Full transcript: Dr Ben Goertzel on London Real
3:13 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 artificial general intelligence and have submitted over 150 technical papers and published 25 scientific books on the subject.
3:30 In 2017, you founded Singularity Net with the goal of creating a decentralized, democratic, inclusive, and beneficial artificial general intelligence. Your latest project, Zarka, is a supercharged, intelligent, neurosymbolic, large language model at massive scale and promises to take on OpenAI’s chat GPT but go much, much further. You believe that your advanced neurosymbolic AI techniques will revolutionize and disrupt every industry, taking a giant step towards artificial general intelligence or AGI.
4:03 Ultimately, your goal is to realize a positive and beneficial singularity for the good of all humanity by building sentient beings of the future to be loving, caring, compassionate, inclusive, and democratic. Ben, welcome back to London Real.
4:17 Ben Goertzel: Hey, yeah, it’s been a while. It’s good to be here.
4:20 Brian Rose: Great to catch up. Look, before we get started, I just want to say that if anyone is interested in learning more about your new project, Zarka, then go to LondonReal.tv forward slash Zarka to request more information.
4:33 We will be investing in Zarka through London Real Ventures and we’re really excited to support you on this venture. Ben, you’ve been busy lately. I know you recently tweeted that you were shocked to hear that people were choosing not to have children because of AI. I thought I’d jump in there. Why are people refusing to have children because of AI?
4:53 Ben Goertzel: That was a sort of random off-the-cuff tweet. I was amused in just browsing through online news to find a blog post by some random, fairly intelligent, well-educated person explaining why they were afraid to bring children into the world because of what AI might mean for the future. I’d heard stuff like this before regarding climate change. People are saying, well, I’m afraid to bring children into the world because we may all be boiling to death in the ocean or something.
5:36 I hadn’t heard that particular line of thinking, or really I should call it line of feeling. I hadn’t really heard that regarding AI before. Of course, it’s an indication of AI is like grabbing the zeitgeist to have a neck and shaking it back and forth. People are trying to just conceptualize and understand what the evident rise of AI means. I realized how utterly opposite that is to my way of thinking.
6:11 When I look at what I had to go through in my own childhood and early adulthood, it seems on the one hand tremendously easier than what my great-great-grandparents went through in various farming villages in Ukraine and Lithuania before my great-grandparents emigrated to US. On the other hand, what I went through in my early life seems much, much more difficult than what my young children now are going to go through. I have a two-year-old and a five-year-old, and I think they’re probably not going to have to work for a living.
7:01 They’re probably not going to reach their 50s like I have with the knowledge that unless some radical advance opens within a few decades, death is the likely outcome. I think my children who are young now, they’re going to grow up in an environment where material abundance is the reality and the norm and where sickness, disease, and death are historical things, like sort of how we look now back at the black plague or cannibalism or something. So my default perspective is, wow, this is an amazing time to bring children in the world, and what a gift to grow up now instead of in the 1970s or the 1500s, right?
8:01 And certainly, there’s tremendous uncertainty about the future, so we can’t say we’re guaranteed a rosy outcome, but it did strike me how my emotional setting on this was quite different than some people. And I felt like this particular column, this came upon their pessimism without a tremendous amount of education or understanding of AI. It’s more just like weird new stuff, scary, right? And now, on the other hand, if you said, I don’t want to bother to reproduce biologically because it will take a bunch of time and effort, and I should be putting my mind toward bringing about a beneficial singularity, right?
8:53 I mean, then, well, yeah, after the singularity, you can probably have all the kids that you want. Their life will be even easier. You can have a robo nanny, so you don’t have to get up at 3 in the morning and rock them back to sleep, right? So, I mean, I think that would be a more appealing reason to defer reproduction. I mean, I’m happy with my own decision to continue the good old human reproductive process.
9:24 Like, in parallel, working toward the singularity. I mean, because we want to maintain our humanity, right? We want to be human. We want to feel human. We want to be in the thick of human life, and we want that to be the context in which we sort of segue into the next phase when humans will coexist with non-human intelligences.
9:53 Brian Rose: Yeah, I think it just shows there’s this level of uncertainty and fear out there that you get a reaction like that. But I appreciate your perspective on it in a big way. My kids are 6 and 7. And last night, I kid you not Ben, my 7-year-old turned to me and said, what’s going to happen with AI? And he didn’t hear it from me.
10:12 I was like, where did you hear that? He’s like, I heard it on YouTube. And so we started having these conversations, which is interesting. I do think about their teenage years, 10 years from now, and how they could be fundamentally different than any human experience in the last 200,000 years. And it does motivate me to make sure that it’s ushered in in a beneficial way.
10:33 It’s one of the reasons we’re working with you and investing in Zarka as well. But I’m concerned, but I’m not trying not to have kids, and I’m not trying to ship them off to Mars anytime soon.
10:46 Ben Goertzel: It’s funny. My three adult kids who are in their late 20s and early 30s, they grew up hearing me talk about AI, nanotech, robots, quantum computing, teleporting yourself to other dimensions, blah, blah. So now that these things are starting to become in common parlance, they were pre-bored by these things by the time they’re 5 or 10 years old.
11:10 They’re like, well, yeah, yeah, of course AI is going to get smarter than people. What else do you think is going to happen? Of course death is going to be conquered. Come on, isn’t that obvious, right? So really, your view on this is just it depends on your background.
11:27 And what you were shaped by as a kid. But it is interesting how technology is now unfolding faster than most people’s minds are able to change, right? So in that sense, the ones who are 2, 5, 6, 7 years old have a big advantage. Because I mean, if you’re asking what’s going to happen with AI when you’re 6, you’re probably going to be a little less shocked when super AGI rolls out. Whereas people who are around the age of you and me, I mean, you and I are fine.
12:05 We’re fairly flexible minded, but there’s a lot of people our age who are going to have a difficult time adjusting, including a lot of people our age who are in big power positions in the establishment, like running the world, who will have a difficult time mentally adjusting. So yeah, the real worry isn’t that some random newspaper columnists are freaked out and paranoid. The real worry is that the people running governments, corporations, military organizations and the educational institutions are freaked out and paranoid and don’t really understand what’s going on.
12:42 I mean, that’s same psychology, but more dangerous because of the position that people are in.
12:48 Brian Rose: Yeah, that’s a definite worry. You mentioned military, Ben. And there was an article that came out, I think, yesterday in the Wall Street Journal talking about how the Pentagon has now come out and said they have a $1.8 billion program to create these AI weapons to fight China. And this woman went out on a boat, I think in the South China Sea, and there were these autonomous drone boats with autonomous drone flying drones on top of them, and they were all connecting with AI.
13:20 And they were talking about how this is the future. I wanted to get your thoughts on that, and I know you just spent a month in China. So when you hear something like that, what’s your first reaction? And also, having your boots on the ground in China, how would you say that the vibe is there when it comes to AI? So these are two different and very interesting questions.
