[00:00:12] Nicolas Babin: All these professions that are going to be disrupted—because ChatGPT is definitely a disruptive tool—these professions are going to transform themselves, will need to transform themselves.
[00:00:25] Tom Garrison: Hi, and welcome to the In Technology podcast. I’m your host Tom Garrison, with me as always as my co-host Camille Morhardt. And today we have Nicholas Babin. He is president of Babin Business Consulting, a European Commission Digital Ambassador, member of the top link expert group of the World Economic Forum, and a globally recognized key opinion leader in new technologies.
He’s a serial entrepreneur with experience that spans startups to multinational tech giants globally. He has successfully helped innovative companies achieve profitable growth and key business goals. So welcome to the podcast, Nicholas.
[00:01:07] Nicolas Babin: Thank you very much, Tom. Thank you, Camille. It’s great to be here. Really looking forward to this.
[00:01:11] Tom Garrison: Yeah. So today we want to take a topic that generally everybody’s heard about if you’re in the tech industry recently, and that’s ChatGPT and more generally, generative AI. So can we start first with how do you describe ChatGPT and generative AI?
[00:01:29] Nicolas Babin: So ChatGPT is a great tool. It’s not the only one. You have others like Notion AI, you have Bard from Google; but ChatGPT is the one that got the most traction and engagement today. I would describe it as a human-augmented tool. So let me explain a little bit about, about that.
I was part of the team that invented aibo. It was a talking dog, launched back in 1999. The first AI-based robot, and we thought we would look into ethics around AI because it was the first one, right? So we had to invent everything from scratch. We invented the way to communicate with a robot, and the way we did it at the time–because we didn’t have the technology, nor the bandwidth, nor the latency, and all these that we can find today–was to pre-record about a hundred sentences so aibo would understand a hundred sentences.
So you could ask aibo, “take a picture” in English in Japanese. Uh, we also had French software and German software. And Spanish software, obviously, and we could communicate in simple words and simple sentences with aibo. What we find today is the fourth version because it’s GPT4 that was just launched last week of a, uh, human augmented tool that has already indexed absolutely every single information that exists digitally on the web and can get the right information at the right time, but differently depending on also your own use of it. What I mean is that Tom’s use or Camille use of ChatGPT or GPT4 will actually be different than mine because I have a different way of communicating. I have different interests, so no two people asking ChatGPT at the same time, the same question will get the same answer.
I hear a lot of people saying it’s a revolution. Bill Gates said that said, the most important revolution today in terms of tech. To me it’s an evolution. We’re on a journey. That journey started even before aibo, back in 1955 when they invented the term artificial intelligence at Dartmouth. It’s part of this evolution and it’s part making sure that at least human is augmented with. What I mean by that is that humans will get the boring tasks, the mundane tasks done, but human needs to be at the center of ChatGPT. That’s basically how I see ChatGPT today.
[00:03:55] Tom Garrison: Can you describe a bit just about the technology itself? Most people don’t really understand how this works. They hear about generative AI. But it’s a bit of a black box. So can you help illuminate what is it? How does it work?
[00:04:10] Nicolas Babin: I can explain how AI works and how generative AI works, but in details about ChatGPT, I’ve not looked into it; but basically the way it works is you have an architecture of an algorithm, which is based on the neuro architecture of the brain. They started to work on it in 1993, and the idea was how to get information processed and how to get information stored because the way the human brain works is you answer based on, in your knowledge, you know, I guess we do it naturally, but basically the brain knows where the information is stored, gets the information from there. And it goes extremely fast. I mean, obviously it, there’s no latency. You know, Tom asked me a question and answer immediately because I, I know how to use my brain. And that’s what we did when we started to work on, on artificial intelligence.
So what I understand from ChatGPT is that it has index, so it’s, you’re talking about a very, very large database, what we call the big data. Any type of information has been entered into that database or structure. And what’s really interesting is the fact that today you could get any type of data–so you’re talking about emails, you’re talking about videos, audio data, Word documents, Excel spreadsheet, anything with technology the way it is can now be recorded into this architecture and based on that, the natural language processing element of it is capable of understanding the sentence, understanding the main subject of the sentence, and going into that big data in a very fast way, and that’s how ChatGPT does it. So that basically is really a 10,000-mile view of what ChatGPT does.
