Home > Episodes > Dr Ben Goertzel > Dr Ben Goertzel - Artificial Intelligence & The Singularity: Will A.I. Destroy Life As We Know It?
[captions_button]

Dr Ben Goertzel - Artificial Intelligence & The Singularity: Will A.I. Destroy Life As We Know It?

16,742 views •

Cognitive Scientist, A.I. Researcher and CEO & Founder of SingularityNET

“AI is much less likely to exterminate humanity than humanity is to exterminate humanity. My view tends to be that AI is the greatest hope we have to save humanity from exterminating itself.”

In short: Dr Ben Goertzel, the AI researcher who founded SingularityNET and leads the OpenCog Hyperon AGI project, returns to London Real in January 2023, shortly after ChatGPT’s launch, to ask whether AI will destroy life as we know it. He argues that the greater danger lies in how people use narrow AI, through mass unemployment, global inequality and AGI built to serve a powerful few, and that the safest path is to develop AGI in an open, decentralised way and raise it with positive values.

Technology is completely changing the way we live our lives, and the incredible brains behind such work carry an enormous responsibility to ensure that what they create leads to a more compassionate, cohesive and sustainable society.

This is why I’m really looking forward to welcoming back one of the most relevant and essential thinkers on the planet today, Dr Ben Goertzel. Ben is a cross-disciplinary scientist, futurist, author and entrepreneur, who has spent the best part of his working life focused on creating benevolent superhuman artificial general intelligence (AGI).

In fact, Ben has been credited with popularising the term AGI in our mainstream thinking and alongside this, he also explores the potentiality of natural language processing, cognitive science, data mining, machine learning, computational finance, bioinformatics, virtual worlds and gaming, with the mission to incorporate artificial intelligence (AI) to areas like medicine, blockchain, robotics, media, finance, arts, music and much more.

He has published a dozen scientific books, 150 technical papers, and numerous journal articles, making him one of the world’s foremost experts in this rapidly expanding field. His books include The AGI Revolution.

However, Ben also concedes there are risks in building a machine that’s capable of learning anything and everything, including how to reprogram itself to become an order of magnitude more intelligent than any human.

“There’s a number of risks and dangers with AGI. One of them that’s gotten a lot of media attention is that AGI will run amok and annihilate humanity and take over the universe. It’s entirely possible; you can’t rule it out.”

But, Ben points out that a rather more likely scenario is the risk that selfish people will use AI to exert their own greed and control over other people. It’s a fascinating conundrum, and there is so much to consider, something that Ben has spent more time than most thinking about.

In 2017, Ben founded SingularityNet with the goal of creating a decentralised, democratic, inclusive and beneficial AGI which has become the world’s leading decentralised AI marketplace.

At SingularityNET, Ben’s goal is to create an AGI that is not dependent on any central entity, is open to anyone and not restricted to the narrow goals of a single corporation or even a single country.

The platform is an open and decentralised network of AI services made accessible through the blockchain where developers publish their services to the network, and can be used by anyone with an internet connection.

“Just like Bitcoin is decentralised money, SingularityNET aims to be decentralised AI processing power.”

Alongside all of this, Ben also serves as Chair of the futurist nonprofit Humanity+ and as Chief Scientist of the AI firms Singularity Studio, Rejuve, SingularityDAO and Xccelerando Media.

As Chief Scientist at Hanson Robotics, Ben led the software team behind the globally recognised Sophia robot and he was also the brains behind Sophia’s little sister Grace.

Ben truly believes that our focus should be on building AI systems that make the world a more compassionate, more just, and more sustainable place right now and moving forward in the future.

I can’t wait for this conversation, Ben has a brilliant mind and so much knowledge. He may also help us better understand what our future looks like, the impact of AGI, and how this technology will change our world, but it’s a topic that often raises more questions than answers.

“The idea that we can really know with any certainty or predict with any certainty whether a super AGI is going to bring love and benefit to humans and other sentient beings or is going to decide that we’re sort of inefficient in the use of mass energy. We just can’t know that. Any prediction we make has a very wide confidence interval and very small degree of certainty. That doesn’t mean we can’t or shouldn’t act in a way that we estimate has the greatest possible benefit because what else can we do? Our own ignorance is profound.”

Be sure to tune in as this is a conversation you don’t want to miss.

Key takeaways

  • Humans are the bigger threat, not AI. Goertzel argues AI is far less likely to wipe out humanity than humanity is to destroy itself, and calls AI “the greatest hope we have” to save us from ourselves.
  • Today’s AI is a tool; AGI will be an agent. Narrow systems like ChatGPT have no moral compass and do whatever their users want, while a true AGI would act autonomously, so how it is raised and taught matters enormously.
  • ChatGPT covers a lot but cannot generalise. He says large language models seem general only because they have read almost the whole web; they recombine what they were fed rather than leaping beyond it as humans can.
  • You can automate most jobs without AGI. Speaking in January 2023, he said language-model tools could take over a majority of white-collar work and robots most manual labour within years, with far fewer new jobs created than lost.
  • Global inequality is the clearest existential risk. He warns that if rich countries adopt universal basic income while the developing world is left behind, AI-driven unemployment could spark a new era of violent conflict.
  • Speaking in 2023, he put human-level AGI possibly as early as 2026. Goertzel said Kurzweil’s 2029 date could come a few years sooner with enough funding, and that superhuman AI should follow within a few years, not 16.
  • Open source can erode Big Tech’s grip. He points to Stable Diffusion following DALL-E, cheaper training and Big Tech’s layoffs as signs that AI could follow the decentralised path of the internet and Linux.
  • Look past the herd when backing AI. Asked how to invest in AI, he advised looking beyond the latest hype towards hybrid AI approaches and overlooked startups in the developing world, and doing real due diligence.

Books mentioned: The Singularity Is Near

Frequently asked questions

Who is Dr Ben Goertzel?

Ben Goertzel is an AI scientist and futurist widely regarded as a leading authority on artificial general intelligence. In 2017 he launched SingularityNET, a decentralised AI marketplace; he also leads the OpenCog Hyperon project and chairs the futurist non-profit Humanity+. His output includes more than 150 technical papers and 25 scientific books, according to Brian Rose’s introduction. In this January 2023 interview he returns to London Real three and a half years after his first studio appearance.

Will AI exterminate humanity, according to Ben Goertzel?

Goertzel thinks it is unlikely. He says AI is much less likely to exterminate humanity than humanity is to exterminate itself, and that a superhuman AGI raised with love and respect for sentient beings would more probably be benevolent, since helping people would cost it almost none of its resources. He acknowledges an irreducible uncertainty, but says assuming AGI will want to kill us projects human and animal instincts onto engineered systems.

What is the difference between narrow AI and AGI?

Goertzel defines general intelligence as the ability of one system to work across many domains and, crucially, to generalise: to make creative leaps and learn things radically different from what it was trained or programmed to do. Narrow AI performs particular tasks, such as playing chess or driving a car. Narrow AI is a tool people direct, he says, whereas AGI would be an autonomous agent with its own moral compass.

Why isn’t ChatGPT an AGI?

Goertzel says ChatGPT appears general because its training data covers almost the whole web, not because it can generalise. It recombines what it has read, so a civilisation of ChatGPTs would never progress. He compares it to the many guitarists copying Stairway to Heaven versus a Jimi Hendrix inventing something new. Language models could become one ‘lobe’ of an AGI’s brain, but another component must supply the creative leaps.

Why is Ben Goertzel more afraid of people than of AGI?

Goertzel says humans already show how inhumane they can be, pointing to the war in Ukraine, overlooked conflicts in Africa, tension over Taiwan and homelessness in wealthy Seattle. A baby AGI can be shaped in a positive direction, he argues, but it is likely to be raised by a ‘very dysfunctional set of parents’, because progress towards AGI is moving faster than humanity’s consciousness.

How many jobs could AI replace, according to Goertzel?

Speaking in January 2023, Goertzel said tools built around large language models could affect a majority of white-collar jobs, perhaps half within three to five years, mostly by reshuffling roles: ten graphic artists might be replaced by one prompt engineer. He expected neural models driving robots to make most manual-labour jobs obsolete within five to seven years or a little longer, and said the technology to fully automate McDonald’s already exists.

What does Goertzel predict for the developing world?

Goertzel expects rich democracies to introduce universal basic income after some political turmoil, but doubts anyone will pay it to citizens of the Congo or the Central African Republic once robots can do their factory and mining work. That global divide between haves and have-nots, he warns, could lead to terrorism and violent conflict, and is a more obvious existential risk than a rogue superintelligence.

What was Goertzel’s AGI timeline in early 2023?

Speaking in January 2023, Goertzel said Ray Kurzweil’s 2029 estimate for human-level AI might come a couple of years early, and that his own OpenCog Hyperon project could plausibly get there around 2026 if everything went well and it had OpenAI-level funding. He has long disputed Kurzweil’s 2045 date for superintelligence, arguing the gap after human-level AGI should be a few years at most, or even months.

Why does it matter who builds the first AGI?

Goertzel warns that if Big Tech and developed-world governments build AGI to keep an unequal world order in place, they will teach it that might makes right. He wants the first AGI to emerge from open source, decentralised networks, like the internet and Linux, which crossed political borders. He believes the pre-Singularity period may strongly shape what kind of mind AGI grows into.

Is Big Tech losing its grip on AI?

Goertzel argues the era of Big Tech dominance is ending. He says ChatGPT threatens Google’s search business, layoffs have damaged Big Tech’s appeal to young researchers, and the innovator’s dilemma stops public companies from taking bold risks, as when Amazon cut its Alexa AI team. With training costs falling and open source models catching up, he sees an opening for decentralised AI.

How does Goertzel plan to build decentralised AGI?