13:42 Ben Goertzel: So my first reaction to you coming on the Pentagon is like, $1.8 billion is chump change when you’re talking about large-scale AI meets large-scale military operations. So I mean, how much has been spent on Ukraine more so far, which isn’t even that high tech, right? So I mean, I think if that’s really all the money the Pentagon is putting into advanced AI, then the US is in military trouble. But I don’t think that’s the case. I mean, as you know, Obama magically synthesized $4 trillion in 2008 or so in the context of the financial crisis and sort of fed it to Wall Street banks and insurance companies.
14:25 So I mean, certainly DARPA, the Defense Advanced Research Project Agency, which has been the US military’s primary research funding agency, they not only invented the internet. I mean, they also basically invented AI, right? And it was DARPA funding over many decades that fueled AI through 60s, 70s, 80s, 90s, and so forth. So US military is all over AI. They’re super close with MIT, Stanford, Berkeley, everyone else.
15:00 They’re super close with Microsoft, Amazon, Google, and so forth. So I suspect the specific funding chunk you mentioned barely scratches the surface of what they’re doing in advanced AI across the board. And I mean, on the one hand, it’s what they have to do given geopolitics and the position that they’re in. On the other hand, and I would say there’s no one more serious about AI ethics on the planet than US military. I mean, they’ve been publishing papers and doing prototypes and tests on ethical AI for many decades now.
15:45 And they’ve worked on encoding ethical principles and formal logic systems connecting these with neural net. So there’s really quite a mathematical science behind what the US military does in terms of when can an AI system decide to kill someone? And they’re not currently doing that, to my knowledge. But when can an AI system decide to let someone die through inaction, which is a harder one, right? What support should the AI system give to what parties of what rank in what kind of circumstances of limited communication?
16:24 I mean, so they’ve gone through that in a very serious and careful way. On the other hand, it’s still done within the high-level goals of the whole military organization. And I mean, with the way geopolitics is, it’s necessary. Like, there’s terrorists out there who want to blow up US cities. We’ve got to stop them, right?
16:49 On the other hand, US military is far more adventurous than I would like historically in going and blowing up people in Middle Eastern countries for no apparent good reason. So there’s a lot of subtlety there. But I think there’s way more money than that going into it. And there’s so much sophistication going into it. Now, going to China, where I think China is outdoing the US is in AI for surveillance, actually.
17:24 I mean, I think for military, US, for better and worse, says by far the best military on the planet, and US military invented AI, we’re going to have by far the best military AI on the planet. It’s not universally a good thing, but there are thoughtful people and sophisticated people involved with it. So it’s not likely to lead to a rogue AI. It can lead to AI that efficiently executes what the US military wants, which is a plus or minus, depending on your view on their various adventures, right?
18:02 In the intelligence community, I think the US historically is very fragmented, and the intelligence community has lacked really high-powered tech savvy. And I mean, of course, NSA has a lot of savvy about encryption technology, which is a sort of island of genius. But overall, you couldn’t say, like, you know, CIA, NRO, FBI, like, INSCOM, you couldn’t say these US intelligence organizations have tech teams to rival Google or Microsoft or Facebook or something, right?
18:40 And typically, different Intel agencies have had siloed data stores that they each want to sit on and own so that they’re the ones who get the goodies for discovering some bad guy, right? So they are, of course, aggressively rolling out AI across the US intelligence community. But they’re, in some ways, slowed down by these aspects I’ve mentioned, just fragmentation and lack of internal tech savvy. Like, not people with the right top secret clearances who are super advanced in AI.
19:20 Because if you have super advanced AI knowledge, you can make more money of a less annoying life working in industry than getting a TS clearance and working for some intelligence agency. Not like in cryptography. Historically, like, NSA was the best job, right? But it’s never been the best AI job to work for US intelligence agencies. China, on the other hand, for better and worse as well, has a quite different setup, right?
19:50 So in China, they don’t have a big problem with data integration across different branches of their government. And they actually have rolled out blockchain in a very impressive way across all levels of Chinese government with a view toward reducing corruption and having sort of decentralized peer-to-peer checking of data accuracy. So blockchain is a very flexible tool. You could use it to make a truly richly decentralized network like we’re doing with SingularityNet and HyperCycle and Nuna and other projects we’re doing.
20:26 You could also use it to make a system that enforces accountability and transparency in a system that’s controlled by a few guys at the top, which is how Chinese Communist Party has deployed blockchain very effectively. They have no problem integrating data across all their major companies and government agencies. They also have the ability to take the best AI people in Tencent by do Alibaba and have them work directly on military and intelligence AI projects as part of their day-to-day job within their tech company, which is not what US government does.
21:04 I mean, US government has close partnerships and collaborations with every big tech company, but they don’t actually say, well, give me a hundred of Google’s best guys this year to do this espionage thing, right? I mean, they spun off Palantir from Google Alumni and there’s a lot of back and forth of Google and Palantir. So I mean, they do stuff sort of like that, but Chinese government could just do that, right? So I mean, I have little doubt that the biggest and best large language model in the world is the one that looks at all the data from WeChat, Taobao, and Alibaba, which is Baidu, which is proprietary to Chinese government and is used for them by all sorts of purposes.
21:54 And another amusing thing, just anecdotally going around in China, you can’t use a Visa card hardly anywhere anymore. And you have to fight with people to get them to take cash. Like everything is WeChat pay and it’s a pain in the ass to connect your foreign credit cards to WeChat pay actually. So I guess just cause there hasn’t been that many foreigners there in the last few years. So they’ve sort of let the infrastructure for letting foreigners pay for stuff sort of dwindle, not out of intention, but just in attention.
22:30 But everything was on WeChat three years ago. Now it’s even more insanely the case. Like every payment you make is there, every message you send is there. Then if you want to use like the weightlifting machine out in the park, you scan your WeChat QR code to activate the weightlifting machine. Then it tracks how much you lift it.
22:54 It gives you information about how many calories you burned. It also registers in some government database like who you are, where you were, and what is your body condition, right? So what’s striking is like it’s really, really, really convenient. Like being able to pay for everything with WeChat pay, it’s far better than fucking around with your credit cards or Apple pay or cash or something. Like it’s super simple.
23:18 You just whip out your phone or don’t even whip it out, like leave it in your pocket and bing, bing, bing. Everything’s taken care of. And even the fitness tracker thing, I mean, sure, you can be at the gym, you can be at the park, you can be in your house, all that data about what you did is there. You can feed it in. You can graph your body’s condition, right?
23:38 It’s very convenient. It helps your life. It’s all good. People don’t complain about it because it’s actually helpful. Very good user interfaces, good data integration.
