's a it's really focusing on the creative process, what I enjoy the most.

00:00:48:21 - 00:01:09:13
Behrad
Okay. Because basically that was the the topic of your Ted, Ted talk as well. How we can basically, Activate, more creative world that we can create new things. Right? So, maybe you give me a rundown of your talk on that front. That was really fascinating to me.

00:01:09:15 - 00:01:33:13
Bob Van Luijt
Yeah. So when they invited me to give the talk, of course, I said, what's the. Is that, like, a topic or anything? And they're like, no, no, no, you can talk about anything you want. And I have a tremendous fascination with language and not language from the perspectives of, and literature. I mean, I'm, I'm also interested in the literature, but just as a reader.

00:01:33:15 - 00:01:56:04
Bob Van Luijt
But I meant more from a, from a structural point of view. Or you know, how later in, in that we've seen language, philosophy and those kind of things, how language is structured. And one of the things that I find fascinating in society, and I was combining that with how we use software, is that, our language is constantly changing.

00:01:56:06 - 00:02:26:05
Bob Van Luijt
We can constantly come up with new ideas and concepts, but we have a finite amount of, letters in the alphabet that that's fascinating. So if we look at the as the Western alphabet, which, you know, just has a less than 30, letters in it, but with less than 30 letters, we are able to constantly, over time, express new ideas and make new things, which I find fascinating.

00:02:26:05 - 00:02:58:05
Bob Van Luijt
Now, what I did in my Ted talk was I said, what other creative? And so you see it in literature. So in literature we capture that. For example. But there's something else where we use language to write stuff down, and that's software, because, software as a formal language is how we define things. And so the point that I was wanting to make is like anything that we can dream up or that I said that we can talk about is something we can capture in software.

00:02:58:07 - 00:03:20:00
Bob Van Luijt
And so the relation that, I mean, like we can make anything we want in the digital world because in the physical world, you know, we're limited to the atoms and the and how we structure the, that kind of stuff. Right. But in the digital world, we only use language. And we, we know how to express our ideas in language so we can build anything we want in a digital world.

00:03:20:03 - 00:03:51:23
Behrad
That's very fascinating. I, I was thinking about this last night when I was playing around with Paul's and a bit between windsurf, which is now everyone talks about it. Then I, paused for a second and I realized that, I was pondering upon the skeptic camp when it comes to whether I can code versus the, optimistic camp.

00:03:52:01 - 00:04:21:15
Behrad
Then I realized that, okay, what I is really good at, I is really good at, basically parsing a topic that you offer to into pieces, understand, that creates connections based on, you know, and then do it over and over, lead through backpropagation and gradient descent to get to a level that say, okay, if you give me new data, I can, with great confidence.

00:04:21:15 - 00:04:48:07
Behrad
I can make a good correlation between the new data and what you train me on. And then I realize that, okay, if I is really good with language, programing is a language to. So if it can do a really good job with English, it will do a good job with Python, with any other language that, is at the disposal of programmers.

00:04:48:07 - 00:05:09:01
Behrad
So that was an moment. Probably is something very primitive to you because you're dealing with code in a, I, you know, I think that's your expertise, that's your basic bread and butter. But for me as a non-technical guy, then I realized that, okay, if it doesn't really good job for an English, it's going to do a good job on any other language that you present.

00:05:09:01 - 00:05:46:12
Bob Van Luijt
I that latest conclusion I agree with. I think understanding programing with AI is, that it can be simply explained by an example. And, and it will also explain why programing languages exist and why I think that programing languages will stay after I and I, I can I can explain that by an example. So. If I ask an AI model, generate a website for me, one page website, one page.

00:05:46:14 - 00:06:08:00
Bob Van Luijt
And the only thing I want is in CSS and in styling, right? I want a red circle. That's that's just the prompt that I'm presenting it. Chances that the model are gonna generate a piece of code or in this case, if we if we really, you know, go into the details, then HTML is not code. It's a markup language.

00:06:08:00 - 00:06:29:13
Bob Van Luijt
But so if it generates a markup language with a with a red circle in the middle, then I can load that circle in my browser and then I will see a red circle. But now the thing is I get to get all the circles to big. So I go back to my model is like, okay, can you generate that again, make it a little smaller, and then the model is going to of course I can do that.

00:06:29:13 - 00:07:00:06
Bob Van Luijt
So it makes a new and it makes a little bit smaller. And I load it again and I look at it again. And then I will go back to the master. No no no no no not that small. Little bit bigger. And so what we do, the only way to make the model generate a circle in icMl. For me that is exactly to my liking is I need to define a little bit and it just say like it needs to be on a the smallest atomic level on my screen are pixels.

00:07:00:07 - 00:07:33:13
Bob Van Luijt
So I need to tell it, well, make a red circle that is, I don't know, 100 pixels wide. And now while I'm at it, let's talk about the other subjective thing in that prompt as well. Right. We actually have a formal language of defining the colors in the, on our machines, and that's in hexadecimal. Cullercoats. So I can ask the model, generate a 100 pixel wide red dots.