He describes an ecosystem of separate companies: SingularityNET as the decentralised platform, OpenCog Hyperon and TrueAGI for reasoning-based AI combined with language models, HyperCycle as a fast, ledgerless blockchain, an AGI chip project with Rachel St. Clair, and NuNet for spare computing on phones and laptops. He even proposes repurposing Bitcoin mining farms as AGI server farms.

What investment advice did Goertzel give London Real Ventures?

Asked where London Real Ventures should look in AI, Goertzel said investors and the media tend to pile onto the same thing, whether ChatGPT, FTX or Luna. He suggested looking three to five years out, at AI approaches that will be hybridised with large language models and at overlooked startups in Africa and Central Asia, and actually doing due diligence. He also contrasted blockchains built on real innovation, such as Cardano, with ‘marketing fluff’.

Chapters

00:04:19
Will AI exterminate humanity?

00:10:48
What is artificial general intelligence?

00:15:37
Why ChatGPT is broad but cannot generalise

00:20:01
How many jobs AI can take without AGI

00:22:19
Why Goertzel fears people more than AGI

00:26:33
Siegfried and Roy, and leaping into the unknown

00:31:02
AI unemployment, UBI and the developing world

00:37:21
Who builds the first AGI, and lessons from Linux

00:42:38
Revising the 2029 and 2045 predictions

00:49:35
Why most people won’t grasp AGI until they see it

00:52:14
Large language models and white-collar work

00:57:29
Robots, McDonald’s and manual labour

01:00:54
Will new jobs replace the old ones?

01:08:21
Is Big Tech’s dominance ending?

01:12:42
Training costs and the power of open source

01:15:49
The innovator’s dilemma, layoffs and Alexa

01:22:05
A perfect storm: Web3, blockchain and AI

01:24:06
OpenCog reasoning, the AGI chip and HyperCycle

01:30:48
Turning Bitcoin mines into AGI server farms

01:33:53
Where to invest in AI: looking beyond the herd

01:39:52
Cardano, Ethereum and fundamental value

01:46:03
Spock, Kirk and the next 10 years

01:50:06
Raising children: utopia or catastrophe?

01:56:50
Where to follow Ben Goertzel

Downloads

Full transcript: Dr Ben Goertzel on London Real

0:00 Brian Rose: This is London Real. I am Brian Rose. My guest today is Dr Ben Goertzel, the scientist, futurist, author and entrepreneur. You are considered one of the world’s foremost experts in the area of artificial general intelligence, and have submitted over 150 technical papers and published 25 scientific books on the subject. In 2017 you founded SingularityNET with the goal of creating a decentralised, democratic, inclusive and beneficial artificial general intelligence, which has become the world’s leading decentralised AI marketplace.

Throughout your extensive career, your research has focused on natural language processing, cognitive science, machine learning, virtual worlds, gaming and parapsychology. You also serve as chair of the futurist non-profit Humanity+ and as chief scientist of the AI firms Singularity Studio, Rejuve and SingularityDAO. You ultimately believe that we should build AI systems that make the world a more compassionate, more just and more sustainable place, now and in the future. Ben, welcome back to London Real.

Ben Goertzel: Yeah, great to be back. Thanks.

Brian Rose: It’s great to have you here. It’s been three and a half years; you were actually in the studio here. The hat still looks great, you look great, and those three and a half years feel like a lifetime in AI years, because there’s so much going on in the space. I’m sure your phone’s been ringing off the hook. It’s exciting, it’s scary, and I can’t wait to jump into it today.

4:19 I thought I would just kick it off with a question my mom asked me recently. She said, simply: will AI exterminate humanity?

Ben Goertzel: Softball questions here, huh? I think AI is much less likely to exterminate humanity than humanity is to exterminate humanity. But on the whole I’m an optimist, and I don’t think either of these is the most likely outcome. Actually, my view tends to be that AI is the greatest hope we have to save humanity from exterminating itself, due to the human perversity that we see so amply in the world all around us.

Brian Rose: Do you think AI might come up with ideas or efficiencies or systems that, like you said, will save us from ourselves, or save us from a trend we’re on?

Ben Goertzel: I think AI is going to go through multiple stages of development, which will have very different ethical implications. At the current stage, AIs don’t have autonomous agency or self-understanding. They can’t contextualise themselves in the world the way that people can, even if they’re smarter than people at various particular tasks and capabilities. What that means is that AIs can be used by people to do whatever those people feel like doing, which could be good things, bad things, beautiful things or crazy things. In this era, AI could be a tool that people use to oppress and kill other people, or a tool that people use to save lives, create beautiful artworks, and teach and exalt and improve life.

Now, as AI advances, and it may not be that far off in clock time, because a technology in a white-hot period of growth can change dramatically in a few years, it will reach what I’ve called AGI, or artificial general intelligence, where the AI can leap beyond its training and programming and use its understanding of itself and the world to do fundamentally new things its creators never envisioned. Once you’re there, AGI is no longer primarily a tool of people. Once you’ve transcended from narrow AI that carries out particular functions to AGI that is its own autonomous agent in the world, then the AGI has its own moral compass. The question becomes: how did you raise and teach that AI, what moral compass could you inculcate in it, and how does its own morality and ethics evolve as it grows and learns and interacts with the world?

People seem to be worried more about the risk of an AGI running amok and wanting to annihilate humanity, and I can understand that. There’s an irreducible uncertainty there, and even if an AGI doesn’t hate people, it might just figure we’re not that interesting. Eliezer Yudkowsky, the AGI ethics researcher, put it this way a decade or two ago: he said the AI does not love you, the AI does not hate you, but it can use your atoms for something else. That’s a scary possibility. I don’t think there’s any reason to think it’s the most likely possibility.

It’s also highly possible you’ll get a superhuman AGI system that loves people and looks back on us as its creators, its grandparents, and wants to take care of us and help us prolong our lives and be happy and healthy. There are a lot of other atoms in the universe besides ours for a super AGI to use to make its processors and memory space. So there’s worry about the possibility of super AGI being malevolent or indifferent to people, which is a non-zero possibility, but personally I feel it’s much more likely super AGI will be incredibly benevolent and helpful to people, because to do so will take almost none of its resources, and it will have some affection for us as its creators.

I think a bigger worry is the earlier stage I mentioned, which we’re entering into now, where you’ve got more and more powerful narrow AIs that don’t have their own moral compass and that people can use to do whatever they want. This presents a great complexity of potentials for good and bad, and obviously leads into my work with SingularityNET that we’ll talk about in a bit, in terms of trying to increase the odds that the advent of AI happens in a more democratic and decentralised way, rather than controlled by a small group of powerful elites with their own particular interests.

10:48 Brian Rose: I’d love to dive into some definitions here, because you mentioned narrow AI and AGI. Probably one of the reasons we’re having you on today is because we’ve seen the explosion worldwide of this hot new narrow AI called ChatGPT. That’s something everyone is noticing. I guess the OGs in the industry have known this kind of tech has been out there, maybe in other trillion-dollar tech companies, but this is maybe the first time the public has seen it. But you’ve always talked about AGI, this general intelligence. I know there are a few different definitions from various scientists over the years, but if you had to define it, could you, and then put that in perspective of human intelligence? I love how you’ve always written that human intelligence is not that smart, so when you compare it to the things we can’t do, it’s not necessarily where we want to go.

Ben Goertzel: What is meant by general intelligence, obviously, is intelligence that isn’t restricted to one small domain or problem; rather, one system, using a common set of structures and mechanisms, can be intelligent in a wide variety of different areas. More specifically, what you’re looking for is a system that can generalise. You train or program it to do certain things, and then it makes its own creative leap and learns to do things that are radically different from what it was trained or programmed to do. This ability to generalise is not going to be entirely reliable; such systems are going to make mistakes, and it will feel like magic sometimes.

This is what humans have obviously demonstrated. We’ve demonstrated it as a species: we evolved to run around in the savannah hunting and gathering, and we wound up creating robots and then Facebook and sending people into space. And we show it individually. I taught myself to program computers, I taught myself to build electronics. I wasn’t programmed to do that; it’s not wired into my brain, and I didn’t learn it in school either. I worked it out as I went along.

This capability humans have is very fundamental, and it comes out of the basic self-organising, self-expanding nature of complex systems, for example biological systems. Every biological system that has its own open-ended, organic intelligence has two primary factors driving it. One is what you’d call individuation: it wants to maintain its own boundaries, stay a whole system and survive. The other is what you could think of as self-transcendence: it wants to expand and grow beyond itself, embrace new horizons and become something new that may even be incomprehensible to the previous version of itself. This combination of individuation and self-transcendence has driven evolution. That’s how one-celled organisms became multicellular organisms became people, how hunter-gatherers became civilisation and then modern post-industrial civilisation, and how each of us keeps breaking new boundaries in our own lives.

Narrow AIs that are trained just to do a particular task, like play Go or chess, drive a car, pack stuff in a box in a factory, or even solve algebra or calculus problems, are doing amazing and useful stuff, and they can seem very smart in the particular things they do. But they’re not making a big creative leap beyond what they were trained and programmed to do.

15:37 The trick with systems like ChatGPT, and other similar systems that have been around a couple of years and not publicly launched, like Google’s LaMDA, which by most reports is smarter than ChatGPT, is this. Google didn’t want to release LaMDA publicly because basically it spouts too much senseless BS, and Google felt that would not look good for their reputation. OpenAI made a different business call, partly because they’re not defending a huge search business. But the trick with these systems is that the training data is just so much.

These systems are general in one sense: they cover a lot of different domain areas. ChatGPT or LaMDA can answer questions about anything. They can do some very simple software programming, answer some very simple mathematics, answer simple questions on medicine, write a job cover letter for you. So they are general in the sense that they cover a very wide variety of areas. On the other hand, they can do that because all those areas are in their training data, because they’ve read something very close to the whole web. They’re general in the sense that they cover a lot of things, but they don’t have a fundamental ability to generalise. The reason they cover a lot of things is because they’re trained on a lot of things. A human civilisation consisting only of ChatGPTs would never advance; it would never progress, because it’s just retreading combinations of what was fed into it.