23:49 On the other hand, it’s also all feeding in to a big centralized database that’s coordinated by advanced AI, which can in many ways be used to optimize the economy and be for the good of the people. Like you can see the old complaint against centrally planned economies where, well, it’s just too much data. No one can coordinate all this stuff. So now, okay, you’ve got advanced AI, you’ve got a centrally controlled blockchain for coordinating all the data. Like we’re getting to a point where that objection is not that meaningful, right?
24:30 And maybe you can do much better with a centrally planned economy using current tools. It still, however, comes down to the vision and ethics of said central planners, right? Which is a whole different topic. The other interesting thing there is three years ago when I was last in China, and I lived in Hong Kong for 10 years, as you know, before in early 2020, relocating back to the US. AGI was treated with some suspicion.
25:08 There was sort of an attitude like the government doesn’t like AGI because it sounds hard to control. They used euphemisms like we like intelligent science, right? I mean, they like applied AI. They just didn’t want the general thing. They want AI to be doing very specific things in a way that was easily delimited and easily controllable.
25:28 Now, dozens of conferences in 2023 all had AGI in mainland China, and probably 70 different companies have trained to put out their own Chinese large language models. These are among the presenters at these dozens of conferences focused on general intelligence, right? So suddenly in China, AGI has gone from being something you weren’t supposed to talk about or think about to being a big focus. Now, that doesn’t mean I think China is likely to be where a big AGI breakthrough comes from.
26:09 Actually, I still think it isn’t because I think the environment there is very unwelcoming to technology risk. It’s very hard as an entrepreneur or as a young academic seeking research funding. It’s very hard to get funding for an idea that’s not proven somehow. It’s much easier to get funding for improving something that was done before or taking something that was done outside China and putting it into China. And I mean, of course, everywhere in the world it’s easier to get funding for copying stuff or improving stuff than doing new stuff.
26:51 But in China, that’s even more strongly the case than in the West. I would say on the business side, investors there are very comfortable with market risk and very uncomfortable with technology risk. And this is in large measure… It appears to me, and I’m not Chinese, so what I say is probably worth shit. But I mean, it appears to me this is in large measure a cultural issue rather than just a government issue.
27:21 It seems like there’s a certain level of mistrust. Like if I invest money in this unproven thing, are they going to just piss away the money or like put it in their overseas account or spend it building some side business? How do I know they’re really spending it on the research project that I’ve funded? Because it’s so hard to quantify incremental progress in research. And it feels like there’s a bit more trust of this nature in the West that if you fund some crazy young guys to work on a crazy idea, at least they’re really going to burn the money working on that crazy idea instead of using it to start a restaurant or something, right?
28:01 So that seems to me… Seems to me there are some cultural factors that will change only with generational change, meaning over decades, not years. Seems to me there are some cultural factors that make it hard for the Chinese system to fund early stage innovation. Even though like the young grad students there have as many wild innovative ideas as anywhere on the planet or more so, right? It’s just when I look at the young people I knew in China five, 10 years ago who had brilliant crazy ideas about AGI, longevity science, quantum computing.
28:44 When I see what they’re doing now, they’re mostly working on advanced and high tech, but relatively sort of boring and conventional jobs because that’s what they could get paid to do. And then there’s other cultural aspects like as a young person in China, you sort of have to get married, have a kid and buy an apartment or have your parents buy you an apartment. There’s not as much acceptance for whiling away your late 20s and early 30s living in your parents’ basement hacking on random shit that may not come to anything.
29:21 And that’s sort of an accepted thing to do in US West Coast at least and more so across the West, right? So I think, I still think the AGI breakthrough is by far more likely to come out of the West. I mean, I think it’s most likely to come out of my own team, of course, but setting that aside, if we’re not the ones to do it, I think it’s going to be someone else in US or Western Europe, almost surely. On the other hand, as soon as that breakthrough comes, China is so fucking poised to take that, borrow it, improve it, deploy it at massive scale and use it to control their population, optimize their economy and work toward the sort of hegemony over their region of the globe, right?
30:15 So that which, on the other hand,
30:18 Brian Rose: you can see their economy is struggling significantly. Can you see that? Can you feel that when you’re there? Because we hear the articles.
30:30 Ben Goertzel: Yeah, definitely.
30:32 I think it’s very much a class division thing though. Like if you’re in Beijing, talking to AI researchers or software entrepreneurs, I mean, it’s glory days, right? There’s loads of money around for projects, so much money flowing for AI, so much interest, so much awesome stuff being built. The pollution is much, much less in Beijing than it was before. There’s like dozens of brands of electric cars out in the street, right?
31:04 I mean, in some ways, everything goes through WeChat and the phone. In some ways, it feels like the future. Then you come back here, we’re like fucking around with paper money and credit cards. Even in Europe, many places, you’re screwing around with cash all the time. China, everywhere has face recognition.
31:24 It’s spooky in a way. On the other hand, it’s very, very convenient. You take it for granted, you go everywhere, your picture’s taken, all the systems know where you are and things are fast, simple and efficient, right? I mean, I think if you’re a technical professional or a scientist or something in China, it’s not hard times at all. Incredible times, which it also is in the West, right?
31:46 I mean, it’s an amazing time to be an AI professional anywhere. On the other hand, in my wife’s home village, which is a rural village with like a thousand people in Anhui Province or in the small cities around there, there’s clearly high youth unemployment and high underemployment, right? So you can see among less educated people who are maybe not in the big cities and who are not working in the high tech industry, when I first started going to China in 2007 or 2008, all these people were getting rich on real estate.
32:23 They’re like, yeah, why are you bothering to do technology? I made so much money just like reconditioning this apartment and rolling it over, right? So I mean, the attitude then was just people with no education were getting rich from doing sort of small scale local entrepreneurial stuff, largely real estate related somehow or another. Now, a lot of those people are unemployed and struggling or employed for a lower wage than they would like or something. So I mean, I think you’re seeing a big class division there and to some extent an unraveling of the communist safety net.
33:04 Like health care is not subsidized nearly as well as it used to be, right? So I mean, the best health care there is great. And in many ways, the average health care there I think is better than in the US, which is a fucked up health care system. You can get a lot of stuff done by very competent doctors at low cost and they’re better at preventative medicine and chronic disease than the West. But if you’re poor there, it’s no longer quite guaranteed to get the medical care that you need in a reasonably efficient way, which it kind of more so used to be, right?
33:39 Brian Rose: So that’s good observations.
33:41 Ben Goertzel: Yeah, I wouldn’t say like the economy is unraveling or something, but people don’t think it’s like universal glory days anymore. Like there used to be a vibe like China is going to overtake the US any year now. Like, ha ha, we’re taking over the world, right? And not in the military sense, but in the sense of just becoming the biggest country, because there’s the most people there, it makes sense, right?
34:12 And it’s a country that manufactures the stuff everywhere in the world uses. There’s definitely an awareness that growth is slowing down now. And the whole thing of like emasculating the billionaires and Jack Ma and all this, on the one hand, it makes total sense. I mean, Jack Ma did not do enough to deserve billions of dollars. No human being ever has so far.
