In this thought-provoking episode of Deep Learning with PolyAI, hosts Damien (Communications) and Nikola (CEO & Co-Founder) delve into the topic of Artificial General Intelligence (AGI) and its broad implications. Nikola shares his expert understanding of AGI, contrasting it with popular views of AI and shedding light on the effects and challenges it poses - from redefining deep learning to altering the landscape of anything from contact centers to entire societal structures, impact economies and form the next frontier of technology. 00:39 Defining Artificial General Intelligence 01:17 Understanding the Concept of AGI 02:53 The Impact of AGI on Society 06:04 The Controversy Surrounding AGI 06:13 The Potential Benefits of AGI 10:23 The Future of AGI and its Control 15:31 The Role of Open Source in AGI Development 18:46 The Impact of AGI on the Deep Learning Industry 22:27 The Role of AGI in Contact Centers Follow PolyAI on LinkedIn Watch this and other episodes of the Deep Learning pod on YouTube
Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.
Transcript
From the show. Tap a time to play from there.
NikolaIt's a resource and that's what people want it. You could have free healthcare, you could have free education, you could have better infrastructure, more optimized, And I'm firmly in that camp, right? I think that we are betraying our own evolution if we don't develop technology. For the betterment of our lives. Right.
DamienHello and welcome to deep learning with poly AI. The only podcast that has two humans talking about machines that are probably better than us. I'm here with my co host and Nikola Mrkšić, who's also the CEO and co founder of poly AI. Nikola, welcome. Good to see you.
NikolaIt's you, Damien. Thank you for hosting.
DamienOf course, I'm stepping in for Michelle. I'm typically behind the scenes, but today I'm bringing it in front of the camera. Hopefully I do her justice. But we're talking about a very exciting topic today, and that is artificial general intelligence or AGI, obviously making news in the headlines because of all the hullabaloo that's going on around the open AI environment as of this morning, I would say we have a CEO that's back seated in Sam Altman after two interim CEOs one of which was a heavy advocate and almost an acolyte of the concept of AGI.
DamienSo we thought, what better topic to discuss a little bit than this potential end state of deep learning, which is artificial general intelligence. So Nicola, without further ado, you got a doctorate in this stuff. Why don't you tell us a little bit about what the concept, it's not clearly defined, but what the concept of artificial general intelligence is. How it differs from most people's view of artificial intelligence. And, then we'll go into some deeper questions about what it means.
NikolaSure. Yeah. I think the last weekend and the week and everything that's happened I think I saw one Twitter post, which said, this is the closest thing to keeping up with the Kardashians for nerds. To be, to being one of them I spent the flight back from San Francisco to London listening to about six different talks from Ilya Sutskever, trying to like just dig at like, why did the high priest of deep learning and the main advocate of making progress in the space. Was there chanting in the background?
NikolaDid we hear the feel AGI feel the AGI in the background. I didn't feel the AGI while I was listening
Damiento it, but yeah, no burning effigies or anything either
NikolaI was in a plane, I was in a, yeah, like that might have been a bit of a security risk. But yeah, back to AGI look, the definition is I think. difficult. A lot of people tell you that artificial general intelligence is something that can do most tasks that humans can do as well or better than them. That by itself is difficult because we already have superhuman systems, some of them not even like necessarily AI that perform. Many tasks that, our economies depend on better than we do, right?
NikolaAnd we're right. And I think the whole concept and I think really where the fear factor comes in and also like the main plot of every sci fi film ever is the singularity, right? We're like a system. Can improve itself, right? It's an intelligence. It's almost like a self sufficient Unit the way that a human is and then you know, it can improve it can continue to advance and evolve You know leaving our carbon past behind, right? And I think that frightens a lot of people you'll hear elon musk talking about being a species and how this is a threat and I think that it's definitely something that gets a lot of headlines I Think that the whole discussion of safety and, what AI will do to us in the short term.
NikolaIt's a bit like, I would say that there's like a near zero probability that, sentient AI goes after humans, right? What I think it seems almost, it's
Damienegotistical to think that, that we would be a top priority. And it's not just on the hierarchy of like spatial, dominance. But I, one of the things that was always funny when you look at, when you look at Skynet or other interpretations of an intelligence that becomes like superior is that of course the first thing they're gonna do is create robots or create robot or robotized human beings. Exactly. Like we should be so bold as to think they would model themselves after us.
