How has A.I. evolved? Deep-dive with Richard Boyd (Part 2)

DISCLAIMER: Below is an AI generated transcript. There could be a few typos but it should be at least 90% accurate. Watch video or listen to the podcast for the full experience!

Alexander Ferguson 0:01
Welcome to UpTech Report series on AI. I’m Alexander Ferguson. In this episode, we continue our conversation with Richard Boyd, founder of Tanjo. in Carrboro. Richard is an entrepreneur, author and speaker on a range of topics from virtual worlds to machine learning, education and healthcare. Here, we ask, where is AI headed? What is its future. And while he says he doesn’t like to make predictions, he does have some very interesting things to say.

Richard Boyd 0:37
I was in a car recently with someone who said, you know, I’m glad that I’m about to retire, because you know, I don’t want to use AI. And I’m like, we’re sitting at an intersection, you’re using AI right now. What do you mean? No, I’m not. I’m not using that Siri or any of that stuff. And I’m like, no, no, the traffic light, there was a time when, when we start first started having automobiles here in the country, where there was a policeman at every corner, you know, at every intersection, who’s waving people through, eventually, we started replacing them with lights. And today, hopefully, the lights have a little bit of intelligence, even if it was just the basic intelligence around timing, to time, you know, when you know, lights are on or off. But now they have sensors attached to them. So they can detect when there’s lots of traffic coming in one direction, and they can let those people through until someone else pulls up. And then it switches, right. So when you take sensors and just combine them with basic logic, Gates sorts of capabilities, that is AI, we just don’t call it that anymore, right? Because it’s, you know, biggest part of the ambient, you know, ether, now that we’ve had it for so long.

Alexander Ferguson 1:44
How are machine learning and deep learning, changing things?

Richard Boyd 1:49
You know, if you talk to most technologists, like my partner, David Smith, or Ken lane, who’s my CTO had been working with this for a while they like, yeah, you can call it whatever you want. But in the end, it’s just code, right? It’s all just code, whether it’s rigid sorts of, you know, logic that’s been programmed into a system where it’s behavior trees for characters, that give them sort of branching behavior that begins to look like intelligence. But again, it’s still pretty bounded. Like I said, machine learning, though, is in deep learning, when you start getting into those worlds, it is, to me a fundamentally different approach to solving problems that we’ve never, we’ve never had those tools before. And it’s only from my perspective, it’s 2009, is when I kind of got the Epiphany, and, and learned about machine learning. And again, machine learning has existed, since I don’t know 1958, or something like that as a concept and as a term. But the reason it’s different today is because of a lot of people contributing like some of the basic systems of basic machine learning libraries, you mentioned that natural language processing, their text parsers, their image processors that have been trained on some of the subordinate models of understanding of elements of it, when, when and when they sort of build up and combine they, they build something that looks like intelligence. And for all we know, that’s how our brains work, right? It’s just a whole bunch of small understandings that build up into a, into a larger construct that looks like intelligence. But certainly deep learning, if you’ve got the, the processing available to you, and you can do those, those are iterative, computational, deep dives into into raw sets of data, you can achieve some really interesting things, I can tell you that when we’ve experimented with a lot of that stuff, very often it gets very weird. And so you end up getting if you’ve ever played with Google stuff, where you actually have DeepMind actually do create art for you. And you can tell it to iterate on things you find interesting. It very often goes into really, really strange directions that might be disturbing to a lot of human beings. But today, with deep learning, you can have systems that create music, create art, beginning to write, although I haven’t seen anything really satisfying there yet, because I think it’s just a it’s a deeper problem. But I do believe it will be solved eventually, where you’ll have a system that can write like Shakespeare, or dos de offski, or Tolstoy or anybody else. We’ve already done that ourselves. Well, we resurrected Victor Hugo, for example.

We had our system read and we’re not even using deep learning, we’re using what I would call shallow learning, which is just building an interest graph and a sentiment model around like reading everything a person wrote, and then everything written about them, which topics have the deepest sort of emphasis, and where are the positive and negative sort of inclinations around the different topics and then And what kind of language so they like passive language, active language, that kind of thing. And those very simple little pieces put together again, start looking interesting. And that we’ve had Victor Hugo living on the internet as a what we call a Tonto animated persona or tap, I can go visit Victor Hugo every day and see what he thinks about current events, what he thinks about Donald Trump, what he thinks about climate change what he thinks about what’s happening in, you know, in, in literature and art today. And he has some strong opinions. And he changes over time, which is very interesting. So we’re watching him evolve, because like, just like humans, he’s affected by the the content he consumes. So again, I don’t know, what do you do with that? It’s very interesting. So I did the same thing, of course, with my father. So in 2017, my father died. He was, you know, 85 years old, had a very successful life. He was a lieutenant colonel in the military in the United States Air Force, he won a couple of Commendation Medals. But he wasn’t on LinkedIn, he wasn’t on Facebook, he did not have any kind of sort of data exhaust footprint, like we were talking about earlier. So instead, I had to take all of his personal correspondence, his military records, OCR, scan all that stuff in, have the system build kind of a weighted word cloud around his interest from that. And then of course, I went in by hand and edit it. So he’s really a sort of a model of my view of who he was right? Is it? How accurate is that? I don’t know. But within the first week of doing that, after he died, within a week, I’d created this model of him. And I can go visit him right now any day and go see like, what does he think about? What’s he, what’s he attracted to? What’s his mind attracted to today? And he’s reading hundreds of 1000s of articles on the net, and then scoring each one of them and telling me what he’s interested in. Now, that was cool. So we had to go to the next step, right? Which is, what if I could write to him. So now I can write to him, and he’ll tell me what he thinks about what I’ve written. Now, the final step, which we haven’t gotten to yet, is let’s go to natural language processing where I can talk to him, and he can talk back to me.