34:40 So far, really. It’s just these sort of positive feedback cycles in the capitalist economy that mean the more money you have, the easier it is to make more money, right? So in a big picture sense, okay, it makes sense for society collectively to decide, you know, maybe you being in the right place at the right time doesn’t mean you personally deserve a hundred billion dollars. Maybe three billion dollars should be enough for you, right? On the other hand, the way that was rolled out definitely sort of tempered the entrepreneurial dream mentality of a bunch of young Chinese because even though the reason most people become an entrepreneur isn’t because they want to make a hundred billion dollars, right, just because they want to make their thing and put their thing out there.
35:31 And yeah, they want to do our right financially along the way. Still, these tech billionaires, you know, like Elon Musk, they have a sort of iconic status and to see the government sort of taking these people down in a few notches definitely biases young people to not take huge risks in their life because there’s the feeling like you may take a huge risk and have a big success and then you accidentally take the wrong turn and the government squashes you, right? And then you start to think, well, you know, why bother?
36:04 Why not just get an engineering job for Huawei, which pays well and won’t get me in any trouble?
36:09 Brian Rose: What’s the feeling about Xi there? Like, what do people think about the premier? Do they talk about him? Or is there a feel?
36:15 Because obviously he’s kind of grabbing more power, turning more communist than capitalist. Obviously, you know, battering down the Moz is there.
36:24 Ben Goertzel: I wouldn’t say he’s turning more communist than capitalist. I think he’s breaking new ground. Actually, it’s not very communist what he’s doing.
36:33 Marx wouldn’t like it terribly much, right?
36:38 Brian Rose: How so?
36:39 Ben Goertzel: I mean, the power is not going to the workers, right? The power is very much with the OE. I mean, it is a highly centralized sort of paternalist state capitalism. And he wants to foster small business very much. He wants to foster tech entrepreneurship very much.
37:01 He just thinks once the company gets too big, it essentially becomes a common good in a public utility and the government should regulate it. And that’s not entirely stupid, right? Like, in a way, Google is a public utility. Facebook is… Yes, they are.
37:19 These things have tremendous influence on politics, on the way the whole society thinks. So it’s not that stupid. To think, well, okay, maybe the government that represents all the people should rule this rather than a few rich executives, right? So I mean, that doesn’t bear too much resemblance to anything that Marx said because Marx didn’t know about computers and modern networks and social media and all that, right? So I think he’s going in a new direction.
37:50 It’s not like he’s organizing worker collectives and empowering them more and more like you would if you were a real old-style communist. He’s really a technocrat, right? I mean, I think the feeling about Xi in China is, from my own reading, which again has limited meaning, is probably complexly mixed. And I think there’s an attitude across Asia, not just in China, but you could say in Singapore as well in Korean democratic countries, there’s an attitude there that you want a strong leader who will keep order on things and plan things in a sensible way and that’s how you make progress, right?
38:35 And that worked for Singapore for a long time, right? And they keep democratically electing people who act like what in the West we’d consider almost dictators, but they do democratically elect them, right? So I think in Asia there’s a lot more liking for strong leaders, not just in the right wing, but like among everyone. And they do view Xi as a strong leader. But I say that there’s also been pushback against many things she has done by the Chinese people, right?
39:09 And I think that he’s been responsive to that pushback, which is, in the end, it’s a form of democracy, although not the Western style, right? So about COVID, I mean, they locked down tightly, but then everyone was getting COVID anyway. I mean, you had protests in the streets, you had half the messages on Weibo, we’re all about, these COVID policies are dumb, they’re failing, we’re all getting COVID, just let us out. And he did, right? I mean, he totally reversed the COVID policy in response to popular demand.
39:41 So now, in the last few days, he’s trying to make new restrictions, like we want to restrict any public speech or art that will be emotionally traumatic to the soul of the Chinese people or something, right? So it’s almost like a wokism thing or something. And there’s huge pushback against that also. People are like, well, this is vaguely defined, it doesn’t mean anything. Like, how can we know what will wound the soul of the Chinese people?
40:14 This is stupid. It means you’ll be afraid to say anything. So I think he will respond to that pushback and will sort of at least water down that edict, right? So I think the people appreciate though, like they can see when they rise up in mass and complain about something, he’s not actually shooting everyone. He’s actually changing things based on what they say.
40:40 And in some ways, that works better than Western elections, right? Like, I mean, it’s a different process. People gripe on social media and go out in the street and yell, then the dictator changes things. But that can actually be faster and more responsive than our electoral system, which is very slow and inefficient at responding to public sentiment, right? So it’s quite interesting.
41:04 But getting back to AI though, one conclusion I came to from talking to my friends in the AI world in Beijing and Shenzhen. And I spent a while with a good friend named Zheng Yi, who’s probably China’s leading AI researcher, who has a project called BrainCog, but a bunch of other AI people there also. It did seem to me that both open source and, you know, open science, like posting your research on arxiv.org and so forth, these are going to be very valuable sort of connecting tissue in the next phase.
41:44 So, I mean, BrainCog is open source. OpenCog Hyperon is open source. SingularityNet is open source. When Zarkar rolls out its code for training neural symbolic AI models, that will be open source, although all models may not be open source, depending on what the training data was that went into them, right? So I think we can then collaborate on integrating these different AI code bases in the open source domain.
42:14 And no sanctions are stopping that, right? Because it’s not directly an economic thing. It’s just people, you can even collaborate in an open source project with people in Iran and against which we have all these sanctions or North Korea if they were able to get online, right? So, and the AI world has really run on open source, right? And we discussed this in our previous calls.
42:35 This was the vibe behind Google’s, like there is no moat. Yeah, the leaked memo. The leaked memo about five months ago about that. Yeah, it was a leaked memo where some internal Google engineers were saying, look, the problem isn’t that open AI will outdo us. They may for a while, but we can catch up to them.
42:55 The problem is that 100 projects in the open source world will do as well as us and then outdo us given a little bit of time because all the key research ideas are published in open literature and the key code bases are open source also. And once a paper is published, even if the code behind that paper wasn’t opened, someone will reverse engineer from the paper and put it out there, right? So I think the no moat paper was correct. And I mean, I think this is partly how all these Chinese companies built their LLMs, right?
43:32 But I think it also is a real bright spot geopolitically because it means, you know, we can put sanctions on chips and hardware and we can try to control flow of money. But the internet is decentralized. Linux is decentralized. Open source code projects are not approximately economic things. And so shutting those down would take unprecedented measures, right?
44:02 Because like even the sanctions against Iran don’t stop you in collaborating on an open source project with researchers at University of Isfahan or Tehran or something, right? So I think fostering open source collaboration globally is going to be quite interesting and important. And we’ll go hand in hand with open science research leading to co-authored research papers on algorithms in open source code. So this is one thing I’m going to be looking at a lot in 2024 with OpenCog HyperOn project, which is my big push toward AGI, which will be leveraging Zarka’s large language models for its LLM component and then adding other components like symbolic reasoning and evolutionary learning.