You are talking about years and years and years of investment from Google, from Microsoft, from all these very large companies that have poured in billions of dollars.
[00:06:08] Camille Morhardt: We’re seeing some emergence already of commercial applications like Microsoft is using the basis of ChatGPT for the search engine Bing Next Generation.
[00:06:18] Nicolas Babin: And because what ChatGPT has done, which is extremely smart, is that they’ve done some bridges between programs and so you can just put as a base ChatGPT, and then literally have it in Word, in Bing, in anything. So it’s like asking Bing, find me information about your bird story, right? and then automatically you will get that information, uh, done. So it’s, we’re really at, at a level where it’s much higher than what we’ve experienced until now with Google or Bing, or whatever, uh, searches.
[00:06:52] Tom Garrison: What uses do you see of ChatGPT? I think there’s like these sort of carnival use cases that people are kind of coming up with right now where it doesn’t really do anything. It’s just sort of a little entertaining–computer seems to be sort of talking back to you; but how do you see the ChatGPT use cases evolving in what I would say is maybe a more professional context?
[00:07:15] Nicolas Babin: Every single profession today can use ChatGPT. In education, students could use it; professors could use it. And I know I hear a lot of people say, “yeah, but that’s very scary” or “we’re gonna lose our jobs” When we first launched aibo, I had the exactly same reaction. People saying, “oof, a robot. Oof. That’s scary. Ooh, what is gonna happen? Is it going to take my job away?”
And this is exactly where we are today, where, you know, all these professions that are going to need to be disrupted–because ChatGPT is definitely a disruptive tool–these professions are going to transform themselves, will need to transform themselves in order to maybe focus on what’s most interesting in that job.
Doctors can use ChatGPT because at the end when they have to do a report on the procedures that they have performed, or on a simple, uh, appointment that they had with a patient, they can use ChatGPT. Customer support–the customer experience–gonna be improved because we’ve all experienced chat bots at the beginning where press one, if you wanna talk to somebody in English, press two if you wanna talk to somebody in Spanish. And then after that, I think we’ve all had that experience. Well, with ChatGPT, it’s not gonna happen again.
So, you know, I think the question could have been different saying, which job is not gonna be affected by ChatGPT? And at this stage, I don’t think I could answer that question.
[00:08:36] Camille Morhardt: So a few months ago, I downloaded one of the beta versions and asked it a question about kind of a science question, bird friendly power line design. Can you give me some examples and tell me more about this? And it provided a fairly detailed answer. And I asked it for references and it provided me references and I couldn’t find the references. And I came back and asked it, “I can’t find the references, help me.” And it said, “I apologize for the mistake and the previous answers; none of the references I listed exist.”
So it actually generated what looked very much like scientific references because it knows how to do that. Yep. And maybe it wouldn’t do that today, cuz this was a few months ago. But it’s a natural language model and it’s using the internet to learn. And then I think you also described at the edge, it’s then maybe doing some federated learning to customize for every one of us who’s talking to it.
And then also is generative, so it can concoct things that maybe, you know, out of all of the language and all of the internet that maybe don’t actually exist. Can you help us understand like the implications of that and how we can protect against that?
[00:09:46] Tom Garrison: Nicholas, before you jump in there, just, just real quick, Camille, that question to me, in any other setting, I would say it lied to you. It said something that in the end it knew it wasn’t true and then, then it admitted it. So it’s probably better than a four-year-old; a four-year-old to just keep lying. But at least this one, it lied to you and then it said, “oh, I’m sorry I lied to you.”
[00:10:09] Nicolas Babin: Well, to be honest, it couldn’t lie because it couldn’t invent a story. The story has to come from somewhere, right? So I have to say that the last version as well, ChatGPT3, one day a client asked me for a bio. And I had really short period of time, so I just put on ChatGPT, “write me a bio about Nicholas P. Babin, born on April 1966 in France, and the beginning was superb.
And then the end, he got me mixed up with somebody else. So suddenly I found myself with a bio that I couldn’t use and I have still no idea who that other Babin is.
