Demystification of AI: Will Singularity happen? -- Mark Esposito // Nexus FrontierTech

Welcome to the Data Driven Podcast, an I Hear Everything production. In this podcast, we explore how to transform your company and career through data-driven decision-making. Want to become a data storytelling aficionado? Then sit back, relax, and get ready to unlock the true potential of your data. Here's the host of the Data Driven Podcast, Dominick Bohan.

Welcome to the Data Driven Podcast, where we dive deep into getting more value from our business data. Whether you're a data professional, leader, or just curious about developing data skills, the Data Driven Podcast is here to guide you along your journey. I'm your host and the co-founder of StoryIQ, Dominic Bohan. Today we're going to discuss artificial intelligence and its implications for the future. Joining us today is Mark Esposito, who is the co-founder and chief learning officer at Nexus Frontier Tech. Nexus is a machine learning firm and a research lab dedicated to the creation of value through artificial intelligence. Today, Mark and I are going to discuss demystifying AI and will a singularity happen. Okay, here's my conversation with Mark Esposito, the co-founder and chief learning officer at Nexus Frontier Tech. Hi, Mark. Thanks for joining us on the Data Driven Podcast today.

Hi, Dominik. Thank you for having me.

Okay, so today, pretty interesting topic: demystifying AI and will a singularity happen? So this could be a pretty interesting, pretty confronting topic for some folks. Can you tell us a little bit about what a singularity is and how AI could enable that or even make that happen regardless of whether we want it or not?

Right. Well, thanks for kicking this off, Dominik, with this question. So, in simple terms, the singularity is the assumptions that the total computer powers of machines will largely outperform human capacity at one point in time. And so, the fact, you know, the whole narrative about machines taking over, we are gonna be subjected to machines and all of that. So, the hypothesis is that eventually, because of the speed of technology growing so fast, every single day. And we're now into the trillions of trillions of trillions of IP addresses. The internet having largely grown after 2020, and so like COVID being one of the major catalysts for the expansion of the internet. So all this conversation has brought to believe that there could be a moment in time where machines are simply going to largely outsmart us. And so that's kind of the singularity, not a new concept. It was a I hypothesized many years back, and I think back in the days when it was hypothesized, we were looking at the estimate of 2040 or so where machine powers would basically take over. Now, the number can of course change because you can't anticipate the speed of development when you make these predictions, but I think the concept is still there. Why do we have to demystify is because I think there is a lot of space for fantasy these days around singularity. And maybe we should try to draw a line and having a better conversation about what could really possibly happen. And if we are in the presence of a singularity, what would the singularity really mean, right? Not necessarily what we see portrayed in Hollywood, but maybe there will be a different kind of singularity that will not be so threatening as we tend to see.

Okay, before I leave there, so singularity is a point at which the total power of computers or the total capabilities of computers exceed that not just of a single human, but of all humans, where they have an absolute advantage in all cognitive tasks.

That's right. So it is indeed the fact that because of the automation, right, the fact that we automate in tasks across the world, right, you can say in the physical world, We've been largely mechanizing and automating tasks, and many of the things we do today are driven by machines, from simple things like elevators and escalators and cranes, right? They're mainly automating motion. You can argue that industrial production today is largely warehouses where a lot of robots are working. You could argue that even vehicles, when they're becoming smarter, right, they are a form of automation because instead of us driving, now it's self-driving cars and all that. I mean, that conversation, of course, is happening. It is happening because machines are great ways for us to increase efficiencies, reduce the cost, and of course reduce the efforts over time, right? So that story, I think, has been going on for many, many decades. I think what we are probably fearing the most is that when you automate in decision-making, some of the agency of humans, what we basically tend to decide on a daily basis, might be automated as well. And so today you see, so like we're tinkering with the conversation when it's about consumer goods. There's a lot of predictive modeling on our retailing spending. We tend to know in platforms that it's easier to lock the customers in because the benefit to stay inside of a platform are higher than to stay outside. The platform are equally becoming some form of 21st century monopolies, right? And so What if now, for example, payments are automated because a decision has been made on an algorithmic nature, right? What if, for example, I have money that is becoming digital and therefore there is a clear intent to limit what I can buy or not with that money, right? The whole idea about if money's becoming a code and I can trace it, could I, for example, preempt you from doing something that I don't want you to do? Can I simply say no longer buying that kind of product, no longer buying that kind of food, right? So somehow you can argue the spectrum of the fantasy is pretty big. But I remember when in 2020 we saw suddenly the world coming to a halt and basically everything was shut down. It was outside our imagination, right? So suddenly we would never imagine that we could reach that kind of, I would say, overpower governments, right? So the same would happen. Imagine a government is shifting more to a tech surveillance government What will happen when technology now largely automates the decisions, right? I think that's the side of singularity where you're shifting determinism to machines that is probably the freaky part, right? I think the question that we need to go and distill, is it possible whatsoever, or are we really in the presence of a flight of fancy, right?