44:54 But we’re going to launch the alpha version of OpenCog HyperOn March or April of next year. I mean, we have a pretty we’re already working with it with the pre-alpha version, but I think we’re going to get all the documentation and test cases and examples and stuff together to declare an official alpha version around March or April. At that point, I want to make a big push to get an open source developer community in mainland China to help with taking OpenCog HyperOn to the next level and in Russia as well and everywhere on the planet.
45:31 And I think this intellectual collaboration when creating the AI software cuts past all these political and money things in a quite interesting way, which to my thinking is for the good of the planet.
45:44 Brian Rose: And part of what you’re doing there, you’re connecting with the young Chinese developers who are very modern minded, right? And who, if you gave them the choice, would probably radically democratize China, although it’s not like where their emotional it’s not what they’re investing their time is now because there’s no practical way to do it, right?
46:08 But I think you’re connecting the more young modern and globally minded Chinese when the young modern globally minded people elsewhere in the world and you’re doing it in a domain that’s sort of off to the side of all the military and trade sanctions and power brokering stuff. Would that make OpenCog kind of the biggest open source AI project in the world then if you do what you want to do by next year?
46:36 Ben Goertzel: Well, I think to get it to be the well, Linux is the biggest oh, Linux is not an AI project, but it’s the biggest open source AI project in the world that open source project.
46:48 And I think I think to become bigger than things like SciPy or PyTorch say which are machine learning toolkits we would need to demonstrate a tremendous AGI breakthrough, right? So if we do something with OpenCog, Hyperon, and Zarka, which is probably way smarter than GPT4 or 5 or whatever the latest version is by that time, I mean then yes, I think that rapidly becomes the biggest open source AI project in the world. But I also think calling in a global community of AGI developers can be helpful to get to that breakthrough faster.
47:40 Brian Rose: Yeah, that’s really exciting.
47:42 Ben Goertzel: I think we can get to that breakthrough ourselves also with just my own team who’s working on it. On the other hand, now that AGI has become more popular and it’s sort of socially acceptable thing to work on. I mean, I think you can pull in geniuses from around the globe to help. And I would love to do that at this moment.
48:08 And of course, if someone wants to jump in and help with OpenCog, Hyperon right now, the code base is open source. We’re happy to welcome anyone into the project at this moment. I’m sort of holding off on making a big recruiting push to pull in open source developers until we’ve formally launched the alpha version early next year, just because the barrier to entry technically is somewhat high now to the early stage nature of the code. I mean, the code base is actually quite capable and learning and reasoning.
48:49 Well, we haven’t put much focus on usability, right? So if we’ve got a bunch of junior developers in there now, we’re going to spend all their time holding their hands and not enough time not enough time finishing the alpha. So I think we’ll have a huge push on that in the beginning of the second quarter of next year and probably start doing AGI hackathons in at least Shenzhen and Beijing, probably Shanghai as well. I mean, as well as in US and across the West. I think it’s super exciting.
49:22 Brian Rose: And I think I want to know a little bit more about this because like you said, there was this leaked memo from Google. We have no moat and the biggest threat is open source. And now you’re talking about opening up yours or they’re already open, but OpenCog being worked on by people in Russia and Iran and China all to make something great in AGI. Meanwhile, you’re looking at the open AI system, which is probably a bunch of people in Silicon Valley competing with Google bard or whatever Google brain, a bunch of people pretty much in Silicon Valley.
49:54 I mean, how does that play out?
49:56 Ben Goertzel: There’s a lot of them right here in Seattle area.
49:59 Brian Rose: Seattle too, okay. How does that play out? How does that play out though over two years?
50:05 Like do you think they are destined to fail? Do you think that their closed model will mean that they can’t innovate and have all these emerging developments like we saw with that leaked memo? There’s rumors that OpenAI is spending tons of money just to keep that model going. Do you think that the end game will always be won by open source or what are some of the challenges they have in their own setups?
50:33 Ben Goertzel: I think in principle, yes, in principle I think open source has a big advantage for breakthroughs and very difficult things.
50:41 And it’s the same reason that Linux is the number one operating system on the internet, right? And I think open source has not been the best way to make like usable retail and end user products, right? But although, I mean, by now, by now I’d say the Android phone ecosystem in many ways makes things that are more usable than Apple. And that’s an interesting mix of open source Android with like proprietary stuff by Samsung, OnePlus, Xiaomi, blah, blah, blah. But on the whole, for user facing stuff, commercial companies have done great.
51:25 But for like very deep, difficult back end stuff, open source has always done better in the software world, right? So in principle, yeah, the open source community is going to outdo anything any big tech company can put together. And they know that, which is why they all have major open source strategies. So they’re trying to play a habit both ways game, right? So Facebook, for example, they open source LAMA2, which is almost as good as GPT-4, not quite.
52:04 And their code LAMA, I think is better at generating code than GPT-4. And they’ve just opened that model up, right? And MusicGen, Facebook opened up. I mean, Google made MusicLM but didn’t, and made it usable by API but didn’t open the model. Facebook opened up MusicGen.
52:28 Now people are building all sorts of stuff on top of that and I’m playing with it myself and with my musician hat on, right? So Google hasn’t gone as far as Facebook and opening stuff up. But obviously they’re working, I mean, they released TensorFlow. Actually Torch, which was Facebook, has been more successful than TensorFlow overall. Now, Microsoft famously has pivoted toward Linux in their operating systems after being allergic to Linux for a long time.
53:06 There’s a lot of Linux under the hood in Windows now. And Azure, which is Microsoft hosting services, a lot of Linux, right? So OpenAI, in spite of their name, is one of the less open entities out there in the modern AI world. On the other hand, Microsoft, who they’re so closely allied with, has quite sophisticated open source strategy. So what it means is that the economy you posed isn’t quite how the world works, right?
53:43 It’s not just open source project versus closed proprietary project. If it’s that, yeah, the closed proprietary project loses, sure. But the question is more grassroots open source project versus corporate driven open source project like a PyTorch or MusicGen or something, right? Or Llama. And that’s in the big picture you’d say less obvious, right?
54:12 Because these corporate driven open source projects, they also get a big ecosystem to build on top of what they’re doing bringing a lot of collective brain power. Yet the updates of the core open source software releases like Llama 1, 2, 3, 4, blah, blah, or they are orchestrated by the big company, Facebook, in the case.
54:36 Brian Rose: So how’s OpenAI doing right now, do you think? And what do they have to do?
54:42 Ben Goertzel: Great, I mean, if we were running OpenAI, we’d be very happy.
54:46 Because why? They’re getting tons of uses, users, they’re getting funded by Microsoft. They have gotten mindshare that is almost unthinkable, right? I mean, they dethroned Google in terms of being like the corporate AI leader on the planet. But Google’s responded.