NikolaYou're right. And actually, physically, like, why would that be the supreme? Yeah, an actor and higher beings would probably live much more in the abstract plane. When you look at all your classical, pure mass professors they're in the clouds thinking about their own things and the facts of mere mortals do not really really care for impact them unless they impact them physically. So I think we are projecting it is. You're right. It is, I think, a very bold assumption. The biggest risk is that as they continue to evolve, we are, a detail on the side, what animals are to, to us.
NikolaWe love animals. We care for them deeply. We feed them, we feed our pets like cuts of their favorite food. Push comes to shove. If you need to build an airport somewhere or a highway, then, you might actually just destroy someone's home and you won't even ask them, right? Maybe it's it's possible that they will move and whatever, and it's not like life threatening, but the feeling, the fear, I think, comes from the fact that the level of power of decision making. Will be such that, it will be God like, and that it will be impacting our lives in ways that we're not comfortable with.
DamienJust talking about the fear of it is obvious because like you said, it's been fostered and fed by a lot of mainstream media leading up to this point. It never, it was always a, it was always a science fiction possibility. It was never an actual possibility, but now we're on the cusp of that. Possibly even there's speculation around in social media that's what caused the big rift in leadership is that there was a question around how to govern it. What, so we know that we know the sort of fears behind it. Why is it still a controversial topic? Why are we still moving, dedicating so many resources towards moving to that state, towards discovering and realizing AGI?
NikolaI think that what we have is models and like things that can do things we previously couldn't. Imagine, but that really means that we have a tool that can enable us to build all sorts of things that we currently are unable to build with existing resources. It's a resource and that's what people want it. You could have free healthcare, you could have free education, you could have better infrastructure, more optimized, like everything could be way better than it is right now. And I'm firmly in that camp, right? I think that we are betraying our own evolution if we don't develop technology.
NikolaFor the betterment of our lives. And I think that's what everyone felt. And everyone felt, a physical sense of loss when with Sam Altman, they saw opening eye, like retreating from that vision. And even if it imploded and they were all gone, it wouldn't matter. Cause I think, the genie's out of the box. People see, the fundamental insight. I do apologize for the Nikola Tesla
Damienreport. Oh yeah. We didn't get a chance to ask where in the world is Nikola today.
NikolaNicholas at the Nikola Tesla airport in Belgium catch catching a flight to somewhere else. But that was just my life. But Where were we? Yeah, just when you think about what people realized and maybe just to summarize what, like these models are, like GPT for, it's a vastly bigger version of previous neural nets that we used to, create language models and stuff that have now reached a scale. Where they have what people call emergent properties, where they are able to do things that you didn't explicitly train them to do, right?
NikolaThey're learning and creating connections, inferences, right? So I think that's like the really exciting, but no one really expected that to happen. If you listen to Elias Lutzkever, right? The guy talks about one of the main qualities in AI research being faith, which is not what you'll hear many scientists. Thank you. And then he says, it's fate that what you're doing is right. And if you believe in it, you will make it happen. And the truth is, it's a wild experiment. You're committing a lot of resources on the line. You're fundraising, you're committing more and more like GPUs to experiment.
NikolaAnd, it could just flop and give you nothing. And
DamienThe concept of faith being something that's a bit antithetical to science and the concept of like true neural networks going into action. It seems like it's not a far stretch. It's hard to quantify. And it's hard to measure the foundational like strength of that sort of claim, but really the human brain as it. Currently exists as a very complex mechanism. We're not using as much of it as we could. So maybe that is a component who's going to prove them wrong.
NikolaTotally. And I think it's just, it's an alternative model. It's not really like AGI in the, in that futuristic sense that people are afraid of, but I think we're on the cusp of having models that, you could potentially to a GPT five or six or a specialized version that you feed the right knowledge, you can say, Hey, create a pathogen that, you know. Eliminates all males on the planner, right? Or you could say be a malicious actor on the stock exchange and go crash the New York and the London Stock Exchange, right? And I think that's what people are more afraid of.
NikolaI think, when you look at the people who are more on the kind of like decelerate and like just slow down the whole thing. I think that's really what's on their mind, right? Not so much like Terminator going after Niko and Damien and ending their lives.
DamienBecause we would be the top two targets. I say, if there is an existential threat to the future of AGI, it is us. It is the two of us right now. I'm ready to go. I'm ready to go to battle, but if
Nikolayou think about what we do, we're the useful idiots, right?