00:07:33:15 - 00:08:11:18
Bob Van Luijt
But if I don't like the type of red before I know it. I'm writing a programing language because I'm saying in its HTML of this version, use wrath of this color code with the white of pixels that is this big. And then slowly, but certainly a programing language emerges and we already start to see certain projects. An example that comes to mind is the Espie from Sanford that people start to formalize, prompting to solve this problem, and another way to think about it.

00:08:11:18 - 00:08:16:16
Bob Van Luijt
And then I'll I'll stop talking. Is that.

00:08:16:18 - 00:08:44:06
Bob Van Luijt
If I prompt a model, generate an HTML page with a red circle, then I've only done a good job as a as a prompt engineer. If I can take that prompt, give it to another model, and then it generates the exact same thing, because if it doesn't generate the exact same thing, then I don't know what's going to generate and then I can use it and the like it.

00:08:44:06 - 00:09:04:11
Bob Van Luijt
We call that formal languages. And a programing language is a formal language. And what we've basically done with each other is at the lowest level on the computer. It's binary language. We can read that. That's ones and zeros. So we can't we can't read it. So we need to be making a little bit less abstract. So then you have assembly language.

00:09:04:11 - 00:09:27:12
Bob Van Luijt
Well by the time you wrote something in assembly language, that's, that's a lot of time just to get a little bit dumbed. And then so then we got the C language and then we got C plus plus Java and then JavaScript, Python. But the distance we're making from the raw ones and zeros, since we're generating all the way from Python.

00:09:27:12 - 00:09:50:10
Bob Van Luijt
And now even up there to the models, we make the abstraction so large that at some point it's so large that what's being generated here is not what we try to express over here. So we need some formality in the language we use. If we talk to computers and make them do something that's beautiful.

00:09:50:12 - 00:10:19:11
Behrad
So your what you're prophesying that there are already some words being done, but there will be more structures, guidance and guidelines and instruction potentially in the in that language, the widespread language for prompting that makes it a lot more efficient to speak, to interact with all these different algorithms.

00:10:19:11 - 00:10:46:08
Bob Van Luijt
Exactly. And that is just that's the same thing with even if you go before computers with spelling, the reason we have spelling is because if people wanted to use the printing press, we needed to have agreement on how to print stuff. That's why we say that's formalization of the language. If you look at the, the old letters of President Lincoln, right.

00:10:46:10 - 00:11:17:00
Bob Van Luijt
The the guy makes and I'm doing and making air quotes here, he's making spelling errors left, right and center. That's not because he was not a good writer. That was just because before that, spelling wasn't a thing. So that only became later when we started to introduce technology. So as long as we have technology, we will formalize the language we use to make the technology do what we want, because it's it's still not at a human level that it understand that.

00:11:17:00 - 00:11:47:18
Bob Van Luijt
So even if I would ask you, there are a picture of a person playing a tuba, right? I now see a person playing a tuba. You see a person playing a tuba. But those are the things are completely different. So if we now ask a model, generate an image of a person playing a tuba, then it will generate something and we might like it or we don't like it, but if we do not like it, we need to somehow explain to the model how we want to change the picture.

00:11:47:20 - 00:11:49:10
Bob Van Luijt
And that's a formal language.

00:11:49:12 - 00:12:29:14
Behrad
That's that was a beautiful analogy that you mentioned that we didn't feel the need for introducing spelling before the typewriter came into a creation. Then we realized that it's a highly efficient process that everyone has the same, as different perception of how to write pretentious ideas. For example, we need we need to systematize it so we to be able to write a letter that if two people write the same letter with the same content, it will be within the same length.

00:12:29:16 - 00:12:58:05
Bob Van Luijt
Yes. And if you would write me a handwritten letter and you would say, Bob, I enjoyed the conversation, but it was very pretentious. And you make a huge spelling error in pretentious because you make you write it pretty. Us, right? I know what you mean. When I read that. That's because of the context in the sentence and because you wrote it like with a pen, you know, it doesn't matter.

00:12:58:06 - 00:13:18:17
Bob Van Luijt
But if we have type centers, which we got with the printing press, then you go like, oh, I've made a spelling error, so now we can print them. So we now need the spelling that we have that we, you know, align our way of writing stuff down with, how the printing presses operate. And that's a, and it's fine.

00:13:18:19 - 00:13:53:18
Bob Van Luijt
And that's a, that's a, that technology does that all the time. But, you know, let me tell you, let me tell you another story related to this, which I find even more interesting. I've been to a conference a couple of years ago before Covid was it was in Iceland, in Reykjavik, and the name of the conference, I don't know if it still exists, is or was material, and the idea was to have people like us who are into software and technology go come to Reykjavik and get presentations from people who do not work in software at all.