55:07 Google’s responded. They’re working Bard. They’re working that in the search engine. Not badly, but Bard sucks worse than GPT-4, right? It’s not the best.
55:14 Brian Rose: I use it. It’s not the best.
55:16 Ben Goertzel: Yeah, yeah. I’m not entirely sure why. Because earlier versions of Lambda inside Google from people I know who’ve played with them were better than Bard is now in many ways, right?
55:31 So I think Google in some ways is still suffering from an internal conservatism which led them to be beaten by, beaten to the punch by OpenAI in the first place. Like that same internal conservatism is still bogging down Bard a bit, right? Like it’s not that easy to refactor an organization and make it agile and responsive like a startup. Like Google has tried that in so many ways over a long time always with non-zero but limited success. So I mean, so yeah, I think it’s a good time to be OpenAI.
56:08 It’s a good time to be Microsoft. On the other hand, their path to dramatic success is really predicated on the hypothesis that transformer neural nets are the key to AGI. So if transformer neural nets are the key to AGI, even though transformers were invented in Google brain in Mountain View, I mean, if transformer neural nets are the key to AGI, well, there’s a decent chance that OpenAI is going to get there, right? I mean, I think DeepMind may be doing some interesting things by putting together the technology behind Mu0 and AlphaZero and AlphaGo with large language models.
57:02 So I wouldn’t count them out either and I wouldn’t count Facebook out either. But on the whole, if transformer neural nets plain and simple are the way to get there, OpenAI doesn’t have like an unassailable lead but they obviously have a lead. And Microsoft’s AI team is way stronger than OpenAI’s AI team, right? So they’ve got so many brilliant AI people in Microsoft and they have Azure to deploy things and they have a lot of money in the bank. I think that the risk to them is really that they’re wrong about what’s the right path or what’s a workable path to AGI and they’re sort of doubling down on that path.
57:50 They’re hiring people who are true believers in that path, right? I mean, sort of like IBM with Watson, like if IBM had been right that expert systems were the path to AGI, IBM would be God right now, right? They’d be the biggest company in the planet because they doubled down on expert systems. They hired the world geniuses and expert systems. They built the world’s best expert system, which was Watson.
58:18 It wasn’t great, but it’s the world’s best expert system, right? It was the best AI based on encoding expert rules and trying to use that to answer people’s questions. But they were wrong. Expert systems were not the holy path to AGI and so now they’re having trouble playing catch up in the AI space. In spite of having an amazing genius AI team, it’s just the composition of that genius AI team in IBM.
58:48 It was heavily standard sort of toward a certain type of AI which ended up to be interesting but not as successful as other things. So that can happen to all the big tech companies now, right? Like they’re hiring everyone who loves transformer neural networks and paying them huge amounts of money and they’re building great infrastructure for running transformer neural networks. Now, transformer neural networks are awesome. They’re going to revolutionize the economy.
59:16 They’re going to make people a lot of money so they’re doing amazing things. On the other hand, if they’re going to hit a wall and they’re not really the path to AGI, what you’re going to find is these big tech companies have all built infrastructures and human organizations that are barking up the one tree of transformer neural nets, right? So then if a better path to AGI is, say, simulate the brain like Zheng Yi in Beijing is doing or take a transformer neural net and put it together with a logic engine and an evolutionary learning system like Zarka and true AGI and singularity that they’re doing, right?
59:57 If one of these other paths turns out to be a better route to AGI than just beefing up transformer neural nets, plain and simple, then you’ll find these big companies they will either be left by the wayside or they’ll just acquire again, right? So, I mean, Microsoft was being left by the wayside in a sort of IBM-ish way but they made a smarter acquisition than IBM has done in AI. I mean, IBM made a smart acquisition of Red Hat Linux which sort of saved their enterprise server business but they never made a smart acquisition of that magnitude in AI whereas Microsoft did that, right?
1:00:43 But then the success of that sort of acquisition is mixed and I’d say the success with Microsoft and OpenAI is not yet known, right? I mean, the integration of Microsoft’s AI team and OpenAI team is certainly far from complete if things go perfectly there’s amazing synergies there but from what I can see from people that I know in Microsoft it’s not being a wholly smooth integration so far and we’ll see how that unfolds.
1:01:24 Brian Rose: What does chat GPT-5 look like? If you had to guess what does it do?
1:01:31 Ben Goertzel: Well, GPT-4 is a mixture of experts model which has now been leaked and it’s pretty clear so it’s a bunch of different networks that are sort of glommed together to give a composite answer and I think we’ll see GPT-5 also be a mixture of experts model because that’s what OpenAI knows how to do. I think we’ll see tighter integration of various plugins doing more sophisticated specialized types of AI so the Wolfram Alpha plugin is quite good for science queries of math queries even some like business spreadsheet queries and so on. I think you’ll see Wolfram Alpha various tools for doing spreadsheet stuff and numerical calculations you’ll see a bunch of more specialized tools integrated into the guts of the system where you don’t need to install a special plugin to use them which will let the GPT-5 do a lot more different things.
1:02:48 I think you’ll see different kinds of memory integrated with it so it will have more of a working memory where it just keeps in mind everything that’s gone on in the dialogue with you so far which doesn’t do very well now. I think it will have a better episodic memory where it just keeps track of everything that’s happened to it in its life and tries to learn from it. It will have better reflexive modeling like better sense of who it is and how it’s related to you but only a little bit better because that’s hard within their model and I think a bunch of reasoning techniques like chain of thought and graph of thought modeling which are used now by special prompt engineering to try to get the LLMs to think better.
1:03:39 I think these can be baked more into the system so it’s a bit better at thinking step by step. You’ll also see multimodal integration though, right? So, I mean, we haven’t yet seen cross-modal LLMs with images and video and text like all glommed together into one cross-modal system but we know how to do it. It’s just money and Microsoft has a lot of money, right? And it’s scalable infrastructure, right?
1:04:11 So, it’s like how do you do low-level coding to make your NVIDIA GPUs to all this video stuff in a reasonably efficient way, right? I mean, Microsoft has many people who are good at that kind of thing. So, I think between all these improvements, there’s a lot of room to make LLMs smarter at reasoning, better able to handle different kinds of data, better able to sort of impersonate a coherent agent when you’re interacting with them. There are many steps you can take there still without having a system that is on the direct path to AGI rather than being a sort of off-ref to AGI as Jan Lekin said because even with all these improvements, I mean, such a system will not be able to write a high-quality scientific paper.
1:05:08 Such a system will not be able to do the strategic thinking needed to run a company. Such a system will not be able to write poems or compose music that are not dull and banal and derivative, right? So, I mean, I think there’s still a lot of mileage to get from LLMs. We’re not hitting the wall this year by any means.
1:05:35 Brian Rose: But where do you hit the wall?