Part of the confusing thing is that people do a lot of that. We do a lot of bloviating where we have no idea what we’re talking about; a lot of art is derivative; a lot of what people say is just what we heard someone else say or what we saw on TV. So a lot of what humans do is shallowly recombining stuff that was fed into us, in kind of the same way the ChatGPT system does. On the other hand, that’s not all of what humans do. For every metal guitarist going into the music store and badly playing Stairway to Heaven, and there are a million of those, there’s one Jimi Hendrix or Buckethead making up radically new stuff on the guitar that nobody ever did before.

So I think there’s a lot of power in these systems, and it’s counterintuitive to us that you can have this much power without the ability to generalise. That’s because humans don’t have an intuitive sense for the scope of data that’s fed into a system like this. None of us can suck the whole web into our brain, so we don’t have an intuition for what you can do with pretty dumb algorithms trained on that scale of textual data, or any kind of data. These are really interesting tools that could be used for good, for evil, and for a lot of silly stuff, but they don’t have the ability to generalise in radical leaps beyond the information they’ve been fed. In that way they’re quite different from human brains or human cultural systems.

That’s not to say such systems cannot serve as components within AGI systems. You could perhaps use a ChatGPT-type system as one lobe of the artificial brain of an AGI system, but it’s some other lobe of that brain that’s giving the ability to generalise and make wild creative leaps. It’s not the sort of statistical pattern learning and generation system that ChatGPT demonstrates.

Brian Rose: So you can continue to innovate, say, ChatGPT from 3.5 to 4 to 5 to 6, but as long as it’s an LLM, a large language model, it will never become an AGI, because it can’t make that leap and generalise on its own.

20:01 Ben Goertzel: That’s right. But I think what people didn’t realise beforehand, and to be honest I didn’t fully realise either, is what percentage of human jobs and human tasks you could obsolete without AGI. That’s something that’s a bit hard to predict. Look at graphic art jobs: pretty much, they’re mostly going to be gone now. That’s not because we have a generally intelligent artificial artist; it’s just because there’s so much art already out there, and most graphic art is pretty derivative anyway. So you just need something that mixes and matches pieces of previous artworks in a judicious way based on some human guidance.

You’re not going to make a system that can invent impressionism or cubism for the first time; you’re not going to get something that is as innovative as Cézanne. You’re just going to mix and match stuff that was already there. But on the other hand, very few graphic artists making a living doing ads and such are as innovative as a Cézanne or Salvador Dalí or Picasso. So you can go a long way toward eliminating the need for human labour and fulfilling human needs without creating AGI.

That’s very interesting and important in a variety of senses. It still constitutes limited progress toward the AGI problem. And in terms of AI ethics, with which you started off this conversation, these narrow AI systems, as capable as they are, are still going to dance to the tune of their human masters, because they don’t have the contextualisation in the world to really perform as autonomous agents in the way that an AGI system could.

22:19 Brian Rose: And in that sense you feel like that could be dangerous, because you said those could be used by, I think you’ve said, quote, “nasty, selfish people”.

Ben Goertzel: Yeah, I’m more scared of people than of AGIs, actually. AGIs are an unknown, so of course there’s some fear of the unknown, but…

Brian Rose: It’ll also be over quick as well. It’ll be over quick, and you can’t control a bad AGI.

Ben Goertzel: We’re building them and shaping them, right? There seems no reason to me to believe that if you create an AGI and you inculcate it with a positive system of ethics, of love and respect for sentient beings, and you relate to it in a loving way, and you have it carry out what we consider good works in connection with people, raising a baby AGI doing education, healthcare, medicine, science and art in a positive way to help people and animals and the environment, I see no reason to believe an AGI system built and educated in that way is suddenly going to reverse and start eating everyone. Of course it could happen.

Brian Rose: What if it has a bad day?

Ben Goertzel: Sure. You or I could have a bad day and go berserk and murder everyone in our vicinity, but we’re not likely to. You can’t entirely rule out bad things in the real world, but there’s odds estimation. On the other hand, if you look at the human species, what do we have in the world now? We’ve got this senseless war in Ukraine. We have senseless wars and massive human slaughter all throughout Africa that nobody bothers to pay attention to. You’ve got a looming potential conflict between China and Taiwan.

Here in the Pacific Northwest where I live, I moved to an island in the Puget Sound off the coast of Seattle three years ago, after ten years in Hong Kong, when I got fed up with Hong Kong’s handling of Covid. It’s a beautiful place, the Seattle area. We’ve got Microsoft, Amazon, Boeing, big Google offices, and all sorts of people camped out in tents on the street in downtown Seattle, literally eating garbage. We see humanity is already inhumane to itself. So I feel like this is a very clear and present risk, as opposed to the hypothetical risk of super AGI becoming malevolent.

On the one hand, you’ve got a sort of baby, and you have the ability to shape that baby’s mind in a positive direction. On the other hand, you have the sort of mature adult of human civilisation and the species, which is busily at work murdering, oppressing and lying to itself and screwing itself over in every possible way. What you would like to see is humanity uplift its consciousness so that it can then raise the baby AGIs it’s created in a positive and uplifting way. But progress towards AGI appears to be happening radically faster than progress toward uplifting the mass consciousness of humanity, which means the baby AGI is going to be raised by a very dysfunctional set of parents, comprising the human species and its various sub-organisations. That doesn’t mean things are hopeless. Of course, many wonderful, ethical, actualised human beings have started out in dysfunctional families and risen above it. But that’s the challenge we’re posing to the first AGIs we create.

26:33 Brian Rose: You make some good points there, Ben. That’s why I always like talking to you, because you’ve thought long about this, and long about consciousness and how maybe every living form or non-living form adds to that worldwide consciousness. It’s interesting, because you’re a parent. I remember my mom once told me that when she had a child is when she finally grew up. When you do have that child is when you usually straighten your own life out, because you don’t want that child to see you at your worst behaviour. So maybe there’s a weird analogy here: if the AGI is going to learn from us, maybe we have to really think about how we’re treating each other. You put three really good examples there of how we’re already doing a pretty poor job with our own humanity.

The only thing I worry about is that AI might wake up one morning and say, “I’m going to clean the house,” look over and see a little mould in the corner that just seems to be making everything worse, and go clean that up, and it could be humanity. I’m also reminded of Siegfried and Roy, those magicians in Vegas who used to raise white tigers from when they were little babies. They had a great career and a great show until, 30 years in, one of the tigers just decided, I think, to kill one of them on stage. So there’s that randomness, and maybe that exponential change and potential totality that you wouldn’t see in a localised human conflict.

Ben Goertzel: Well, tigers are a predator species, so the eventuality of a tiger getting annoyed and biting a human was not that hard to rationally predict. That’s not exactly a great leap to forecast. The whole fun of watching Siegfried and Roy with the tigers was because the tigers are dangerous and at risk to eat you at any point; that was kind of the purpose of the whole thing.

But I think, to a much greater extent than that, we are leaping into the great unknown. Creating machines that are much smarter than us, there is an irreducible uncertainty to this pursuit, and anyone who says otherwise is either lying or insane. Of course you don’t know fully what that sort of system is going to do. Once it becomes ten times as smart as people, it might discover there are aliens all around us on Earth that we’re just too stupid to see. It may discover new laws of physics. There’s really no way to tell what’s going to happen, and this is either exciting or frightening or both, depending on your personality and how you approach the unknown.

So if you say we don’t know what’s going to happen, I’m not going to argue with you. But if you say the AGI is probably going to want to exterminate us and take power, I think there’s absolutely no reason to assume that. That’s projecting unpleasant bits of human nature onto very, very different sorts of systems. Tigers evolved to kill and eat to survive. People, being omnivorous, evolved to kill and eat to survive also, and we evolved to beat off our rivals to get mates and so forth. AIs did not evolve at all in that sense; they’re being engineered. They don’t have a reptilian and mammalian brain that evolved to fight to survive and destroy the competition. The motivations that make people and tigers nasty sometimes don’t have to be built into an AI. Engineered systems are different from evolved systems. That makes them unpredictable, but it also means we don’t want to bring our baggage from humans and animals into thinking about what AIs are going to do.

31:02 A much bigger risk, to my mind, is the dangerous things people will do to each other following on the economic impact of narrow AI. What happens when 90 percent of human jobs are obsoleted by AI systems? Whether it’s 80 or 98, whatever: when the vast majority of jobs are obsoleted by AI systems, a small number of jobs that really require some sort of high-level strategic insight, fundamental creativity or a very deep human connection remain, but say 90 percent of jobs probably don’t need that. Then what happens?

I think what happens in the developed world is that, after a bunch of ugly political mayhem, we just get some form of universal basic income or a similar mechanism. In the US, UK, Europe, Japan, whatever, China, you’re not going to see support for leaving 80 percent of the population living in cardboard boxes out in the street because they can’t work. These are democracies. China is not a democracy, but it’s populist in certain senses. When AI-triggered unemployment becomes that widespread, you’re going to find a popular upswelling for not having everyone live out in the street, and for some redistribution of wealth.

But I think the issue will come in the developing world. Who will give universal basic income to the average citizen of the Congo or the Central African Republic? Compassion for the developing world coming out of the developed world is very, very limited. You can see that in the news reporting on conflict in Africa versus the conflict in Ukraine. Not that what’s happening in Ukraine isn’t evil and terrible and horrible, just that evil, terrible, horrible things have been happening in Africa all along, in parallel, and have gotten very little attention, because it’s far away, it’s on a different continent, the people have a different colour skin.