00:13:53:20 - 00:14:20:03
Bob Van Luijt
So people who were working with wool, people were working with wood people. So who have these apprenticeships and hundreds of years of of knowledge because in software we build something and then we throw it away. And they gave us the it's two great examples in the opening speech. It's the one thing. The first one was this. Do you remember a couple of years ago that you had these websites with flash?

00:14:20:05 - 00:14:45:03
Bob Van Luijt
Do you remember that all these. Yes. All these people using flash, building flash, etc. then it got killed. So that knowledge of all the things that we could build with flash, it's just it's gone. It doesn't exist anymore. And the second example that they gave is that they showed a picture of a huge crossbow that DaVinci had drawn.

00:14:45:05 - 00:15:08:17
Bob Van Luijt
And DaVinci explains in the drawing how to build the thing. But the problem is that we don't know how to build it today, because the knowledge about woodworking on how to build such a thing is gone, is lost. So we're going. So. So what I just said with the formal language, we capture all kinds of information. Everything is going so fast.

00:15:08:18 - 00:15:30:12
Bob Van Luijt
And if you don't like it anymore, we just throw it away. But then people come up with creative ideas, generate knowledge e.g. in flash or everything that's happening now what I but then we don't care about it anymore and it just is gone. And it's so much creativity, so much knowledge is just gone. And I find that very interesting concepts.

00:15:30:14 - 00:16:02:07
Behrad
We talk a lot about right now, creativity. And then there is now the elephant in the room. I, I would like to get your take on. So there is a lot of anxiety in the creative, communities about whether AI is going to replace the, but I personally, you know, See a lot of areas that I cannot touch, right?

00:16:02:07 - 00:16:31:23
Behrad
When it comes to, basically, you cannot replicate the types of creativity that you humans bring to the equation. Right? I would love to get your brain, how do you differentiate human creativity from what I does? And how is AI complementing humans when it comes to the creative work?

00:16:32:00 - 00:16:43:06
Bob Van Luijt
Those are excellent questions. Let me, let me let me try to because I have many thoughts on it. But let me try to answer defiantly. So.

00:16:43:08 - 00:17:09:08
Bob Van Luijt
As a quick preamble to answer questions people in the creative industries and I've studied music, so I know a lot of people in that industry who complain about new technologies taking over their jobs. Again, air quotes I don't care about that one bit because that is burnout that has been happening since day one of technology in the creative industries.

00:17:09:10 - 00:17:36:12
Bob Van Luijt
So when people invented the vinyl records and the record player, people who played in restaurants were concerned that they would put on a record rather than have a band play and so on and so forth. When I was studying music, there was a lot of software. So for example, orchestral, music that, you know, all of a sudden I could create this whole orchestral pieces just by using the software and in sound are really good and, especially for movies and that kind of stuff.

00:17:36:12 - 00:17:57:12
Bob Van Luijt
And then people were like, oh, you know, and I'll play the violin, and they're not going to ask me to, you know, to, to, to, to to play an orchestra anymore. That is just since forever. And then. Yes, I'm sorry, but that's just how that's just how the world's been working since we invented fire. So I you have to be creative and and try to solve that problem.

00:17:57:12 - 00:18:23:20
Bob Van Luijt
Right. So that's one thing. And sorry for being so slow to have such a bold statement, but I really think that and, the second thing is, to and now to, to the answer, but it's related to, to the questions that you're asking. Is that so we need to ask ourselves the question of the, the esthetic quality of creativity.

00:18:24:02 - 00:18:44:23
Bob Van Luijt
Right. So we measure we try to measure. We haven't figured out like a universal measurement yet, but the language we use to explain if something is well or not, though, we say like how much esthetic value does that? And one of the things that we've learned in under humans, in esthetics is that it's always in the eye of the beholder.

00:18:45:01 - 00:19:09:12
Bob Van Luijt
So you might look at something and go like, oh, above this beautiful look at it, and I might look at it and it might look, right. It's very hard to, to, to align those kind of things. So if I, if, if there's a creative work, however we define it that's generated by a machine and you appreciate it, but then that's great.

00:19:09:12 - 00:19:36:16
Bob Van Luijt
Good for you. And if I appreciate it too, then that makes two of us. And if not, then, then it doesn't. Human creativity. So a machine can never do human creativity because it's never human. A human can't do elephant creativity because that's what an elephant does matter. So a human can do machine creativity because that's what a machine does.

00:19:36:18 - 00:20:08:19
Bob Van Luijt
The only difference with a machine, a human and an elephant is that as far as we know right now, the machine, I mean, I would say we have zero proof to assume that the machine can assign any form of esthetic value of what it's doing. Right. So you could argue, well, if it creates something that's beautiful in the eye of the beholder and, and it's seen as beautiful.

00:20:08:21 - 00:20:31:22
Bob Van Luijt
Okay, great. But it's not the machine doing it. We know that an elephant has some understanding of the world, and therefore could assign some tiny bit of esthetic value to what it's doing. And based on other mammals on the planet, humans are just really good at it, right? So we just do the lot. We do all the time.