1:05:38 Ben Goertzel: It’s GPT 5 or 6, depending on how they define the boundaries. Then they hit a wall and then they realize you just need something else, right? And Microsoft research has more people who understand the breadth of the AI field than Google does or Facebook does, actually. It’s been very heterogeneous. Those people may not be highly friendly with the open AI guys at this moment in time, right?
1:06:16 But you can see where the wall will be hit and you’ll need to introduce quite different ideas than what they’re doing in GPT 4 or what we’re doing in 5 to go around that. But again, big companies typically solve that problem by acquisition and you can’t say that couldn’t work. What I am working toward, though, is that by the time they hit that wall in a couple years from now and they realize they need to do something different besides glomming together transformers, once they hit that wall, like we’re already very advanced in developing AGI systems with a richer architecture, right?
1:07:06 So if by the time they hit that wall, we have an advanced OpenCog HyperOn system that can do more advanced reasoning, that can do more fundamental creativity that’s running on a decentralized infrastructure, right? So if we’re already there by that point, well, guess what? They can’t buy us. We’re an open source project centered on a decentralized blockchain network. We do have commercial companies associated with it.
1:07:34 But even if one of those commercial companies doesn’t acquisition exit, it doesn’t let anyone buy the whole decentralized AI ecosystem, right? So I think then they have to play with our ecosystem rather than trying to buy it or destroy it in the same way that they’re playing with Linux rather than trying to buy or destroy Linux. But they fundamentally think they can get AGI with the chat GPT 567 or at least that’s what they’re betting on. Sam Altman appears to seriously think that I feel like Ilya Sutskever probably thinks that.
1:08:14 So I think there are a lot of people who think that around in big tech companies. So Yann LeCun who runs AI for Facebook totally doesn’t think that. Okay. He thinks you can get there with a neural network, but he thinks it needs to be a quite different sort of neural network than the transformer neural networks are. Demis Hassabis from DeepMind, he thinks you need to put a reinforcement learning system and a sort of strategy system like you have in mu zero.
1:08:49 He thinks you need to put those together with something like an LLM to get there. So I think in open AI, you have a lot of people who really think the GPT series can get to AGI. If you look in Facebook or Google, they probably think LLMs solve more of the problem than I do, but they also think you need a bunch of other quite big sophisticated components to plug in there. And then you get to the next level of technical intuition where Yann LeCun and Demis Hassabis think the best route to get to AGI is to take large language models and add other sorts of neural networks onto them.
1:09:39 So you have a hybrid architecture, but each component of the hybrid is a deep neural network. They just don’t think the transformer is necessarily the right kind of deep neural network to play the central role in an AGI architecture. Whereas the intuition that I’m following and my colleagues in OpenCog Hyperon’s architecture AGI singularity, the intuition we’re following is the best route will be to take an LLM, combine it in a hybrid architecture with other components, not all of which are neural nets, but one can be a logic engine, one can be an evolutionary and learning engine, right?
1:10:20 So then that’s a different divergence of research intuition. And of course, I have my own strong opinion from an empirical view. I mean, the jury is out because no one’s gotten to AGI yet, right? And we’re at a stage in the evolution of the field now, though, where a number of different approaches are being actively pursued in parallel with reasonable amounts of funding behind them, which is an unprecedented situation in the history of AGI.
1:10:59 Brian Rose: How confident are you that you’re right, Ben?
1:11:03 Ben Goertzel: I’m very confident we can get there. I’m 99% plus confident that we can get there with the architecture that we’re pursuing. The question is whether, say, DeepMind also has a way that could get there, right? Like, I don’t think OpenAI does. I don’t think Facebook does.
1:11:28 DeepMind has a pretty rich and diverse combination of AGI researchers, and their approach is not so naive. They’re taking many different neural net components, doing useful things, and connecting them together, right? So, I mean, I’m much less confident in their approach than I am in the one that we’re taking in my own team, but I can’t rule out that they may also have a different workable approach, right? I mean, just like Linux is great. Okay, Windows is not as great, but it does work, right?
1:12:07 It’s a workable operating system. There can be many different ways to solve a problem. And the other thing the Linux Windows example shows you is Linux is fundamentally better, but Windows became better for some purposes first, just because it had more money behind it, right? And it had a focused development team with a pile of money. So, there can be multiple ways to solve a problem, and the way that’s fundamentally best isn’t always the way that goes fastest.
1:12:40 Brian Rose: But in AGI, like, whoever develops it the fastest wins, is it a winner-take-all here?
1:12:47 Ben Goertzel: It’s not necessarily that way, it has that aspect, but because, I mean, to take an analog to human cognitive development, which isn’t entirely perfect, but let’s suppose someone has a breakthrough and they create an AGI that can do everything a three-year-old or four-year-old kid can do, right? Not necessarily the same robot body, but mentally, you can do everything a three- or four-year-old kid can do, right?
1:13:18 So, that’s a huge breakthrough, right? And that would convince everyone, wow, we have a real AGI, like this little avatar in this virtual world, it’s as smart as a three-year-old kid. It can talk to me, it can build stuff, it can move around and get stuff, right? So, that would be a huge breakthrough, bigger breakthrough than chat GPT by a significant amount from a science view anyway. On the other hand, once you got there, I mean, there’s still work to do, right?
1:13:48 So, what do you have to do to get that to be like a smart human adult? Well, maybe you need a bunch more processing power than you need to buy a bunch of NVIDIA GPUs or AMD or whatever, and then you have to solve some problems about how to scale up your computer network or maybe some of your reasoning algorithms actually don’t extend to what a 10-year-old does and you have to do some research to figure out how to beef up the reasoning algorithms in your artificial three-year-old, right? So, during that time period, the knowledge of how to make that artificial three-year-old will absolutely leak to everybody else, right?
1:14:27 I mean, whether or not you open-sourced it, this is not going to be like a one-man project, there’s going to be a lot of junior developers even if they don’t leak the code base, they’re going to be hired away by someone else and tell them how to do it, right? So, in the interval between that three-year-old AGI and getting to an adult level AGI, right, there’s going to be lots of time for the core ideas to leak out and for the thing to be replicated by different people who will then add on their own ideas, right? So, there is a certain point where it’s winner-take-all which is like, okay, whoever first gets to an AGI that is as smart as, you know, you or me and can then rewrite its own code base and improve its operations and design new hardware for itself, right?
1:15:20 Like, you could argue whoever gets to that level first, okay, that AGI system is going to revise and improve itself and know how to defend itself against others also, right? But on the other hand, it seems there can be a window of probably years at minimum months in between having the first really clear AGI breakthrough of which an artificial three-year-old would be one kind but there could be other kinds. There’s going to be some meaningful gap, not like five minutes, right, between getting the first real AGI breakthrough and then having a system that can self-improve and self-understand and rebuild itself to the level that, yeah, that guy is having a hard take-off and because they were the first mover they’re in the lead.