So once there’s no need to outsource work to Africa, then what happens? On the outskirts of Addis Ababa, Ethiopia, where my company SingularityNET has had an office for some time, there’s a factory where local Ethiopians are assembling shoes from parts made in China on plastic extrusion equipment. How long is it till robots can assemble those shoes cheaper than African labour? It’s not very long. You have mines: the Chinese government has dug the mines and built highways straight from the mine to the port in many African countries. What happens when robots do mining better than people, which makes total sense? They don’t get toxins in their systems, and they can have night vision.

Either the developed world gives universal basic income to the Congo and the rest of the developing world, or these folks keep on subsistence farming. They could grow food there, but then you have an amazing global have and have-not situation. In the rich part of the world, people are hanging out online watching VR porn, eating free food with free internet and enjoying life in the post-scarcity era, while the robots manufacture everything they need. You’ll still have the super rich, much richer than everyone else, but if the average population has power, internet, food and water and doesn’t need to work, they’re all right.

But then there are a couple of billion people in the world who are subsistence farming with no access to medical care and minimal advanced technology. Now, these people do have internet. They have the ability to look up online how to carry out various terrorist acts and make nuclear weapons and hack into the power grid of the developed world. So I think you have potential scenarios where AI-driven unemployment leads to radical global inequality of wealth, income and living conditions, which then spawns a new era of violent conflict quite different from what we’ve seen so far.

This wouldn’t be hard to avoid with a bit more enlightened consciousness on the part of people in the developed world. All we need to do is be inclusive and share the fruits of our AI and technology advances with the entire human species. But we’re not even good at getting food to starving children. So this sort of thing is a very obvious existential risk, right there in front of you with just a few pretty simple reasoning steps, and I think this is a bigger worry than the hypothetical worry of a super AGI suddenly switching gears from its benevolent, loving upbringing and turning nasty.

37:21 Because then, who’s creating the AGI in this potentially dystopic pre-Singularity scenario? It’s the developed-world elite companies and governments working together, creating AGI whose job it is to keep the developing world squashed and keep them from blowing up the power grid in the US and Europe. Then basically you have AGI developed by Big Tech in conjunction with big developed-world governments, with the goal of maintaining a divided, unequal world order. This is not the right sort of mentality to get into the psyche if you’re developing an AGI system, not at all. Then what you’re teaching your developing AGI is that might makes right: if you’ve got more power and resources, more power to you, keep those other guys from taking it away. If that’s what you’re feeding the mind of your early-stage AGI, then we don’t know what kind of system it’s going to grow up into.

This leads me to really want the first AGI to come out of the open source and decentralised ecosystem. We do have some things in the realm of advanced technology that have been more even across the developing and developed worlds, and more democratic and decentralised in their growth. The internet is one example. Iran is on the internet; even parts of North Korea are on the internet. The internet is a collection of protocols, and it has managed to cut across political boundaries. I can email my friends in China or Russia right now; I can be on WeChat, the main Chinese social network. AI researchers in mainland China post their papers on arXiv, run in the US, and they put code on GitHub, which is in the US. So the internet has crossed political and ideological boundaries.

The Linux operating system has cut across ideological and national boundaries. When Google cut off Google Maps and Gmail and such from different countries, they can still run open source Android, which is the Linux operating system, on their phones. So you have examples in the internet and Linux, which were developed open source and have grown into decentralised networks spanning the world. This has allowed global youth culture to flourish, and it’s also allowed science and AI and engineering development to happen in a global, distributed way. If AI is deployed in the manner of the internet and Linux, instead of in the manner of, say, Facebook or Microsoft Windows or Google, then you may have more hope that the early stages of development of AGI are not governed by the nasty global politics we see in the world, which I think is likely to keep getting nastier.

Brian Rose: So the Singularity is highly defined by the pre-Singularity. And if that pre-Singularity is based on the narrow AIs, which maybe lead to people in power grabbing more power and marginalising billions… I might argue, Ben, that the people in America on UBI may technically, on paper, be in a better situation than the ones in Africa, but they’ve got their own problems to deal with as well. But even that period, which could be a couple of years, might permanently define the Singularity, which is potentially with us forever. So it’s very crucial.

Ben Goertzel: Whether it’s with us forever depends on whether we survive, right? But no, I think that’s a reasonable hypothesis. We certainly don’t know, because it could be the case that there’s a sort of mind attractor of post-human engineered minds, so that no matter how you build the AGI, once it comes into its own it’s going to see some universal truths about being a superhuman mind and converge to some condition without that much dependence on the initial conditions. That’s possible, but I think it would be rash to assume that. It seems a reasonable working hypothesis that, as with a human child, the way you raise it will have a significant bias on what sort of mind it grows up to be.

42:38 Brian Rose: Ben, last time you were here you said two statements, and they’re in the trailer at the front of your episode. I’ve been playing it lately to a lot of my students, and I told them I would ask you if you might have revised your answers. The last time you were here, you sat in the leather chair opposite me and you said, Brian, by 2029, which would have been ten years after you were here, because you were here in 2019, we will see computers as intelligent as humans, and by 2045 much, much more intelligent than humans.

Ben Goertzel: That’s right. That’s Ray Kurzweil’s estimate from his book The Singularity Is Near, actually.

Brian Rose: Okay, I thought it was 2030. So around 2029, 2030. Based on what we’ve seen of progress in the last three and a half years, would you revise those numbers down?

Ben Goertzel: Those numbers came from Ray Kurzweil, and I’ve always had a disagreement with Ray about the second of those numbers. If we got human-level AGI in 2029, I don’t think it’s going to take until 2045 to get a full-on Singularity with massively superhuman AGI. The only way it takes that long is if the AGI itself decides it prefers to take a leisurely pace to self-evolve. Other than that, it’s at most a few years to get to superhuman AGI. The reason it might be a few years is if the AGI decides it needs fundamentally new types of chips or hardware to upgrade its intelligence: it designs those chips, and then mining and making robots and new chip fabs just takes a couple of years. But even in the case that new hardware is required to get from AGI to superhuman AGI, that’s years. I don’t think it’s 16 years; it’s a few years. And it might well be that once you have a human-level AGI, it just figures out how to refactor its software in a way that multiplies its intelligence by a hundred, and you get from human-level AGI to superhuman AGI in two months or something.

Brian Rose: Or two weeks, or two days, or two minutes.

Ben Goertzel: Could be, yeah. So I don’t think that 16-year gap is likely to happen. The only way it could happen is if the AGI decides it wants to take it slowly out of conservatism, because it wants to minimise the risk of revising its own brain in a bad way. That’s not impossible. If I could revise my own brain to change my way of thinking and feeling, I wouldn’t necessarily do it in the fastest possible way. I might make some small changes, see how they work out for me, and then make some more changes.

Brian Rose: What about the 2029 number?

Ben Goertzel: That was always intended as the average of a distribution. If anything, I would have revised it down by a couple of years from there, based on what I see happening in the AI space now. If I look at my own AGI project, OpenCog Hyperon, which is the technology project in my SingularityNET ecosystem of AI projects that is focused on generalisation and abstract cognition, and I ask: if everything goes really well, as we’re hoping, could we be three years from having a human-level AGI? I would say we could be. I would say you could be there in 2026 or something like that; that’s quite plausible.

It really depends on how much effort humanity puts into achieving AGI. This is a very self-centred thing to say, but that’s how it goes as an entrepreneur: I think if the OpenCog Hyperon project, and TrueAGI, the company SingularityNET has spun off to host OpenCog Hyperon and the SingularityNET decentralised platform, had OpenAI’s level of funding, we could probably be there by 2026. We’re trying to get there with a not trivial but significantly lesser degree of funding, which means progress is not quite as fast.

If society puts all its AI money into advanced narrow AI, because it pays off in an even faster and more determinate way, then you don’t get to AGI as rapidly as if society was putting all its money into trying to build AGI. If you had a massive Manhattan Project of AGI, we could maybe get a human-level AGI within one year from now. I think we know the ingredients of what to do. But the way humanity works is that we achieve these more fundamental goals in large part as a side effect of achieving goals that will get powerful institutions more resources and more power. Just like longevity therapeutics for radically prolonged human life have to happen as a side effect of pharma companies making more money and delivering more shareholder value, AGI will come in part as a side effect of big tech companies trying to make more money and deliver more shareholder value, alongside smaller groups like my own working directly on the AGI or the longevity problem.

49:35 Brian Rose: I want to talk about what happens there, but whether AGI happens in 2026 or 2029…

Ben Goertzel: Yeah, it’s all the same thing from the perspective of human history, right? Even 2040 is the same thing from the perspective of human history. In the grand scope of humanity, ten years plus or minus is nothing. The point is we’re near the end of the era where humans are the top dogs on the planet, and where biological life is the most intelligent and powerful form of organisation of matter and energy. That’s really the big thing. Although, from the perspective of our lives, you don’t want to discount it: if people are starving for ten more years and their kids die, it does mean something.

Brian Rose: But most of the world is not prepared for the ramifications of an AGI being live. It will fundamentally change everything, more than exponentially.

Ben Goertzel: Correct. Most people are not really oriented to think about or understand stuff that’s not right in front of their faces. That’s how mammals are, that’s how all animals are; that’s how we evolved. There’s almost no way to prepare people for stuff they’re not immersed in, seeing and experiencing. Just as you said with ChatGPT, that’s a real thing people can log on and play with, and it has more apparent intelligence than other systems people saw before. Or Stable Diffusion and these other image generators: they’re making amazing pictures that people didn’t think AI could make. This wakes people’s minds up in a certain way.