DamienI've been called that my entire life. Nicola, we, we bring up a really interesting piece, which is you could theoretically like issue these directives, right? So the question is the concept of being able to maintain control. Of AGI of a recognize or of an acknowledged and realized AGI plausible. Is it even possible in the future? We talk about the importance of, especially now with current AI models in business in social applications, the importance of safety. And responsible use and deployment of AI.
DamienIs that even plausible with AGI? So
Nikolatechnology is absolutely controllable, right? It has always been like, when you look at our systems and, how we make them deterministic, even though we use both load of very powerful neural nets. In our enterprise product, like it's absolutely controllable. This whole safety thing started, probably about 10 years ago when people really focusing on biases in neural nets, the fact that like a computer vision thing for screening individuals would discriminate against one race. And that's like actually because of the data. Because we as humans are heavily biased, right?
NikolaSimilarly, the question to ask about controlling something like AGI is not really, can we control the technology? It is, can we hold ourselves back? Can we resist the temptation? Once you are, let's say you and I are like the leader of the Democrats and the Republicans, we're a bit 40 years too young for both, but Hey, thank you very much. Very nice. And If you think about that, then okay, my bot farm and your bot farm are creating fake content, fake stories about, your 50 illegitimate children and about the hundreds of billions that have stashed away in the Cayman islands.
NikolaAnd the whole thing is just exploding. Will we resist? I think like the next American election is going to be really interesting, because the way that the internet is getting electrified now about current issues, you see that technology already being put to the test. You see deep fakes all the time with statements that are not true. And you look at it and you're starting to doubt everything right? That's our defense mechanism. So it's not really, can we control AGI, but what does a world in which you have asymmetric capabilities by different actors look like, right? And that's okay. That's I think where it gets really juicy, right?
NikolaBecause at the moment you've got America as the global hegemon, right? And the challenged, but still dominant number one power, right? Sure. with a superior technology, right? So that technology can be used with devastating effects, the more symmetrical it is, right? If you think about the British empire and the opium wars, I'm not sure that the British society was far more functional than Chinese society at that point, but the military technology levels were Sure, it's all very distorted and it was very easy for Britain to impose very humiliating terms in China, right?
NikolaSo I think like what people are thinking about now what I mean When you look at opening eye and just that whole discussion is who do you give that capability to? If they themselves have something that could be used to create that pathogen or to just change the outcome of an election or something like that. How do you like get ready for a world in which that will exist? It will exist. You look at the open source models and they follow a curve that is two years, three years behind. Now, I'm sure these things are really expensive. And it might be impossible for like different groups to create things that you can create with something that's been funded by billions in the U S.
NikolaBut that's not true. There are at least a few centers of power in the world that can fund and create something and will eventually get there. So then I think the real question is just this will happen. That's a certainty. And it's not really AGI, blaming AGI is the wrong thing to do, but like knowing that the level of AI technology is reaching a point in which. There will be a scramble for, dominance and media dominance and, all sorts of economy because you now have a tool that is like a steam engine for X.
NikolaRight. And I think we're now seeing like a Cambrian explosion of all of it. And the world will look very different very soon. It already is. I think we're part of that thing. I think like our rate of growth now, everything that we're doing with our customers, the kinds of conversations we're having with clients. It was unimaginable to me two years ago. Yeah. So that's, I think the exciting bit.
DamienThe rocket style growth of a startup. You could easily equate to the space race, which is who gets there first. It usually makes the biggest splash. The same could be said for, the utilization of these tools. I agree with you 100%. It's going to be how can humans maintain control and not necessarily how does, how do we control AGI, but how do we control our own impulses and our own drive to leverage for a competitive advantage that is sometimes unethical and immoral. And when you have a system that doesn't make those determinations then all of a sudden, that's a recipe for disaster. Unlike the space, the global space race, though, once the leaders were established and they were up among the stars, it was okay.