I think that's the question on everyone's mind. Could this ever happen? Are there fundamental limits that mean there are certain things that machines will never be able to do? I think that really is the big question. Before we get into that question, it sounds like you're gonna share with us the non-Terminator Skynet movie kind of dramatic view and something that's more based on reality on the ground. And you've got deep experience in building machine learning models, so just tell us a little bit about your background at Nexus. Nexus, which is a machine learning research lab.

Learning. That's right. So, uh, maybe what we can do, we can, we can create a bridge between the conversation and the role at Nexus. So one of the reasons why I think we should be cautious with projecting ourselves into a more fear zone where we are fearing the Skynet and the Terminator is because from a technology perspective perspective, we're not there. We do not have any infrastructure capable of even, I would say, projecting or standing the singularity the way it is portrayed by the doomsayer, right? So technology is not even close to that. And fundamentally, because for technology to become a global threat, a number of different factors should be in place. And one of this would be the infrastructure, the fact that we are largely permeating the entire layers of value with automation. But, you know, the world is still going in uneven distribution pockets, right? In many parts of the world, it's still cheaper to hire people than to automate, right? So I would say the reality kicks in and you say, well, before this can possibly become reality, there are so many different steps that needs to happen. It is very unlikely that within the next few years will even going to be considering the singularity. Now connecting to Nexus. So we started Nexus in 2016. Now I don't come from a technical background. I did my studies in social science and economics. I became a professor at the university and then, you know, I basically started to work in academia. And so we started to consider a possibility of joining a venture that was operating out of Asia. And at first we said maybe we build an advisory role, advisory firm, but then we realized that the market was in dire need of firms that could go from the ideation, the proof of concept, all the way down to building the algorithms and the deployment into the client's infrastructure. So we kind of said, well, what if now we're becoming a full-fledged, you know, like I would say machine learning suite, right? And so, we started to tap on the technical resources of some of the founders. Some of us were basically, some of our founders were very deep into the tech. One of our co-founders is a kind of famous scientist in Japan who has sold his company when he was very young for a lot of money to Yahoo, right? One of the chief engineers was renowned to be one of the most sought-after coder in Japan. So we had the technical expertise they wanted to leverage, and we had the business side where we could integrate because AI was going from technical to commercial. So we started to put together a team, and little by little, I think the market was— it was like, I have this analogy, it's like in the dot-com, if you had a web company in the dot-com, it's like half of the job is done. Because you are in demand. And so we were in the AI field at the right time, but, and which is where the conversation that we're having before, when you start working with AI scientists and you try to come up and say, oh, you know, out there they fear this and that, they kind of look at you like, say, where do you come from, right? I remember when I was asking, what do you think artificial intelligence is? And this Japanese scientist said, statistics, right? And so somehow there's such a discrepancy between what you hear from the technical people who are capable of building models and codes and what you hear out there on the media, right? So that's where I started to say, well, at Nexus, I— now I am the chief learning officer, so I do much more about what we call public intellectual activities, right? So bringing the companies out there. I do teach an AI course at Harvard, and I do a lot of my use cases are coming from the experience with Nexus. Some of my co-founders, they come and they are becoming guest speakers. So I kind of bring that exposure to more of a wider range because of what I do, right? But pretty much a lot of the work we do is educational to try to tell the world, like, beware, AI is not necessarily what we think it is, right? There is a difference between the technical capabilities of AI and what we see out there in the world, right? And I've been trying to correct, I would say, the course of action. I would say almost like trying to shift a little bit the narrative from this big flight of fancy, as I mentioned before, to more of a reality check. And as much as I think AI is an incredible set of technologies, there is a clear difference from running an AI company, dealing with clients, trying to understand what they need, setting up a team, launching, making it work, right? Making sure that the predictions are aligned to the value proposition the firm wants to achieve and what you read on media, right? Which tends to sometimes be hyperbolic. So that's kind of where you see the learning educational side and select the functional side of Nexus, right? And in one end.

Thanks. Okay, that wraps up this episode of the Data Driven Podcast. Thanks to Mark Esposito, the co-founder and Chief Learning Officer at Nexus Frontier Tech, for joining us. Join us again tomorrow when Mark and I are going to discuss generative AI and beyond. If you can't wait till our next episode and you'd like to learn more about Mark, you can find a link to his LinkedIn profile in our show notes. You can also contact him on Twitter where his handle is @exp_mark, or visit his company website nexusfrontier.tech. One link in our show notes I want to tell you about. If you didn't have a chance to take notes listening to this podcast, head over to datadrivenpod.com. We've got summaries of all our episodes and contact information for our guests. Of course, you can always reach out to me. Just send an inquiry on our website, storyiq.com. If you haven't subscribed yet and you want a steady stream of data-driven brilliance in your podcast feed, We're publishing multiple episodes each week, so hit that subscribe button in your podcast app and we'll be back in your feed tomorrow. Okay, that's all for today, but remember, until next time, when it comes to data, less is more.