Alexander Ferguson 7:18
How could these new technologies be applied to business?

Richard Boyd 7:22
I know what people want to do within a marketing. Right? We showed that to Gartner, for example, they made us the cool vendor put us on the cool vendor list this year, because it is disruptive to marketing, instead of spending lots and lots of money on focus groups and surveys, create, like have take all of your customers have however many millions of them you have or even if you have a small pool of them, and have this create a synthetic population of who your customers are, what are their values? What do they care about? How does how do those values and interests change based on current events, or maybe just based on the seasons? Or as they move through different stages of their lives? And then ask them those questions. And guess what the data doesn’t lie, it can only represent itself. And there’s a book out there called everybody lies, which was, somebody pointed that to me right afterwards, right after I started talking about this idea. And that’s the problem, like focus groups and surveys did not predict Donald Trump. They didn’t, they didn’t predict Brexit, right. What did Google search data, Google, Google knew what was going to happen with Brexit? And they knew about Donald Trump before any of the rest of us did. 538 didn’t know. Because people lie. And sometimes people don’t even intend to lie. If you say like, you know, how many hours how many glasses of wine do you have a week, you might go like, Oh, you know, no more than two a day. It’s like, well, here’s the data from your purchases from your VIP card at Harris Teeter or food line, or wherever. And here’s how many bottles of wine you bought, right, that you buy every single week. So that data suggests that there’s a different answer to that question, even though you might think you’re answering honestly. And that’s, that’s the power of this stuff. And again, because it’s powerful, we do need to pay attention to how it’s used, who’s using it. And we’re obviously not paying enough attention right now.

Alexander Ferguson 9:20
Where do you feel AI is headed? What does the future look like?

Richard Boyd 9:24
It’s likely that we’re going to see some disruptive point. Now if you’ve heard about this idea of the singularity. Vernor Vinge II, science fiction writer talked about this first, Ray Kurzweil kind of pick that up and he has Singularity University and he does singularity conferences. And it’s that point at which AI has gotten so smart that it it exceeds human intelligence. And most people who subscribe to that theory think that’s the last invention human beings will ever make. Because then it’s all about well, will the AI keep us around or will they not need us? anymore go terminator on us. And we don’t know now most of us, again who’ve been working with it and understand how it works, and realize, and we realize that it’s just a bunch of simple things. That but a massive set of them that work really well fundamentally at the cellular level. When combined, they create these really complex, amazing capabilities. It’s hard to see that taking over or doing anything really disruptive or achieving anything like what we think of as real intelligence, although it’s possible right there. And there are people a lot smarter than me, who are concerned about it. Stephen Hawking, concerned about it. You know, Bill Gates concerned Elon Musk while he changes his mind all the time. But But yeah, so I think what we’re going to see in terms of if you want to look at a 10 year window, and I can’t look beyond 10 years, I think looking beyond five is a little perilous, right? Because things are changing so quickly now. But I would just see, say, prepare for a world where more and more things are automated. And and if I’m a young person today, I would look at the career I’m going to choose and think like, how susceptible is that activity to being automated? And again, like I said earlier, the answer to that question is changing every week, every month. So it’s hard to it’s hard to really predict which of those are the right thing. But a lot of people think, well, let’s go into computer programming, because that’s the technology field. Guess what we have, there’s been a DARPA program for a while trying to have AI systems that program and write programs. They can look at previous programs written in FORTRAN or COBOL, or whatever, understand the conceptual process and intent of that program and rewrite it in JavaScript or Ruby, or, you know, Python or any any of the new languages. That to me, that was something that took me by surprise, even. So when I see stuff like that happening, I say, programming is not going to be something immune to disruption, or, you know, automation. So, right now, I would say it’s human creativity. That’s the that’s what makes us unique. The fact that we can surprise ourselves, create our own little black swans, I think is is where we should be focused. But that’s going to be interesting, and it’s going to happen fast.

Alexander Ferguson 12:33
That concludes the audio version of this episode. To see the original and more visit our UpTech Report YouTube channel. If you know a tech company, we should interview you can nominate them at UpTech report.com. Or if you just prefer to listen, make sure you subscribe to this series on Apple podcasts, Spotify or your favorite podcasting app.