DamienHere we are now what, but this is a different story. This is a very different story, which begs the question when you have that, when you have AGI, that's realized. And you talked a little bit about this, like the open source versus closed source model. Is it something that we, as a society, as a global society, should establish whether or not whether or not open source is the right answer? Is that something that you think that fundamentally we'll all agree on as opposed to having closed walls and letting it be that next generation space race? I think
Nikolayou'll always have everyone who is not the number one player or believes they will soon be the number one player advocating for open source. I think this is so powerful that it is inevitable that it won't be open source at the leading edge of it. I think it's also extremely important that we make open source as closely. It really depends on like how much we want to change the status quo. And if you want to see the world burn then maybe you just think that it's worth upsetting it by releasing a thing that different parties will use in different ways and different forms of success, right?
NikolaThen we're a Dan Brown
Damienbook. We're a Dan Brown villain at that stage. And,
NikolaBut it's happened many times in history, right? If you look at say the Russian revolution the Soviets Empire brought on a lot of the Soviet Union had a lot of progressive reform, but I think that the consensus is that if you probably allowed the Russian empire to continue as it was the fastest growing economy in Europe, it was making huge social reforms, regardless, they abolished serfdom and many other things. Those revolutionaries thought they were creating a better thing. I think by now, we might've done better if the other thing had happened, but I think with with AI now, like people will use it in these ways.
NikolaAnd they, I think like the economies that make the best use of it. We'll do really well. The whole debate of effective altruism, and like that as the, the continuation of ideas like the basic income and all of that, that's where you say, Hey, if Microsoft and Google are the two winners of this, if their GDP if their value exceeds the GDP of America, because. They become the American economy because they're like the source of all value because AI has made them the sole kind of like guardian of the greatest steam engine ever created.
NikolaThen they're cutting the
DamienUBI checks for the citizens of that nation.
NikolaYeah, it's do you think that's good? Is that, just a continuation of our current status quo, right? Or do we try to, encourage the whole world catching up in places that are less developed in other ways, using AI to, circumvent their own issues, I think both will happen to an extent, right? We will improve the lives of everyone on the planet with this. So it's just there's a reason everyone is so excited. And it's not just the sci
Damienfi. I think the concept of effective altruism deserves an entire episode on itself. And I know that's something that Michelle is very passionate about. So we'll come back and revisit that, expand on that a little bit. Yeah. I'm curious, just as a whole, do you. Let's look at the deep learning industry. And market is AGI the end game is that when all this, is it like cut to the NASA control center? And once AGI is realized in public and let's say open sourced and used for the greater good, does everyone just shake hands and say, okay, we've done what we can do. There's nothing more here.
DamienI
Nikolathink we didn't define AGI, so the question is Vacuo, Vacuo says is, but, I think the to come back to the hero or the anti hero over last weekend, I think it was Nips. Back then Europe's now in 2016 where alias discover just showed some model and then click the button and it's a problem is solved, right? And I we sean eddie and I have used it as an internal meme ever since Not really. It's just technology. I think the, to go to Sam Altman and his favorites by Graham, which is talent density, right?
NikolaLike the talent density of, I think the people working on AI right now is as high as it's ever been. It's attracting literally everyone to work on it to make advancements more quickly. Cause we see that the ROI. It's huge. Not business ROI, but just the impact on our future lives. All of our future lives, right? This is the technology of our generation, how fast it will continue to improve at this rate where we feel like everything will get, turned on its head. I don't know. Like it's really hard. Think of the space race, right?
NikolaI think if you watched that thing live, which we didn't, but I think you would definitely assume that in 20, 30 years, There's like a charter flight to the moon where, you pop
Damiencolonies already being established
Nikolaand you jumped around. Yeah, exactly. Aqua ducks, like I think on the excitement there, you're probably assuming that moon is like, you're outing to. Staten Island. And then like Mars is like oof. That's like that trip to Bali that you've been excited for
Damienlike half life. I'm saving up. I've I ever since I saw Total Recall, that's been my vacation fund
NikolaI like Oracle flop in the same way where no, not flop, it won't flop. It'll have a huge impact. But, and as you said, I think just the sheer nature of the product is it's not a singular, Americans got to the moon before the Soviets did, or the Soviets got to space first, like those were like, propaganda items that were really important in the Cold War, right? Right now, everyone can log into chat GPT and use a model that researchers, no matter which lab they were in, could not have imagined three, four years ago.
NikolaSo I think like that thing is there. It's usable. It's more like a steam engine than the moon landing. So that'll be the impact. But the steam engine, it completely transformed our economy. It created, it led to industrialization and this will lead to big societal changes. When I think of our space and what happens in the contact center, it's never. suffered from having too many people. It's always, it's been a chronically understaffed business unit, right? And with AI, you get a lot of extra work that can help, right?