1:16:16 And this interval is when having a decentralized underpinning becomes quite interesting, right? Because if you have a real AGI breakthrough like your artificial three-year-old or say just a machine reasoning system that can solve the entire MIT graduate curriculum, right? I mean, there doesn’t have to be a three-year-old. There’s a lot of ways, right? Once you have that full-on breakthrough, then if that’s made within Tencent’s AI team, well, you know who’s controlling that and they’ll do a better job of keeping the code base secret than anybody else, too, right?
1:16:53 If that happens within Google, I mean, then, okay, there’s going to be a lot of sanctions on any human being who understands anything about that code base and it will rapidly be applied to the U.S.’s military and intelligence applications like going back to the beginning of this conversation. If this occurs within a global decentralized network where there’s humans in a hundred countries who are working on it and know about it and there is live code running on server farms all over the planet, right? I mean, then how things unfold in that period between the first AGI breakthrough and the first artificial superintelligence, right, how things unfold in that interval can be quite different depending on the success of the decentralized AGI initiative.
1:17:53 Brian Rose: Okay, I think I understand. Yeah, I’m always amazed when I talk to you, Ben, of how deep your knowledge is of not only what you do but what everyone’s doing and you seem to be always up to date every time I talk to you. You seem to know what everyone else is doing as well.
1:18:13 Ben Goertzel: Well, no one can know that but I’m trying my best to keep impressed and it’s an interesting time to do it because there’s just so much happening all around the globe.
1:18:25 Brian Rose: Do you feel overwhelmed?
1:18:27 Do you feel energized?
1:18:29 Ben Goertzel: I feel super excited. I spent decades planning for this time to come just like mapping out different algorithms, different implementation approaches, different deployment strategies. So I was thinking about this for decades because I knew this time was going to come and now it’s here and we can sort of unlock all this thinking that I and my colleagues have done over decades. So it’s being a heck of a lot of fun.
1:19:01 There’s just not enough hours in the day to do it all.
1:19:04 Brian Rose: Yeah, well, I can tell you guys are having fun and we’re excited to be a part of Zarka. Again, people can go to LondonReal.tv4.com to learn more and it’s great to see OpenCog and Hyperion and HyperCycle and all these other projects you’re launching. You have quite an ecosystem now and again, this open source decentralized narrative, blockchain narrative is something you guys have been all about ever since you started singularity net back in 17 and it’s turning out to be
1:19:32 A powerful, powerful way and maybe the way to develop this.
1:19:36 Ben Goertzel: More people are able to see the point of that now than in 2017. In 2017, it was more like, well, you’re combining different buzzwords in a very cool way. It must be something exciting. We got blockchain.
1:19:48 We got AI. We got robots. Whoa. Now more people are seeing. Well, wait a minute.
1:19:55 Hold on. AGI might be near and if someone doesn’t do something weird, these big companies and allied government agencies are going to control it. Is that really what we want? So I think more people are seeing the point of what we’ve been doing in gluing AGI together with decentralized infrastructure now than they did in 2017 when we launched singularity net.
1:20:23 Brian Rose: Absolutely.
1:20:24 Well, the more I learn about this space, the more I want to align with you guys and your ethos, your thoughts of decentralization, even today with the fact that you’re saying we can empower all these great minds and all these countries that would normally not work together. And I get more and more suspicious against these big tech companies because I’ve had rough experiences and I and you talk about certain ones aligned with certain powers and I just don’t want to trust them with the future.
1:20:53 Ben Goertzel: No, the problem is just as I found when I was working with military intelligence in Washington DC in an earlier part of my life.
1:21:01 I mean, the problem is these organizations literally have minds of their own, right? So the people working within the company, very few of them are evil people. The researchers they’re trying to do good research. They’re trying to make breakthroughs. They’re trying to build software systems that people will use and will make their lives better.
1:21:21 And the, you know, the business managers, they have a responsibility to deliver shareholder value to the people who put their hard-earned money into by shares of the company. Like people are just doing their job by and large, you know, in a good-hearted, thoughtful, ethical way. I mean, just like when I worked with INSCOM, everyone I met there was like, well, we’re defending the American way of life and, you know, we’re and there were terrorists trying to blow people up, right?
1:21:50 So the thing is the way that organizations are configured, they’re configured in a perverse way so that the good intentions of a lot of people combine together to yield often a bad outcome, right? And that’s the way corporations and governments have self-organized by nobody’s calculation. And we can counteract that by putting differently organized sorts of entities out there such as decentralized networks, including open source communities and including blockchain networks.
1:22:27 Brian Rose: Yeah, great observations. Ben, I always appreciate these check-ins. Let’s do another one in a couple months time. Final thoughts for people listening here, any final thoughts you have for everybody on the state of the union of AI.
1:22:47 Ben Goertzel: Oh, I think that from the standpoint of someone who’s not an AI expert, the most helpful thing you can do right now at this moment is just learn everything you can about AI, what it is and what it’s going to be.
1:23:05 Because if we’re going to have decentralized participatory decision-making about the future of AI, I mean, democracy of all forms works better when the voters know what the hell they’re talking about, right? And the better educated the general public is about what AI really is, what this or that AI system really can do, I mean, the better they’ll be able to participate in guiding the development. And the good news is this can be to your own benefit also, right? Because using current AI tools to do things in your own business or your own job is probably a good idea in many, many people’s cases.
1:23:45 But by doing that, you’re also learning both the strengths and weaknesses of current AI tools. I do think sometime next year we’re going to launch an initiative which is an AI agent running on OpenCog Hyperon and Zarka and TrueAGI Singularity Net Software, an AI agent whose purpose is both to answer people’s questions but also to learn from people. So then everyone will be able to go on and teach the AGI what they know, give the AGI the wisdom of their life experience as a human.
1:24:24 And I think I want to do that to both, because it’s a great way to fill the AI with human wisdom rather than having it just read Reddit and because it gives a way for people to become engaged and to exercise their desire to help out.
1:24:45 Brian Rose: I love it. The future’s bright. I love the way you view the future as well, Ben. It’s super refreshing.
1:24:54 So thank you for being optimistic. Thanks for your team. Janet Adams, your COO and future co-CEO of Zarka is amazing. And Sergey Shylappen and all your whole team’s amazing. So we’re really excited to be on this adventure with you with Zarka.
1:25:07 And just thanks for everything you’re doing. And let’s check back in in a couple of months and we’ll find out.
1:25:13 Ben Goertzel: Yeah, thanks. Thanks for having me. I always love our chats and you’ve gotten an amazing, amazing audience of open-minded people.
1:25:20 So I’m thankful to you for curating them.
1:25:22 Brian Rose: Fantastic. Thanks for the time, Ben. For anybody else, go to LondonReal.tv4 slash Zarka and get more information here. Ben, take care of yourself.
1:25:30 We’ll see you in a couple of months. All right. All right, bye. Take care, everybody. Bye.