00:20:32:00 - 00:20:59:10
Bob Van Luijt
And so. The word sits in the word human is human creativity. Right. So I mean machine cannot do human because machine is human, and human can do machine creativity because, you know, the human is not a machine. So that's how I would look at it. And we and the reason I made my first statement is because I sometimes think, especially when we talk about creativity.

00:20:59:12 - 00:21:10:02
Bob Van Luijt
Is we sometimes mix up.

00:21:10:04 - 00:21:41:06
Bob Van Luijt
People who have a job in what is seen as the arts or creativity musician, the designer, chef. That somehow that has like a right to certain creative elements. But I know a lot of business people that are more creative than musicians that I studied with. They just go to school. They happen to have the talent to play the flute, and they get very good at playing the flute.

00:21:41:08 - 00:22:11:11
Bob Van Luijt
But, you know, they couldn't play something interesting if their if their life would depend on it. Right? So. Everybody in every job can tap into that creativity. You can be a bank teller and be super creative. And how you do job. So that's what we need to think about. And that is measured in the esthetic value. So every time when you, you know, I can go to a shop to, to, to to buy a cup of coffee, go in there and go like, oh wow.

00:22:11:11 - 00:22:41:08
Bob Van Luijt
And tell the owner, this is such a beautiful shop. It's so nice. I want to buy my coffee here. That is an esthetic value that I assign to the coffee shop. You see. And and that there's the human to human, interaction and the machine doesn't necessarily play a role in that. And if then the coffee shop owner says, well, actually, you wouldn't believe it, pop, but I prompted a generative model design, a nice coffee shop, and I liked it so much that I turned my coffee shop into what this thing designs.

00:22:41:13 - 00:22:54:09
Bob Van Luijt
Okay, great. Good for you. But then it's still human to human interaction. That's that's giving it esthetic value. You see. And, I hope this is not too dense in information, but that's how I think about it now.

00:22:54:09 - 00:23:42:04
Behrad
I think the way the way I think it's a beautiful take and the way I perceive it is that no one has claim an ownership and pattern over creativity. Creativity is, the act of creativity is a multi-dimensional process that different forces feed into it. One of them is, I think technology has been, since the founding of the fire feed into the creativity or fueled the creativity and technology will evolve and there will be more types of technology, will feed and feed creativity.

00:23:42:06 - 00:24:16:09
Behrad
And it's up to humans to use these tools at their disposal to basically create works. Right. So if I, a designer or a musician, that doesn't give me the right to dictate what content, what kind of technology should we be using to create work? It's basically it's a free market of ideas and tools and let the best tools and this the best idea we.

00:24:16:11 - 00:24:56:23
Bob Van Luijt
Exactly. And so if what we call creatives, have an issue with, with techno technological advancements, I go like, yeah, it's not it's correlated, but it's not the same thing. It's just it's something different. So, I recently saw some, somebody the, And this, this guy made his life's work to work on and study all the work of Bach, but only that.

00:24:57:01 - 00:25:29:19
Bob Van Luijt
Nothing else, which I find beautiful. I find that such a wonderful endeavor in life. And he is concerned with how it was performed back in the day. He's concerned with the writings of Bach from back in the day, but he's not complaining about he's not only not complaining about the technology, he's using it all the time. He he he digitizes the music of Bach to to to have different ways to listen to it and to structure it, to analyze it.

00:25:29:21 - 00:25:53:16
Bob Van Luijt
He gets access to all the works, to, to the internet and so on and creates podcasts, what have you. And I think that's the right way of doing it. Right. And it's a, it's a, it's a, it's you're in a creative endeavor. Then you use it or you don't. But whatever you do, don't complain about it.

00:25:53:18 - 00:26:15:03
Bob Van Luijt
It's like it's like it's I'm also like if people complain about it, I'm also like, okay, so what? So we now just have to go. Everybody's like, hey guys, the AI stuff. Let's we're going to stop doing this. Just because my friend over here is a writer is not happy now. Of course not. Right. So it's okay. That's not what's going to work.

00:26:15:03 - 00:26:35:06
Bob Van Luijt
So use it leverage it or don't but don't complain about it. It's that sorry that I get so worked up about this, but it's like it because you know and you notice because you you this is such an exciting time. This is not the time to complain. This is the time to be excited.

00:26:35:08 - 00:26:39:19
Behrad
This is this is the time to stay up at night and figure, yes.

00:26:39:21 - 00:27:12:06
Bob Van Luijt
Yes, yes. And the beautiful thing is that he it keeps it, it keeps changing. I was, I was the and yes it goes in waves. So sometimes there's like a, there's like so when I was for example, when I was studying music, there was a little bit of down period. There was not a lot of new stuff happening, but actually a lot of people that I studied with, now became some people became well known artists right from my generation, like, Thundercat and that kind of stuff.