[00:10:41] Camille Morhardt: I hope it didn’t write in your death, too. (all laugh)
[00:10:46] Nicolas Babin: I haven’t found that yet. Um, we’re talking here about, uh, piece of software based on AI. We’re talking about glitches. Definitely. But the thing is, as you rightfully said, maybe now with GPT4 it would be different because I used ChatGPT 1–I’ve been involved with natural language processing since the beginning because again of aibo–and believe me, the, the first version and second version were not even worth talking about. The third one started to become more and more, cuz again, you could have very good examples, but you can still have some bad experience. And this is actually, is comforting for me. It still puts human at the center, so you need to make sure that at least what you did, everybody does it because you’re right. One thing that ChatGPT doesn’t do is that doesn’t give you the source. That, to me, is a problem, frankly.
It’s surprising that he told you that he made a mistake and that the, the, the sources didn’t exist because where does, do they come from? So there was a glitch, I guess. And again, I’m pleased to see that, uh, you know, human needs to stay and remain at the center of all this because otherwise to me it’s gonna be going places where we really don’t want it to go.
[00:12:01] Camille Morhardt: Because it’s a natural language model, and you had mentioned even with aibo–French, Spanish, German, English, Japanese, and I’m not sure a few others, but um, I know this has maybe been an issue or was an issue initially with the internet where vast majority of content going on, on the internet was in English to begin with, and then sort of expanded as the basis or the underlying model of ChatGPT starts generating things like say movies.
I imagine eventually we’ll move from language like audio to video and it’ll be generating things like movies. Mm-hmm. It’s gonna have to be pulling from what exists already to learn from it. So does that mean we’re gonna have more and more content that it’s generating kind of siloed into certain kinds of languages and not being able to replicate other things or, or sort of provide a very curated, specific version of humanity because it’s only able to look at what already exists?
[00:12:58] Nicolas Babin: As of today, absolutely. Yes. But language is not an issue anymore. You’ve all heard of Depot, the, uh, AI-based translation tool. I use it all the time. I mean, I speak fluently, English, French, and German, but I make mistakes in the three languages definitely. And so you can put a whole Word document, you can put a, uh, whole PowerPoint document, and Depot will just translate it fantastically.
My wife is American so regularly what I do is before Depot I used to check by her. You know, it’s like, “can you tell me if it’s really, there’s no mistake or whatever.” And, and now I don’t need to do that anymore with Depot. It’s, it’s just crazy.
So we’re coming at a point where not only natural language, where we can see with GPT4 and things like that, but also where languages are not an issue anymore. I mean, at the European Commission, for example, use an ear bud that translates automatically. Normally you have somebody in a room that has a, has their help trying to understand the- and it’s a very, very difficult job. I’ve, I’ve done it before. It’s extremely hard. But now, uh, with technology, we’ve been able to avoid having these people; if both of you have those earbuds, you understand your language in your earbud, and the person, uh, in front of you understands their language. And that’s also based on AI.
So as of today, definitely to answer your question is definitely with languages, you could find yourself a bit stuck into what’s available on the internet, but more and more, everything is available in every single language. The world is becoming smaller already because we were better with airplanes, with, uh, trains that go extremely fast with things like that. But I think the world is gonna become even smaller because everybody’s gonna be able to talk to everybody thanks to technology, and that’s what I call tech for good.
[00:14:50] Tom Garrison: Well, to me it’s like we’re marching fast towards the future that Gene Roddenberry made with Star Trek. So the earbuds that’s the universal translator, ChatGPT is like the whole Starship–that they could ask at any question and it would always come back with the answer and—
[00:15:08] Nicolas Babin: No, not always come back with the answer. Uh, sorry I’m interrupting here, but sometimes ChatGPT says, I’m sorry, you know, I cannot get the information for you.
[00:15:15] Tom Garrison: Just like the Starship said sometimes.
[00:15:18] Nicolas Babin: Yeah, yeah, exactly. (both laugh)
[00:15:20] Tom Garrison: So, um, I wanna change it up just a little bit and ask you about some of the stuff that we’ve heard–this is a while ago–but it was about really the creative arts where people were using AI to paint a picture or to stylize things. How similar or related is that type of work to what we’ve been talking about around generative AI?