NikolaLike mundane tasks go away, you become more productive. So then it becomes a cold war between different companies to adopt a better. AI solution so that they can win against their competitors. And the context, not just from a
Damiencompetitive advantage. It's going to impact org design as a whole. Like our solution impacts org design where we actually advise customers that, Hey, if you find that some, one we're solving for churn. Like you said, it's not like it's ever it's not like you've ever been overstaffed in a contact center, but now you're right staffing and you're allowing some of the senior or most experienced or talented agents to avoid burnout by taking on greater skill sets by becoming either a tier two agent that handles more complex problems, which reduces the. Mundane nature of those intro level calls where they're getting yelled at because someone can't remember their password or something to that effect.
DamienIt makes them a more skilled agent, which reduces their burnout, which keeps them around longer and also makes them happier. A lot of our clients say that their agents are now happier because they're feeling less more, but more important calls. Yeah. But does that happen on a global scale?
NikolaYeah, it does. And look, I think that. One thank you card we got from the head of a contact center for a client of ours was thank you. We finally have enough time to grab a biscuit in between two calls. Yeah. And that's Whoa, man, that's that's cool. And also wow. Yeah. Wow. These guys work hard. They still work hard. But I think that there, when you think about the executives in a company and, let's use the language of software companies, cause that's what we are like, there is a certain conversions between customer support, customer success, and then all the way to marketing operations and sales, it's all always been connected.
NikolaIt's. In a software company, that would be like your go to market, right? And with AI with our products, right? What are like, our stakeholders get elevated much further towards being someone running marketing or, sales, because they're now able to do sales and marketing at a much greater scale and often better in a superhuman way with our technology compared to And understaffed units trying to do it with human labor alone. So yeah, there's definitely going to be some upsets. And I think, like those countries we discussed, whichever one was first utilize this, the contact center for a long time has been one of the main targets for AI once people got off.
NikolaJust okay, it's cool. And then like, where are we going to deploy it at scale? Google contact center AI has been like. It's been called that for a while and it's one of the main kind of like tips of the spear for google And pushing ai into the whole space right For aws, like amazon connect and everything on top of it is also geared around the ai and Other like behemoths in the space And ccas and ucas they can't get on this thing fast enough, but they're not like they're not even close, right? so I think What we're doing is like bringing this technology into that department, helping it become a lot more important, more influential.
NikolaIf all of a sudden in customer service you have insights that would otherwise be the sole kind of like property of. Marketing, sending surveys and all of a sudden you go, not only am I telling you that we have a problem here and an opportunity here, but I'm telling you that because 7 percent of the people mention it and when they do, my voice assistant actually just goes into free form conversation and has that conversation that you would otherwise pay, like an agency to go and collect on a random sample of people who might not even be our customers.
NikolaSo this is I think it has huge ramifications for. Yeah. And an ability of the company to make really smart data driven decisions about the wider business.
DamienAmazing. All right. I think we're at time. So I really appreciate the insights. Fascinating topic. I'm sure that everything will change tomorrow after this episode.
NikolaI'm probably, I'm probably wrong about absolutely everything,
Damienbut it's always fascinating. Thanks for your time. I appreciate it. And we'll see you next week. Okay.
Three years ago, Simplyhealth's PolyAI deployment was only resolving simple FAQs over the phone. Today, they run voice, in-app, email, chat, and WhatsApp conversations on the platform. Nikola Mrkšić talks with Paul...
An agent that gets better every day compounds into a very different agent within months. That's the idea behind Wren, PolyAI's new release: an agent whose job is to make your other agents better. Some of what it finds...
Companies are pouring money into AI, but not everyone is getting the returns. Ian Jacobs, VP and Lead Analyst at Opus Research, thinks he knows why. In this episode, your host Nikola Mrkšić sits down with Ian to work...
When PolyAI built its Builder MCP, one client helped shape it: Vixxo Facility Solutions, one of the largest facilities providers in the US. In this episode, guest host Damian Sasso talks with Vixxo's CTO Derek Neighbors...
PolyAI just launched Dialog-RSN-1, its first audio-native model, and your host Nikola Mrkšić sat down with its builder, Matt Henderson, to unpack why it’s a game-changer for building voice agents. Most voice AI either...