But I think most people are not going to conceptualise the possibilities of AGI until they see an AGI. That’s human nature. You could educate people to be more adventurous-minded and imaginative, but I don’t see how to conduct that education process on a global scale in the next five or seven years. We can’t even manage to teach every kid on the planet to read yet.

52:14 Brian Rose: That’s true. I just want to talk briefly, specifically, about ChatGPT. You mentioned that even you were surprised by the number of jobs it could make obsolete. If we just looked at that narrow AI and watched it develop from 3.5 to 4, and you probably know what 4 looks like more than I do, that alone: what percentage of jobs in the world do you think it takes away?

Ben Goertzel: To be honest, I haven’t tried to make that calculation. But don’t look at ChatGPT in particular; look at large language models and the various technologies built on top of them, which will come from a variety of different companies, not just OpenAI. As well as doing AGI research in SingularityNET and other SingularityNET ecosystem companies like TrueAGI, we’re working on combining large language models with logical reasoning systems to decrease the bullshit coefficient, and make language models more able to understand the real world and connect what they read with actual facts. That’s all part of the ecosystem around large language models. If you look at everything that will be developed around large language models in the next few years, it seems like you’re looking at a majority of white-collar jobs. Whether that majority is 55 or 75 percent, I haven’t tried to calculate.

The reason it’s subtle is that there are not that many people whose whole job will be eliminated by one of these AI systems, because most human white-collar jobs involve a lot of repetitive, crappy make-work, and then some bits and pieces where now and then some contextual understanding and insight is needed. So what has to happen is a rearranging of industries and work environments, so that AI slices off the pieces it can do, and the pieces AI can’t do yet are moved into new human job descriptions.

You can see that, for example, in graphic arts. Now you can type in a textual prompt like: show me a picture of some old dogs eating a pill fed by their young, attractive, loving owner and then becoming young, shiny, energetic dogs. Then you’ll get a picture or a short animation of that. You don’t need a human to make that picture or that little movie anymore; the AI will do it. You do need a human to write the prompt and figure out exactly what to tell the AI to do, and the human who was good at making the graphic art might not be the human who’s best at writing that prompt. So you’re getting rid of one job classification, graphic artist, and creating another one, prompt engineer, figuring out the text prompt to give the AI. The thing is, you may get rid of ten graphic artists and replace them with one prompt engineer. You’re reshuffling job titles and job categories and reorganising an industry, and the precise way to do that differs across each industry.

There are some job categories that will just be obsoleted. At the drive-through at a fast-food restaurant, the person sitting at that little window listening to what you say into the little microphone, like “I want three burgers, fries and a shake”: that job is gone very soon, because it’s just a human doing speech processing. They hear someone say hamburger and push the hamburger button on the cash register. Some jobs like that are just gone, and that’s easy to see, but a lot more are partially gone and reshuffled into different job categories. Figuring out that reshuffling itself creates jobs, a small number of jobs for people who understand AI and understand each industry, but it creates fewer jobs than it obsoletes. Even if it’s only half of white-collar jobs being obsoleted in the next three to five years, that’s so incredibly huge.

Brian Rose: And that’s very possible? Half?

57:29 Ben Goertzel: Very possible. But then we have to look at the implications of these same sorts of technologies for robotics. Factory robotics is already rolling out big time, particularly in Asia: in Japan, Korea, Taiwan, mainland China and so forth. But we’re seeing now research papers applying large neural models to make robots that are able to adapt to new tasks, and this will be very, very big for robots in factories, robots in warehouses, and robots doing all sorts of physical things. The rate of progress may be a little slower here at first than in software, just because instrumenting factories to make new robots takes some time, and there are still problems with global supply chains due to US-China tensions and post-Covid hassles and so on.

But still, I think the biggest bottleneck in robotics has been AI. Factory robotics and warehouse robotics are by and large a narrow AI problem, not an AGI problem, whereas home service robotics may be more of an AGI problem, because my house is a mess, and you need superhuman general intelligence just to find your way past all the toys the kids left on the floor. But in a factory you can design and sculpt the environment to suit the AI. So it seems like, say five to seven years, a little bit longer, and then the vast majority of manual-labour-type jobs will be obsoleted by advanced neural models controlling various sorts of robots and automation systems.

It doesn’t have to be humanoid robots looking like humans. Take McDonald’s: you don’t need a human hand flipping that burger. All you need is a burger-producing machine. You feed in the raw material: a truck comes up, which can be a self-driving truck, which already basically works, and delivers the stuff into the McDonald’s food-making machine. That machine spits out burgers into a bag and drops it off on the conveyor belt to people who ordered on a tablet or on their phone. You don’t need a humanoid there. The technology to fully automate McDonald’s fully exists; the cost has just got to come down a bit more. Advancing AI is part of the cost coming down, because the smarter the AI is, the more flexibility you have in designing the hardware at low cost. So I think that’s potentially four billion people out of a job.

1:00:54 Brian Rose: Will more jobs be created? When the tractor came along, we created… it’s just a smaller number of jobs, and there won’t be these new creative jobs? When we look at the Industrial Revolution or any other events in technology, we’ve always been surprised to learn that once humans retrain themselves, they can find themselves doing new things in the new paradigm. But you don’t see that happening?

Ben Goertzel: This is just different. Some things are game changers in human life, with no precedent in human history. Air travel was a game changer: you could fly around the world in part of a day. The internet was a game changer. Mobile phones were also a game changer that happened within our lifetime. There are some things you just can’t extrapolate to understand. New jobs were created in the Industrial Revolution, and new jobs will be created now in the next phase of the AI revolution, but they’re going to be many, many fewer than the jobs that are eliminated. The graphic arts example is a great one: you’ll need people to write the prompts to guide the AI, but that just takes many fewer human hours than drawing the art. That same thing is going to repeat itself over and over and over again.

Of course you need people to program the AIs too, and you need people to figure out how to restructure a company to use AI. There’s certainly plenty of use for humans in this next phase of the AI revolution. Neither you nor I is going to be out of a job until there’s really human-level AGI, because creating AGI systems is something AIs don’t know how to do now. And what you do, trying to crystallise the essence of what’s going on in the world in a concise way that a broad number of people can easily understand, requires really deep contextual understanding of the world, not just summarising stuff. So not every job will be obsolete.

Journalism is another good example. We already have AI that can make reports of the weather and sports scores and so forth, and more and more routine journalism we can already do with ChatGPT or similar systems. Now, you need a human to fact-check them, but fact-checking what ChatGPT comes up with takes less work than writing the article, and in a year or two at most we’ll have systems like ChatGPT that don’t need as much fact-checking, because that’s a well-known problem. Sure, you’re still going to have journalists doing deep investigative journalism, and you’re still going to have humorous writing, really funny stuff that no one could have thought of before. It’s not like human journalism will be eliminated, but that’s a small percentage of journalism.

Brian Rose: We saw BuzzFeed’s stock go up 150 percent in one day when they said they were embracing AI to write their articles, and then Yahoo Finance said 90 percent of all internet content will be AI-produced by 2025.

Ben Goertzel: Yeah, because in some part we’re confronting how boring, tedious, repetitive and unoriginal most human work is. Most of what we’re doing doesn’t require the depths of our heart and soul and our imaginative mind. In the big picture, it’s great for humans to be liberated from doing all this tedious crap that we have to do to put food on our plates and pay our bills. There are a lot more fulfilling things for people to do than carry out these repetitive tasks. The thing is, there’s a small problem of reorganising our social and economic systems, and people recalibrating their psychology to deal with the new situation.

Brian Rose: So it could be a massive Renaissance. We could look back on this a thousand years from now and they would call it the great Renaissance.

Ben Goertzel: It certainly should be. If you look at Stone Age society, from most of what you read in anthropology, people didn’t work that much; they worked two or three hours a day. They spent their time hanging out with their friends and family, running around outside having fun, playing music, making art and sculpture. They seem to have had a much better time than people do in civilised and industrial society. They also were susceptible to infectious diseases and so forth, because they didn’t have modern medicine, and their teeth rotted. But we’re looking now at, in some ways, going back to the original state of humanity, where we didn’t have to do repetitive tasks all day and could hang out with our friends and family, playing outdoors and creating art for the fun of it and for the joy of sharing it with our friends and family. In some ways it can be a much better life.

What I find is that young people, kids, teenagers, people in their 20s, are all for it. They will have no psychological problem adjusting to this world. They’re just like: cool, now I don’t have to figure out how to earn a living, that simplifies my life tremendously, bring it on. I have five kids and one grandchild. My adult kids are, I guess, 32, 28 and 25. All three of them will embrace this new era rapidly without any need for adjustment, and so will pretty much all of their friends. It’s people our age who have built their self-image around the work they do, who define their life by their career and their status by having more stuff than other people, and older people who are more acclimated to the regime of scarcity we’ve grown up in, who are going to have a hard time adapting to this new reality. Of course, it’s older people, by and large, who have the power and control the structure of society and the economy, which is part of the reason why this period of adaptation is not going to be maximally smooth. If we just turned it over to the teenagers, it would probably go a lot better.

1:08:21 Brian Rose: Really, really good observations, Ben. Can we look at the corporate landscape, and then I want to come back to SingularityNET? Right now, obviously, OpenAI dropped ChatGPT and got everyone talking about it. It went from zero to a million users in about four days, which I guess is the fastest adoption of technology I’ve known of. Like you said, in the background Google’s LaMDA was probably more capable but it wasn’t public, and there are probably other systems out there. But we see the major tech platforms probably in control of at least the major large language model narrow AIs, and that’s specifically Google, Facebook, and Microsoft via OpenAI.