00:27:12:07 - 00:27:39:01
Bob Van Luijt
And then I'll do all kind of new stuff, right? So they go, oh, wow, this is amazing. And, and that's just a and they leverage technology. These people leverage technology left, right and center whilst being very creative musicians. So again, I 100% with you, it's like, this is the time to stay up at night, to be excited to talk to your friends about it, to build new businesses or new artworks or whatever you want to do.

00:27:39:03 - 00:27:40:03
Bob Van Luijt
It's just it's great.

00:27:40:05 - 00:28:10:09
Behrad
If I tell the audience what you do, they probably, after 28 minutes of talking about creativity, they wouldn't believe it. But for me, as a person who runs the Innovation Studio, you have to hope. Science with digital transform mission. I know how valuable what you are doing at VBA is. Maybe walk us through what we view is what problems trying to solve and why.

00:28:10:09 - 00:28:23:06
Behrad
Is it empowered, and especially now as companies have to by force or voluntarily move towards, data led, AI led operation?

00:28:23:11 - 00:28:45:18
Bob Van Luijt
It's maybe nice to answer this question from the perspective of the earlier conversation we had about language. So for a very long time, the question is like, how can we have machines do something with language? And and I forgot who said this and how long ago it was, but it is a long time ago. It might even be like 100 years ago already.

00:28:45:20 - 00:29:15:08
Bob Van Luijt
Somebody said, like, we can systematically understand language if we look at how what the relation is between words in a sentence. And the example that I always give is that if you have the words AI, the two words Eiffel and tower, Eiffel Tower, then in the sentence, chances are that you'll find Paris more closely somewhere in the sentence than London.

00:29:15:10 - 00:29:39:11
Bob Van Luijt
London might only come like a paragraph later. And when we got cloud computing, I mean, we got deep learning. So there was how we referred to AI before, we call it AI. People got back to this idea again and what they started to do, and it's for the people listening. It's how simple it is actually in in realities might surprise you.

00:29:39:13 - 00:30:01:05
Bob Van Luijt
People just started to select, let's go to the back to this point and just start to conference is going to count words. And we're going to put them in a huge Excel file. And it's not literally an Excel file, but for people to imagine like a large Excel file and so basically they said like, you know, how does Paris followed the Eiffel and Tower and then just, you know, on the big matrix.

00:30:01:07 - 00:30:17:18
Bob Van Luijt
The problem, however, was that if you do this for large data sets, then it gets slow quickly. I just think about using Excel. You know, the bigger Excel get, the slower we get. Now imagine you want to do this a whole Wikipedia right? That's just not going to work. So this idea of this concept was always there. And of placing it in the matrix.

00:30:17:18 - 00:30:33:01
Bob Van Luijt
Because if you have it in the matrix, we can draw it on a map, and then we can just make relations between these words on the map. And and so this idea was always had a concept that we couldn't do it because it just was too slow. And even with all the cloud computing, we couldn't do it.

00:30:33:03 - 00:30:53:01
Bob Van Luijt
And then with the whole deep learning thing happened again. And this we're not talking ten years ago this or little longer. This idea came about like, what if we use machine learning to predict this? So we're not going to build a whole Excel file because we can't? What are we going to predict what these words, how they relate to each other.

00:30:53:03 - 00:31:11:04
Bob Van Luijt
That works very well. So now all of a sudden the we could say something about the words, how we related to each other by drawing them on a map and how do we define a map? We have coordinates, right. So we see like x and y in the map. That's where I end. Or in a three dimensional space we have x, y and z.

00:31:11:06 - 00:31:42:12
Bob Van Luijt
And these models, they output at these coordinates, but not in two dimensions, not in three dimensions, but in hundreds of dimensions. And the reason they did that, that is also easy to visualize if I have a three dimensional space. So let's take a supermarket and you say, Bob, where R is this brand of toothpaste? Then we tell you what, you have to go to this x and this y coordinate, and then this z coordinate.

00:31:42:12 - 00:32:00:06
Bob Van Luijt
So you go up and then you say, oh there it is. And then you can get it. If we now draw it on a two dimensional map, I can still give you access and y. But we lost the information of Z. So the reason that these embeddings are more than three dimensions and into the hundreds is because they keep more information.

00:32:00:06 - 00:32:27:21
Bob Van Luijt
It's a simple step. And what happened was we saw more and more of these models, but the infrastructure and the technology to deal with what we call it's called these calling this, which we call the vector embeddings. That's what we do because we believed if these embeddings are going to be everywhere, which now they are, because machine learning an AI without vector embeddings does not exist.

00:32:27:23 - 00:32:52:19
Bob Van Luijt
I mean, sorry, I think that that that exists, but how we do it today, it's, it's it's omnipresent. And we built a core technology for that. So people built search systems with recommendation systems with its rack systems with its agents, with it everywhere where you want your AI application to remember something. And that's the problem we solve with our database.