[00:15:47] Nicolas Babin: It’s generative AI for art. It’s exactly the same thing. You can say, “draw me a frog wearing, uh, an astronaut helmet” or whatever, uh, “on Mars with, uh, blue colors” and it will do it. So the combination of both ChatGPT and those tools–because you have several of them actually–make something that is really unbelieveable and why today it’s gonna be even more important is that, you know, NFTs, non-fungible tokens, they are on the market already and a lot of people are using them. So you could potentially create a piece of art that is unique because it comes out of your brain and uh, you created it using a tool and then start selling NFTs around that piece of art.
We haven’t started to talk about Metaverse, but metaverse using NFT as the main economical token, if I could say, to exchange to create an economy. You know, and this is why I believe Metaverse is gonna take off because back in 2000 when Second Life came up and Sony had invented also a home system, we couldn’t see it really taking off because there was no economy around it. We didn’t know how to make money out of that. Whereas today you have large companies, I mean, we’re talking about Nike, talking about Google, talking about Meta—Calfo in France, the big retail store–these guys are pouring in money. And also you have people going in exchanging those NFTs. And so you definitely can lose money. You can make money, but you are starting to be in a phygital world, have physical, have digital with its own economy, and this is why generative AI for art, for music too, it’s something that’s gonna be extremely powerful and it, it’s pretty impressive.
[00:17:35] Camille Morhardt: We do have an episode, actually two episodes on Non-Fungible Token or NFT, if people are curious about diving more into that and kind of what it enables.
So what kinds of technologies or policies or checks do you envision emerging so that we have any kind of a sense that. If we’re asking for something that Nicolas Babin said, it’s really you; you can attest to that or we’re sure that you generated it; it’s not just coming up with something based on other things you’ve already said or videos that sees of you online.
[00:18:07] Nicolas Babin: And that’s a very difficult question. Um, the only way to me I would see is, is by using blockchain technology, then you can make sure it’s me because you know it’ll be date. From my own digital signature and things like that, but—
[00:18:24] Camille Morhardt: I mean, I think you’re heading in one of the directions I can imagine. It just seems like there has to be a way, or there will soon have to be a way for any of us to know that what we’re looking at was truly generated by the person that any model is telling us made it, right? This is a piece of art, or here’s an audio clip or a video clip. Is it a video clip? I don’t know. It can generate them right, based on real images. So
[00:18:52] Nicolas Babin: I know that many, many large companies like Google and others are working on these issues about copyright as well. So it, it’s not exactly what you just asked me, but it’s part of the, of the problem because I created this, uh, frog on Mars. How can I make sure that at least nobody else is gonna use that and make money out of it?
[00:19:10] Tom Garrison: Well, but interestingly though, even in that example, you didn’t actually create it. Because the, the AI tool is using pieces that other people actually created, and it’s piecing elements of that together. So it’s like a composite of other people’s work, which–
[00:19:28] Nicolas Babin: But having said that, the final version, it’s mine. And, and that’s where the issue is. So who owns the copyright of that? Is there a copyright for the frog? And then a copyright for the astronaut helmet, and then a copyright for the blue colors, and then a copyright, you know, it’s like ChatGPT who wants the copyright of an article that you’ll write about birds or, or whatever.
Uh, right, because basically, yes you ask ChatGPT tool or any generative AI tools to create that piece of work. But behind it, it has taken it from so many sources around the internet who will get, uh, access to that copyright? I think we should think about how we can solve these problems. But what I’m saying is, uh, it’s part of the destruction process. Again, I started that by saying we’re in a journey and we’re learning. What I’m saying is here, let’s trust humanity and let’s be optimistic for once that we’re gonna find a way to feel comfortable in terms of all these tools, but the only way to me to feel comfortable is to ensure that human is at the center of all these. If we forget about humans, then we’re gonna find ourselves in difficult situation.
[00:20:47] Tom Garrison: Well, Nicholas, I think we’ve come full circle now. It’s been a really interesting conversation and I, for one, am walking away with a hope that maybe I could actually paint a picture for my house now. Because I know in the physical world that will never happen. But I appreciate you coming in and talking with Camille and I. It was a great topic.
[00:21:10] Nicolas Babin: Thank you very much for the invitation.