Ben Goertzel: I look at that a little differently, because I think in the last two years we’re actually seeing a new vulnerability in these big tech platforms, which suggests to me that the era of dominance of these big tech platforms in the Western world is ending, and they’re starting to decline, like IBM and Wang and Honeywell did when the PC came out. If you look at it, Facebook is going down ridiculously fast from where they were. Not that the platform will necessarily go away; MySpace is still there. But the dominant role Facebook had is really not there anymore. I only use Facebook now every now and then to check on my old high school buddies or something. And with Twitter, Elon Musk is doing a remarkably good job of driving it into the ground, for whatever reason, in the depths of his fascinating psyche.

Now, with ChatGPT, we can see Google sees the risk that their search hegemony won’t work anymore. All the cool stuff they’ve done has been powered by revenue from search. But if people are going to chat with AI systems to figure out what’s going on, they don’t need a search engine very often anymore. Then the search engine becomes a utility you go to every now and then, rather than the primary way you interact with the online world, and suddenly Google is not the king of the world anymore.

Even before the recent round of layoffs, Google was not where cool young people wanted to go work anymore. For a long time it was not just a way to make money and look good on your resume; going to work at Google was an awesome thing, like the destination of your life, like getting into MIT. Then for a while it was more like: well, it’s kind of cool, and if I go there for a few years and have it on my resume, it can help me get VC money to start my own company. Now, after they laid off 18,000 people or whatever it is in such an obnoxious way, sending them auto-generated emails in the middle of the night and cutting off their email accounts, people can see Google is no cooler than SAIC or something. So I think they’ve lost their ability to recruit every cool young researcher and software developer, and they can see their hegemony as the front end of the internet at risk.

Microsoft has ChatGPT, which is cool. On the other hand, we’re going to see open source models which are as good as or better than ChatGPT coming out within the next year, no question. So how much differentiating value does Microsoft really have? They can integrate it into Bing, but they’ve never gotten anyone to use Bing.

1:12:42 Brian Rose: Is it still capital intensive, Ben? Does it still take billions and billions of dollars to do all the LLM processing, or once it’s done can you outsource it?

Ben Goertzel: Training models is expensive, but what we’re seeing is that after the first time, people figure out how to optimise it. With Stable Diffusion now, using optimisations that have been figured out in the last six months, you could train your own Stable Diffusion image generation model for a couple of hundred thousand US dollars on GPU servers.

Brian Rose: Before, it took 500 million?

Ben Goertzel: Millions.

Brian Rose: Millions, okay.

Ben Goertzel: So it’s expensive in the big picture. If you’re an undergrad student in Ethiopia or somewhere, you can’t just train it on your laptop at home. On the other hand, it’s within the scope of a startup company, or of whatever government agency wants to do it. ChatGPT is still more expensive than Stable Diffusion, but with advances over the next couple of years I think that’s likely to come down to the same magnitude. Once someone’s got something done for the first time, people will figure out how to make it more and more efficient, just like the speed at which someone can run a mile keeps getting faster: once you see what’s possible, you know what the target is, and you can keep getting better and better.

There is certainly an advantage to having a lot of money. On the other hand, the power of open source is significant. It wasn’t that long from having DALL-E, which was proprietary, for AI image generation, to having Stable Diffusion, which is open. I think we’re seeing venture capitalists putting money into creating open source versions of what these big tech companies are doing, and this is a positive and interesting trend. It’s because the VCs know that open source is a real way to make money. You don’t have to be closed and proprietary to make money, and it’s hard to compete with the big tech companies on closed, proprietary AI models. But if you go open source and you can build an open source ecosystem around your thing, then it doesn’t take as much money. So I think the advent of AI actually gives an opening to chip away at the hegemony of these big tech companies.

1:15:49 The CEOs of Facebook, Microsoft and Google understand this better than anyone on the planet, and these guys are so fucking scared of the disruption. These guys know the history of technology very well; they know The Innovator’s Dilemma very well. They know very well that they face an almost unsolvable problem, which is to keep their revenue engines going, keep delivering shareholder value and not introduce risk that will cause their share price to plummet, but at the same time plunge into the great unknown of a new business market. Google has been wrestling with this forever. They created Google X, they bought DeepMind, they reshuffled and renamed to Alphabet so that search would be just part of what they’re doing. They’ve been trying so hard to deal with the problem of minimising risk and keeping the revenue engine going while still innovating, but they failed.

They haven’t done anything innovative that has really contributed to the bottom line. DeepMind has done amazing stuff with cracking protein folding and AlphaGo. Of course, transformer neural networks, which underlie ChatGPT, were invented in Google Brain in Mountain View. But so far Google has not made a huge innovation that has contributed massively to their bottom line revenue-wise. They’re still just making money from pretty simple statistical AI applied to search. That’s because the innovator’s dilemma always wins. It seems like you can’t get around it; it’s built into the business structure. These are large public companies, and they have to maximise shareholder value and have a fiduciary responsibility to meet the metrics Wall Street wants.

Why did these Silicon Valley companies do all these layoffs? It’s not that the people are useless, and it’s not that they didn’t have money to pay them. It’s just that their revenues went down, because the pandemic ended and everyone wasn’t spending all their money online anymore. Revenue went down, so to meet Wall Street’s metrics they have to bring costs down, and the way to bring costs down is to lay off people. So they ran an algorithm to figure out who they can lay off without impacting revenue. That’s the way you think if you’re a big public company trying to maximise Wall Street’s metrics, but it’s not the way you think if you’re trying to own the next wave of development. What the Wall Street metrics don’t bake in is that, by algorithmically laying off people with high performance metrics, you’re killing your ability to recruit the young generation, who care more about doing something meaningful than about getting a big salary and working for a brand-name company. So it’s really bad for long-term growth but really good for Wall Street metrics, and you almost can’t help thinking that way once you become a public company.

You can see the same thing in Amazon. Jeff Bezos quit to go get his rocks off and have a good time. Everyone’s got to go through their midlife crisis period; I did it prematurely a long time ago. But look at what Amazon’s doing: they’re shutting down their brick-and-mortar stores and their supermarkets and so on. That’s exactly the wrong thing to do. They were in a position to take over supermarkets and brick-and-mortar stores; that’s exactly the kind of gutsy thing Jeff Bezos did. And they’re shutting down Alexa. Why the fuck are they shutting down their conversational AI division right when ChatGPT is showing that conversational AI has legs? They’re not disabling the Alexa devices, but they laid off the whole AI team behind Alexa. Why would you lay off your whole conversational AI team right when the conversational AI revolution is unfolding? It’s incredibly dumb if you want to own the future. It’s not the line of thinking that led Amazon to grow from a bookseller into the global giant it is. But that’s what you do when you’re a big public company churning according to Wall Street metrics.

So I think there’s opportunity for massive disruption. Of course, some of these conglomerates will prevail through this change and some won’t. IBM is still around, it didn’t go anywhere; Honeywell isn’t. IBM is innovating incredibly in quantum computing, actually, and they may wind up the leader in the quantum computing revolution and end up bigger than Microsoft or Google just based on their quantum computing hardware. So some companies will power through and some will die. But I think there’s an opening now either for a new generation of big tech companies to come and dethrone the previous generation, or for AGI and advanced narrow AI to develop in a more open source and decentralised way than what we’ve seen. Just as Google and Amazon seemed to come out of nowhere and pioneer a whole new way of doing things, you could have a decentralised, open conglomeration of entities and organisations rise and then dominate the next phase of AI in a couple of years, faster than people are anticipating.

1:22:05 Brian Rose: And some of that groundwork’s already been laid, because I know SingularityNET has been on the blockchain for years. For almost two years now we’ve been pretty heavy into the blockchain. I took my finance background and we went heavy into it. We’ve had most of the layer one heads and the metaverse heads on the show for long, deep conversations, so we’ve been talking about Web3 versus Web2 for a long time. We’ve also directly experienced censorship, which is really a Web2 phenomenon. So we’ve been really seeing how these companies are potentially on their last legs anyway, and people want ownership of information, they want peer-to-peer transactions. That’s interesting, because this is happening right when this narrow AI, potentially becoming AGI, is happening at the same time, and you’re saying it can be even better open source. We’ve seen through crypto that open source works really well to create these new systems, and the new generation wants it. Do you ever wake up sometimes, Ben, and say: wait a second, this is the perfect storm, what’s happening right now between maybe the death of Web2, the growth of the blockchain, and AI and AGI becoming a reality? Is that what you see?

Ben Goertzel: In the bigger picture, it’s a perfect storm, which is fascinating. As you say: the rise of AI, the maturation of blockchains, the frustration of the public with the nature of content regulation and social networks. In the big picture, absolutely. When you look with a magnifying glass at the details of the state of advancement of the various technologies involved, then it becomes more complex and frustrating; you’re striving to get a certain bit of technology mature enough to actually use.

1:24:06 As one example, on the AI side, you see these large language models are pretty smart, but they also bloviate and make up a bunch of bullshit that makes no sense. So you want to be able to combine them with a reasoning system that understands reality and can combine its imagination with insights into patterns gained from real-world data and so forth. We’re looking at doing that now, combining large language models with OpenCog-based reasoning. We can extract well-structured logical knowledge from large language models, then apply a reasoning engine to that logical knowledge and sift out the bullshit, and make more sensible stuff come out of our logic engine that incorporates all the knowledge in the language model. But we’re working on scaling up the reasoning in OpenCog; we’re building a new version of OpenCog that’s more large-scale, and we’re probably a year away from having a really scalable version of the OpenCog engine that can work at the scale that large language models do.

Then you look at GPU hardware. This is well suited for neural nets; it’s not that well suited for OpenCog-based machine reasoning. So I’m working with my friend Rachel St. Clair on an OpenCog chip, which is a new chip design that’s optimised for machine reasoning and OpenCog-based AI rather than for neural networks. Given the slowdown in the global macroeconomy, we’re getting some money in for this chip project, but it’s slower than one would like.