00:32:52:19 - 00:32:54:07
Bob Van Luijt
And it's called weaving. That's what we do.

00:32:54:07 - 00:33:48:03
Behrad
So the dimensionality that you mentioned that the more dimension that you introduce, the more data you can store. Probably there are more ways to reach to that date. Let's say you want to find a tuna fish in the shop. The more dimension you have probably there are more ways to get to that. This dimension as well basically allows, a AI to deal with massive database of data that are useful for business, pass them in a way that they can access it as fast as possible, and so basically the dimensionality is a core component for AI application to be leveraging analytics.

00:33:48:05 - 00:34:12:15
Bob Van Luijt
Yes. But it it only has to do with the amount of information we want to keep. So every time we we remove a dimension and you can do that, you can like in a three dimensional space, I could still give you a two dimensional map. We're not we're not lost with all information. We still have some information, but not all of it.

00:34:12:20 - 00:34:41:00
Bob Van Luijt
So now you need to somehow come up with that information again. And the trick is how can we, if we look at the world's knowledge, how can we somehow compress that as well as we can in this vector space, that we can efficiently store it and search through it? That is the question, the big question at hand, and that is what the vector embeddings do.

00:34:41:02 - 00:35:14:08
Bob Van Luijt
So, they, they contain information, but not the whole information, because the whole information is always better than compressed information. The problem is we can't store the whole information because as I discussed earlier in my example, we don't have to compute, it just doesn't exist. The compute to get these insights from the raw data. So we need to somehow compress it to still get a lot of insights, but in a way that is manageable.

00:35:14:10 - 00:35:40:04
Behrad
And then the AI, what it does is like basically work with this compressed vector database and then, make correlation. Okay. There is there is Eiffel Tower here. Must be a Paris in this sentence as well with a high probability. Basically filling the gaps whenever it's necessary without access into the database, into the complete databases.

00:35:40:04 - 00:36:06:19
Bob Van Luijt
So yeah. And this is actually so if you look at a generative model, it's actually also what happens. So it's a it's a some people listening might have heard of tokens. Right. So a token is just a representation of a piece of a word. So, you know, Apple might be one token, but apple tree might potentially be two tokens, right?

00:36:06:21 - 00:36:28:21
Bob Van Luijt
Even though it's the same words and, so what the model does, we take the words like in the word apple tree, we convert it into 1 or 2. Let's, for the sake of argument, say one token. This token is again related to a vector embedding. So we now go from the token into the world of vector space.

00:36:28:23 - 00:37:03:19
Bob Van Luijt
We have all these vector embeddings in that space. And then we ask the model, what do you think that the next point in vector spaces it returns that to us. And then we look which token is the closest to that point in space. And then that to that token we can correlate with another word again. So if that's why, if we say, we can ask anything we want, we can say, the core of the apple tree is and it will do something right.

00:37:03:19 - 00:37:25:16
Bob Van Luijt
It will give us something. And that's what the generative models do. But if we want to inject information into these models, which is what we use, the vector embeddings for, we need a type of a database. Because storing data, that's what we call the database. And then what I've database is good at it and is what the vector database does.

00:37:25:22 - 00:37:57:02
Bob Van Luijt
So if you are a company and you have, I don't know, 100 million, customer files or something, then you turn all these things into vector embeddings that you can feed into that model. And that is how we what the database fact the database does and what the generative model does. And the two together generates the RAC that we call the RAC pipeline because it's one and the other, or nowadays the agents.

00:37:57:02 - 00:38:06:06
Bob Van Luijt
That's a little bit more complex than RAC. But conceptually similar under the hood. So that's why we need that.

00:38:06:08 - 00:38:39:14
Behrad
So fascinating stuff. So basically, now large organizations, they can by vectorizing their database, the massive in the database of their customers and what they are doing. Then they can allow a AI to come into the equation and give birth to a new wave of product and service offerings that were not possible before.

00:38:39:16 - 00:39:05:06
Behrad
That includes very smart and contextual recommendations, right? That includes, better, faster search perhaps. So walk me through sort of the innovation on the enterprise is could get as part of committing to this transformation.

00:39:05:08 - 00:39:31:08
Bob Van Luijt
Yeah. So I'm not I'm not a big fan, especially if we talk to to as you said, like from enterprise to think about this from a faster or something perspective. I'm always from it. More from my how do use it perspective. And at the lowest level, the easiest way to think about it is this all organizations in the world, we store a lot of data right?

00:39:31:10 - 00:39:59:06
Bob Van Luijt
Storing it, that's great. But if we can't find it, if you can't retrieve it, then it's useless, right? So, and a lot of data, it's estimated that more than 80% of data fits in what we call this unstructured bucket, because retrieval of data always has been in a structured manner. So let me take that example of what I say, what the Eiffel Tower was in Paris.

00:39:59:08 - 00:40:24:14
Bob Van Luijt
Back in the day, the only way to find a document that's at the App Store was in Paris, was looking for Eiffel Tower and or Paris. Today, thanks to the models, we can retrieve the document by searching for landmarks in France. Zero keyword matches. But still we find it. Thanks to vector space. We do this for images, just for audio.

00:40:24:16 - 00:40:50:05
Bob Van Luijt
We do that for text. We do it for image, audio and text generation and those kind of things. So that's unlocking completely new capabilities of stuff that we just couldn't build before. And let me give you a simple example. So, one of our customers, they build a tool, that helps you to search through documentation.

00:40:50:07 - 00:41:11:12
Bob Van Luijt
And so there are customers, but we also their customers. So if you go to the visit website and you want to figure out how to do something with database, you can find it. You can go there and search for it and browse through it the old fashioned way. But you can also click the okay button and then you say, like, how do I do this?

00:41:11:12 - 00:41:35:14
Bob Van Luijt
And this isn't database, but in TypeScript enter and it explains exactly how you do it. You copy paste, you're good to go. And that's you. And we're going to see a lot of new applications like you know I don't know I make something up a new travel industry right. Which the agents like you know you know, you tell your agents, you know, you've seen where I like to travel to.

00:41:35:20 - 00:41:55:15
Bob Van Luijt
I don't want to spend more than 2000 bucks. Okay. Figure something out for me. Sure. Do you want me to book for you? Yes. Agents. You can book for me to. Here we go. And the new products and things that we will see out of this. And already starting to see out of this is is is unbelievable. It's unbelievable.

00:41:55:17 - 00:42:26:14
Bob Van Luijt
So it's okay. It's it's it's a fascinating time but it's to the people. Also listening to your podcast now. We need their help putting in the creativity to come up with these new plans. So we'll help you with technology. We will help you with inspiration. So we will create example projects, demo projects and that kind of stuff. But it's to our users and our customers that really bring new value by using these technologies.

00:42:26:16 - 00:42:46:12
Bob Van Luijt
And that goes back to what we said earlier. That's why it's such an exciting time, the time to build that is now right. And not it was not one and a half years ago. Must too early. And it's not two years from now or five years from now because might be too late. So that's why it's so exciting.

00:42:46:14 - 00:43:13:22
Behrad
There is goosebump here. I see, I, we we need your products to do and our, our role is basically help clients to design new stuff. But we are so excited about this that I just told our team that we are going to accept less clients now to be able to free up time to build stuff that we would like to build, because right now it's just becoming.

00:43:13:22 - 00:43:45:16
Behrad
So it's just becoming a designer dream come true. We don't we are not entirely dependent on developers anymore. So that's that's a big topic that I would love to, dive deeper further. And, yeah, it's it's absolutely the the time of like the time of creatives to do stuff. So, Bob, I would like to get your take on two things, and I leave it to you how you want to address it.

00:43:45:16 - 00:44:14:16
Behrad
And that would be the ending. And basically we're approaching the end of this conversation as well. First of all, is that, look, you're at the core of, you're building a core technology that allows for new wave of application to come to fruition. So in your mind, in your interview, what's the difference between the pre IE application and Post application?

00:44:14:16 - 00:44:50:02
Behrad
What are the the fundamental difference you see in the architecture of this application? That's the first question I, I leave it up to you how you want to address the. And then for digital innovation officers are listening to this and I have you know, I talk to a lot of them and mean there is a new wave of, hiring across enterprise is to basically solidify and boost their digital innovation I departments to be able to walk towards.

00:44:50:04 - 00:45:24:10
Behrad
And I transform mission. What are the challenges enterprises, especially established enterprises, traditional enterprises, are facing, in your view towards moving to an AI native world? What do they have to do to basically start moving towards it? Because there's a lot of panic right now, and I and I know it's because I'm talking to these folks. So these two questions leave it to how you want to address.

00:45:24:12 - 00:45:49:05
Bob Van Luijt
Yeah. So so let me start with loss. And so, I have good news. I think I have good news and bad news. So the good news is, is that this problem is new, right? So this problem is this is a recurring problem. So there's new technology. And then what do we do. So so so then we that's but then the bad news is like, so, you know, how do we, how do we do it?

00:45:49:05 - 00:46:21:20
Bob Van Luijt
Right. This is hard to do. And, maybe it's not even bad news, but it's just like the challenge and and and, the answer is, pretty simple, because the question is like this, which what? How can we group the people together? And I'm going to, the historical, question try to answer the question. I think we can group these people.

00:46:21:22 - 00:46:55:11
Bob Van Luijt
The name that we've given to this group is, Gen Z, right? So is young people, and that is now recording these early 2025. So a couple of years from now, it's going to be an older generation. But now it's Gen Z. And we need to as organizations figure out how do we make these people not only part of our organizations, but also how do we enable them to enable their generation to build these new kind of things?

00:46:55:13 - 00:47:19:06
Bob Van Luijt
And, and that is that is the answer that somewhere there lies the answer. And so it's a and this is as old as, they're not. So in my case, I'm an early millennial. I've made sound to make money from software very early on. And of course, for me it wasn't that I that it started with, but for me it was the web line.