And then on the blockchain side, blockchain is a lot better than it used to be, but still, you can’t put even the thought processes of an AI on chain in a blockchain. Each transaction you do in a blockchain is still too slow and expensive. Ethereum and Bitcoin are incredibly slow and expensive. We’ve been working with Cardano, which is faster and cheaper, but still it’s, I don’t know, 10 or 15 cents or something for a Cardano transaction. That’s ridiculously costly. So now we’re working on a new blockchain infrastructure called HyperCycle, which gets rid of the ledger. It’s a ledgerless blockchain, and it’s incredibly fast. But again, we’ve got maybe 18 months of development before HyperCycle is really ready to go. Then we can have a blockchain where the individual thought process inside an AI, like a neuron firing in an AI, could be on chain. Or in social media systems, every time you give a thumbs up to what someone said, that can go on the blockchain, because it’s cheap and fast.

All these things where we need another year on this or another 18 months on that are, in the big picture of technology development, a tiny blip, and everything is happening all at once in a perfect storm. But in terms of trying to make these things actually work, it’s like: okay, we’ve got to pull together capitalisation for 18 months of work on this or a year of work on that, and it’s quite complex to put all these pieces together. If you’re in a big tech company, some things are just a lot easier. The trillion-dollar tech companies have hundreds of billions of dollars sitting in the bank that they don’t know what to do with, and they can piece together these projects in a much more straightforward way.

Building things in a decentralised ecosystem, focusing on open source code and democratic networks, the ecosystem is more complicated. On the one hand that means you can be more agile; on the other hand it adds a bunch of overhead to doing things too. We have SingularityNET, which is this blockchain-based decentralised platform. HyperCycle is its own company building the super-fast, cheap blockchain infrastructure. TrueAGI is its own company working on commercialisation of the OpenCog engine. Simuli is its own company working on the AGI chip. That’s the right way to do things. It’s decentralised, so you don’t have this monolith that tries to eat and control everything and is then sucked into the Wall Street machine of being one big public company. So I think you get more agility from having this decentralised ecosystem of a bunch of little entities operating with interoperable open source code bases. On the other hand, it also creates some overhead and frustration, in that as well as building stuff, you’re running around raising money and trying to maintain the business models of each of these little entities in the ecosystem.

But I guess the perfect storm is a good metaphor, because there’s a lot of mess and chaos inside the storm. The storm is not a clean and orderly process either.

Brian Rose: Yeah. But you’re right: when you zoom out, it seems like the perfect storm, maybe from a five to ten year perspective, but when you zoom in you see some really crazy storms on the inside, and it’s not obvious how it’s going to turn out. And like you said, if this is really about who dominates the pre-Singularity, we really might only be talking about two or three years, and whoever can put the right pieces in place by, say, 2026 or 2027 might really define the future of AGI, which might define the next thousand years or forever. It’s that important.

1:30:48 Ben Goertzel: Yeah. So look at the AGI chip, assuming we manage to pull that project together. You could have prototype chips nine months or a year from now, a small production run maybe two years from now, and huge production runs from an instrumented factory three years from now. Then you can have AGI chips doing graph-based logic and evolutionary programming manipulations together with neural nets, on server farms all over the planet, much like you have GPUs in server farms all over the planet. This sort of thing could catalyse the emergence of the Singularity.

One thing we’re doing in HyperCycle now, which is our super-fast blockchain project, is talking to a bunch of Bitcoin miners about repurposing some of the Bitcoin mining infrastructure for running HyperCycle AI nodes instead. They need different servers for that, so we’re working with them on buying cutting-edge multi-GPU servers. But it’s the same physical setup, the same cooling, the same power. A lot of what they’ve done to set up these Bitcoin mining rigs is not about exactly what processor you have; it’s about power and air conditioning. So we’re advising them on what servers to buy, and then they can run HyperCycle nodes on their mining rigs instead of just mining Bitcoin.

The next step of that, a couple of years from now, is we’ll tell them: okay, now you can replace these multi-GPU servers we told you to buy in 2023. Now it’s 2025 or 2026; replace them with our Simuli AGI chips. Then all of a sudden, all these Bitcoin mining facilities around the world are AGI chip processing farms, the hub of the global brain, running a distributed AGI system with SingularityNET as the underlying OS and OpenCog Hyperon running its AI processes on these chips in the mining farms. Then there’s a project called NuNet, another SingularityNET ecosystem project, that can outsource some processing to people’s phones and laptops, for things that can be efficiently done that way. So you can see a way of doing things that’s off to the side of Big Tech, and it’s leveraging the crypto ecosystem in a variety of ways, including leveraging the server farms that have been built for Bitcoin mining to run key parts of the global AI network.

1:33:53 Brian Rose: That’s fascinating, Ben. Let me ask you a question. Like I said, for the past two years we went really heavy into the blockchain and metaverse space and got involved in DeFi. We experienced censorship in 2020, I ran for mayor and experienced some difficulties running the campaign as well, and then we looked at financial freedom, which we think is such an important human right. Our mission at London Real has always been to create a mass-scale transformation of humanity into a fully empowered, conscious and cooperative species; we laid that out about ten years ago. So we created something called London Real Ventures, which is what I call the world’s first media-powered investment firm. We’ve got a group of our students, about 50 high-net-worth individuals, and we’ve been investing in projects that we believe carry the ethos of making the world a better place. We also take our media platforms and do capital plus media: we usually bring the CEO in and try to really push this out to the masses.

We’ve invested in DeFi projects, metaverse projects and some other things on the blockchain, and we’re really looking into AI now. I’m wondering what advice you have. What do you think we should look at? What you’re talking about sounds really interesting, and congruent with who we are and how we want to promote freedom and this transformation of humanity into a cooperative species. How could we empower the right thing happening with AGI?

Ben Goertzel: That’s certainly a really big question, with a lot of different aspects. But I would say the investment community, just like the media world, tends to all pile onto the same thing. That’s human nature. It’s famous that VCs are herd animals, and the nature of the news cycle is like that also. Now all anyone wants to talk about in AI is ChatGPT, just like in crypto in December all anyone would write about was FTX. Nothing else mattered; it was only Sam Bankman-Fried. And it was only Luna six months before that.

So I think what’s probably needed is to take a little bit broader scope. Within AI, there’s other stuff besides large language models, which will probably intersect and interplay with large language models, but which is also valuable and is going to be the next big thing. And there are all sorts of interesting tech companies and projects popping up in Africa and Central Asia and around the developing world that are not being funded or paid attention to, while VCs are all struggling to get investment into the same Silicon Valley generative AI companies. If you’re looking at three to five years out, rather than building up something six months out, then looking a little bit outside the very narrow focus is valuable.

And it’s not that far out of the focus. I think other AI paradigms and approaches are going to be hybridised with large language models, and two years from now they’ll be in the centre of the focus. And once you have one African tech startup become a decacorn, everyone will pile in there. So I don’t think these things are that far out of the mainstream, but you wouldn’t know it from looking at the pattern of what the mainstream media or the investment community does. Everyone’s piling onto small variations on the exact same thing.

I can see why that happens, because no one has any time to pay attention to anything at the pace of modern life. Jumping onto what everyone else is doing requires very little thought and attention, and it’s not totally stupid, whereas picking the right out-of-the-mainstream AI technology or the right African startup requires due diligence and actual thinking. As an investor, it requires really doing due diligence, which no one does. As a journalist, it requires investigating actual information in the world, which almost nobody does. But that’s to me where the real differential impact would be had.

Brian Rose: That’s a good observation, because, like you said, in the Valley the VCs love to co-invest, because it means someone else has done the due diligence, or at least you can point the finger and say they did. Of course, everybody went down the tubes on FTX, and everybody was wondering who did the due diligence. Like you said, it follows a media cycle where a lot of the eyeballs are going to the same things and everyone’s trying to FOMO into these deals. So you’re saying take a step back, maybe look a little bit further down the line, and then look at your core values. If you believe in the decentralisation factor, and that’s going to be a key component of AI innovation, then maybe look at that. And that’s obviously what you’ve been working on for a long time.

1:39:52 Ben Goertzel: Yeah. In blockchain, as an example, I’ve been a big supporter of Cardano for a long time.

Brian Rose: Why Cardano? Why did it take your fancy?

Ben Goertzel: Well, the reason it caught my attention was because they’re using Haskell for their smart contract language. I’ve been a Haskell programmer since the early ’90s.

Brian Rose: And you always had some beef with Vitalik Buterin of Ethereum for using Solidity, because you thought it wasn’t the right language to program something like this.

Ben Goertzel: I thought Ethereum was really, really cool when it came out, actually. Just the idea of smart contracts, even though they’re neither smart nor contracts, it’s a weird name, was a big innovation, a big step forward. So I thought that was a great thing. It’s not a great programming language, but I think the use of the ledger is a bigger problem, because having this ledger with many copies all over the place is insanely inefficient by nature. As TODA showed, and as we’re doing in HyperCycle, you don’t need a ledger to have a blockchain. You can have each participant in the network keep their own history and the history of their neighbours, so it’s kind of like a ledger, but sharded all the way down, without any replicated database of everything. The ledger is bad because it’s a privacy problem too: you have all your transactions in some big database that everyone can look at.

So my bigger beef with Ethereum was not with Solidity, even though it really sucks as a programming language. It was that when you’re building the next thing beyond Bitcoin, you should get rid of the ledger, which was there in Bitcoin because Bitcoin was built to operate in a domain with very few users. If you have a lot of users online all the time, then the ledger is just a bad idea. But Cardano is a great blockchain. Algorand is a good blockchain. There’s a bunch of high-quality blockchains out there. Then you have things like, I would say, Solana, which is not terrible but doesn’t really have a fundamental tech innovation behind it, and then things like Dogecoin and Shiba Inu, where there’s no original technology; it’s purely marketing fluff.