00:47:19:08 - 00:47:41:01
Bob Van Luijt
So, web agents almost didn't exist or they were very expensive. And I had this 14 year old dude that was me that was making, I mean, let's be honest, shitty websites. But still, I was making them and make money with that. So now we need to now transport that to AI. And and I think the same thing is happening there.

00:47:41:03 - 00:48:02:00
Bob Van Luijt
So I always tell my team this, I said like, yes, the older people in the team, they bring experience. But I tell young people, if you think that that is something they have that you don't have, then that's true. But if you think that that's not true in the reverse as well, you're wrong. Because as Gen C or you, you bring sad guys.

00:48:02:02 - 00:48:22:05
Bob Van Luijt
You know what's important now? Because that's just when you're young you have these antennas. So long story short, the answer to your question, in my opinion, is figure out how to get the talented Gen Z ers to join your company. And if you figure that one out, you're going to be fine because they know how to adopt these technologies and use them.

00:48:22:06 - 00:48:32:09
Behrad
That's a very different answer. I wasn't expecting this, but 11I need to chew on.

00:48:32:11 - 00:49:00:23
Behrad
Yeah, that makes sense. That makes sense because I think that a lot of organizations are stuck in the past. And we get we get invited to go there. And I basically give them a presentation about applications. You know, what we see working or not working. And they, you know, we charge them good amount of money. But this is something that I don't think you need to external provider to give you this.

00:49:00:23 - 00:49:15:03
Behrad
It needs to come from the within. And perhaps the reason that it doesn't come from within is that maybe the the people in the organization are stuck in the past, of course.

00:49:15:03 - 00:49:41:10
Bob Van Luijt
And it's also painful. So it's painful. Right. So it's like I'm so I'm 39. So it was about nine years ago, maybe just over that I started to notice that I wasn't in tune as much, which what the young people were doing. And and you notice that too, for example, there was like a new TV series or like it was online something or with new, these, these, these YouTube stars.

00:49:41:12 - 00:50:04:06
Bob Van Luijt
And I was like, I don't get this, I don't understand. I knew it was new, but I was like, I don't I just don't get it. It's just a it's and it's the price you pay for experience. Right? The price you pay for more and more experiences that you get older. And with that, you know, you have to get older to pay the price for experience, which is you just lose your antennas.

00:50:04:08 - 00:50:29:18
Bob Van Luijt
Maybe young people need to have their, naivete about the world to have these, antennas be open and learn about the world that way. I don't know, but it's a and don't get me wrong, I'm not saying young it's better than old because, I mean, you know, if I if I hear, like a bunch of 14 year old dudes talk, I go like, you guys are.

00:50:29:18 - 00:50:54:18
Bob Van Luijt
So it's like, oh my God. Yeah. You guys have no clue how the world works. So I'm not saying that's saying it, but that's experience, right? But the things they talk about, I can't do that. So that is the beauty of in your organization of young people and older people. Right. So one helps you from an experience perspective and the other one helps you from was that guy's perspective.

00:50:54:20 - 00:51:13:20
Bob Van Luijt
And and and and that's a, that's a circle of life. So and then 20 years from now then now Gen Z has 20 years of experience and then the new people coming in, and then that's a, so that's how I look at it. I hope I'm making sense. Why do I.

00:51:13:22 - 00:51:38:18
Behrad
As beautiful like this? This episode turned into very philosophically a philosophical love letter to AI and creativity. And and I'm all for it. Bob, really appreciate you coming on and being open and have some heart. Hot takes. And I think that's what's needed here for people, especially creative people. Any last words?

00:51:38:20 - 00:51:47:17
Bob Van Luijt
It was great to be here. I very much enjoyed the conversation, and I hope that people have some refreshed inspiration to just start building, start making stuff.

00:51:47:21 - 00:52:25:01
Behrad
It's time now. I mean, I have been loud and clear that hey guys, this is a done to your there is no excuses. It's these things are so easy to play with. Like before you had to set up this integration, this integration. Find a developer on Upwork to get something done. Right now all the the the basically what's between you and the application you're thinking about is a long series of prompting and figuring out how you can communicate what you want to do.

00:52:25:01 - 00:52:28:20
Behrad
I and let the the day I does the rest right.

00:52:28:22 - 00:52:37:01
Bob Van Luijt
Right now I'm I'm 100% with you. So it's like, it's it's it's time for people to wrap up the podcast and start building.

00:52:37:03 - 00:52:39:06
Behrad
There you go. Thanks a lot.

00:52:39:08 - 00:52:41:12
Bob Van Luijt
Thank you so much, Behrad.

00:52:41:13 - 00:52:53:05
Narrator
Thank you for listening to UX for. I join us next week for more insightful conversations about the impact of artificial intelligence in development, design, and user experience.