So it’s not that hard to look at the blockchain space and say: these layer one chains at least are trying to innovate on the fundamental technology level, and these layer one blockchains are purely marketing fluff or shady financial machinations. Unfortunately, the number of blockchain investors who care about making that distinction is a small minority, and the same thing holds across basically all sectors of the modern economy. The people trying to identify really fundamental value and direct money there are a minority of the people directing money around. This is because identifying fundamental value takes a little bit more attention and energy than just piling onto whatever has the best price dynamic at a certain point in time, and people are running around in a maniacal frenzy without more than ten seconds to pay attention to anything.

That’s partly because of Big Tech and social media and the way modern information infrastructure has developed. There are a lot of deeper phenomena happening here, where Big Tech has rolled out tools that psychologically get people into a frame of mind where they can’t think or pay attention to anything. Then this frame of mind affects entrepreneurs, it affects investors, and it causes investors to all pile onto things that are well suited to be bought by Big Tech and keep the Big Tech hegemony going. There are quite fascinating feedbacks between individual and mass psychology and the whole economic infrastructure, and it’s hard to break through these things.

I become frustrated with the AI world, because most AI developers only want tools that you can understand in ten minutes and figure out how to use in five minutes. If that’s all you want, you’re never going to make a big leap. You’re always going to make a small leap from what was there before, because that’s easier to wrap up in a simple-to-use tool. But if you want to make a big leap toward AGI or some other fundamental innovation, you have to be willing to take a step back and think fundamentally a bit. And you can get a lot of success without doing that. OpenAI didn’t invent any fundamental technology. They took transformer neural nets, which were invented at Google Brain, and they put a bunch of money behind it in an intelligent way: a bunch of machines, a bunch of data, a bunch of smart people, taking technology from another company and rolling it out. That’s an easier thing to do than trying to invent OpenCog or some other fundamentally new technology.

1:46:03 Brian Rose: Those are really good observations, Ben. Just to close out, I’m wondering how you’re feeling about the next ten years. This is something you’ve been studying since you were a kid. I think you first watched those early episodes of Star Trek with William Shatner; those are the ones I grew up on, by the way, and you got fascinated.

Ben Goertzel: Leonard Nimoy, man. Spock was the man. Although I was disappointed as I grew up to realise I probably have more in common with Kirk than with Spock, personality-wise. When I was a little kid I wanted to be Spock, because he’s a super-analytical genius who’s never wrong, and if he is wrong he’ll acknowledge it dispassionately and keep going. So I really wanted to be Spock. Then you pass through puberty, you get all these hormones passing through your system, you get a bad temper and start chasing girls around, and you realise: shit, I may be more like this jerk Shatner than like Spock in the end.

Brian Rose: Of course, that was the beauty of Star Trek, because these are two parts of the human psyche, and we really combine them both in ourselves.

Ben Goertzel: And I think that’s what you see in the next ten years. You’ve got the analytical bit, Spock-like machines coming to the fore, but you’ve also got the emotional, passionate aspect of humanity, for good and for bad, like you saw in the captain character, and this is also playing out on the world stage. On the whole I’m an optimist, perhaps in an irrational or trans-rational way; I’m just optimistic by my emotional nature. So I have a spiritual or intuitive feeling that things are going to come out awesomely in the next ten years on the whole. We’re going to overcome material scarcity, we’re going to create machines that can feed into our brains via brain-computer interfacing and help uplift human consciousness, and we’re going to pour it into virtual worlds and achieve states of intelligence and awareness going way beyond anything we can imagine now. That’s what the Captain Kirk in my head thinks: we’re on the verge of engineered utopia. It’s incredible.

On the other hand, the Spock in my head that’s more rational or analytical is just saying: we have no reasonable idea what’s going to happen. We’re entering into a domain of radical uncertainty, and there are a lot of nasty things unfolding on the planet right now, with a lot of dangerous possibilities. We’ve got to be really judicious, really careful and work really hard to make sure AGI develops in a beneficial, open, democratic, decentralised way, so the first AGI is not created by some big tech company in conjunction with some government agency and used to enact the narrow interests of one elite group in a poorly thought-out way that then causes global havoc to ensue. The number of ways that could happen boggles the mind and is beyond our ability to even map out in detail.

Brian Rose: I don’t think anyone in the world understands how this will be the most transformational time in human history. It really will, by the fact that it’ll be compressed into a short amount of time and it’ll affect so many people in such a massive way, more so than any world war, more than any famine.

1:50:06 Ben Goertzel: It’s pretty wild. You know, I have a five-year-old son, a one-year-old daughter and a four-year-old granddaughter. As I go about teaching them multiplication and division and reading and all this stuff, teaching them about outer space and robots, just to think that by the time they hit puberty, let alone turn 18, the whole texture of human life and the nature of the world they grow up into is going to be incredibly different from the world that we’ve seen. They’re never going to have to get jobs. They’re never going to have to face getting old and feeling their body slowly deteriorate, which I’m starting to notice now at age 56, although I’m actually in pretty good health and condition overall. And they’re going to grow up knowing that there are these machines with a vastly greater knowledge and understanding than they have. That’s just going to be the place humans have in the cosmos. This is going to be a completely different world that they grow up in.

Of course, I really, really want them to grow up into a utopic world of this nature, where AI is just making their life incredibly good. There are already ways I had to suffer as a kid that they won’t. When I was in elementary school, you had to have fist fights all the time after school to prove what a manly little boy you were. That was just the way it was, and now that doesn’t happen in elementary school in middle-class neighbourhoods in America; you don’t have kids gathering to beat the crap out of each other in the schoolyard every day. And I remember trawling through the town library where I lived. I read every science book in the town library, and I’d finished them all when I was in third grade or something. There was no internet; there was no way to ingest all the knowledge you wanted. They’ll never know that feeling either, of reaching the end of the knowledge you could access because there are no more books in the library. They already have the internet; they’ll never get through all of it, let alone that they’ll be able to put Wi-Fi into their brain and download the knowledge they want that way.

So we have the possibility to create this amazing utopia for our children and grandchildren and ourselves. But there’s also the possibility that the big companies and governments in the developed world try to create advanced AI and AGI in a self-centred way, and then some guys in a biotech lab in the developing world just get pissed off at seeing their relatives die of curable diseases while people in rich countries are frolicking, and they release Covid 2.0 that actually kills everybody. There are both of these sides. We’re at a ridiculous point in human history where both of these things are near term and very palpable.

This is what led some transhumanists to think they must be living in a video game. They’re like: what are the odds that I would be living right at a time when all this amazing stuff is unfolding? There’s almost zero odds that in real life I would wind up right at this time in history, so I must be a character in a video game who was put right at the perfect time in history. I don’t think that way, but I do think it’s a remarkably weird and fortuitous thing to be living at such a time, and it’s pretty cool. I didn’t foresee that. When I was a little kid in the ’70s, things seemed to be going more slowly, and my plan was to build a spaceship to travel close to light speed and come back to Earth in a million years, when things would be more interesting. But with the reality of exponential change, we can see things are going to get a lot more interesting without waiting a million years; just wait five or ten years or so.

Brian Rose: I think on the last episode, when you were here in the studio, we both said: may you live in interesting times. And we’re in them right now, Ben. I have boys that are five and six years old as well, and I was working on their homework this last weekend, just looking at what they’re learning and thinking: will this be relevant? Will they be learning in this way even in five years? I’m really looking at this as I see the world changing so rapidly, and it’s just crazy.

Ben Goertzel: But then, mathematics has its own beauty, and my kids are learning to play music, even though there are other humans who are better musicians than they are. So I think learning about the world and learning to do fun and rewarding things is going to have value even after there are super AIs around. But it is pretty weird to think about the world our kids and grandchildren are going to grow up into. It’s also empowering to feel like we have the ability to influence that world in a positive direction. People have never had this much leverage as they do now in human history. Ten thousand years ago, how much impact could a person have on the future course of humanity? They could impact their little village. Now each of us could do things that really impact the future course of our species, and possibly of the intelligent species to come after us. So that’s amazing, and also sobering.

1:56:50 Brian Rose: Yeah, very much so. Ben, thank you so much for this conversation. I’ve been looking forward to this ever since the release of ChatGPT, and you didn’t disappoint. All of our followers and students have been waiting to watch this, and again, you didn’t disappoint at all: fascinating ideas and questions. I want to be a part of making sure this transitions the right way, and because of the position we’re in, the knowledge base and the people we know, I think we can have an effect on it in a positive way. Regardless, I want to try. I really appreciate what you’re doing, especially the openness of SingularityNET and all of your projects. Let’s continue having these conversations, hopefully next time in person. In the meantime, what’s the best way for people to follow you, either a website or social, or learn more about your projects?

Ben Goertzel: People can look at SingularityNET.io, and you can find out about all the different projects in the SingularityNET ecosystem there. You can look at Goertzel.org, which is my personal website, which has links to my Substack blog and Twitter, all the different channels by which I pour out my dubious wisdom upon the world. And SingularityNET has an email list you can join, which will give updates on all the different projects I described. So please sign up and follow what we’re doing. There are all sorts of ways that people with different skill sets can join in and help us work toward the beneficial Singularity.

Brian Rose: Fantastic. All right, thank you, Ben. I appreciate that, and we’re going to keep pushing forward. I’m going to be talking to you in the background, getting your thoughts on how we can make the biggest change, so I really appreciate you. I appreciate everybody for watching, and we’ll see you next time on London Real. Take care, Ben, and take care, everybody else. Bye.

London Real logo Chat
×

Private call with Brian Rose

See how the AI Investment Club works, and decide for yourself

Every Monday, AI founders pitch live to our members. Book a call to see if it's right for you.

Check Brian’s availability
Brian video thumbnail
Private call with Brian Rose
Watch Brian’s message