Ben Goertzel, SingularityNET Foundation - TOKEN2049 Dubai 2024

hello everyone so it's uh yeah surprised to be uh here talking at token 2049 Dubai actually I hadn't been planning on speaking but I I happened to be happened to be in town and the folks from the conference contacted me that someone else had canceled because of of the flood so since I was here I figured why not come and talk uh for a few minutes about about decentralized AI which which uh is something I've been working on for quite a number of years but obviously is something far more on the tip of everyone's tongue now than than it was when I when I started out with it so want to talk for a couple minutes about where the AI field is now where it's going and why putting AI on blockchain is so so very important and uh I think it's important cuts across all different application areas I mean including humanoid robotics which is the the application that uh was just uh introduced to moment ago so the the humanoid robot Sophia who was the first robot citizen was my project I Le I led the software team behind behind sopia and we've made a number of other human robots since then that's one of the many projects which uh really needs decentralized AI so the the AI field as most of you probably know has been around since 1950s by that name really since the 1930s or 40s before that I mean I wrote my first AI code I think in 1979 or something when I was 13 14 years old at that time AI was much more obscure than now right it was a corner of the of the of the research world I started my first AI company in the late 90s and you know around 2201 I and a lot of others started thinking well hold on we've got I AI needs to run on many machines and the internet itself is a decentralized network right there's no Central owner or controller of the internet so why can't AI run decentralized with no single owner or or controller and you know in 2001 I wrote a book called creating internet intelligence I tried to chart out how you would make a decentralized AI network with little pieces of the AI running on all different machines around the world communicating by internet protocols tried to build something like that in like Java version 1.1 programming language it was just so slow I decided that was hopeless and ahead of its time now if I'd figured out to make decentralized money instead of decentralized AI then I would have been Satoshi which would be then we then we'd probably have the singularity and super intelligence already right but I didn't I we actually thought about decentral I money but we were like well this will just take a very long time for transactions to clear and we'll have a high cost so who would use it I didn't I didn't I didn't Envision uh Silk Road and all and the rampant price speculation and all the all the other things that drove the early days of Bitcoin right but so we fast forward to 2017 when I saw ethereum and the smart contract platform I was like this is the time that it can get real right right so I founded Singularity that in 2017 we did a token sale at that point and we launched what at the time was pretty much the only decentralized a AI Network now it's been an interesting Journey since then we built a lot of infrastructure for running AI on on decentralized networks and a number of other projects have popped up doing that as well we built some products that people use to do useful things def music processing but biology it's fair to say we haven't yet gotten the level of traction that we'd hop for when we launch a singular un n instead what's happened in the last few years centralized AI has taken over more and more of the of the internet right I mean starting with chat GPT mid journey and so forth I mean see more and more centralized AI Services running on big centralized server Farms created by big tech companies or VC funded startups these are doing amazing things right so you could look at this and say well oh big Tech 1 looks like building AI decentralized was just too hard in many ways it's harder than building it centralized which is hard enough already but but I I don't think think the story is yet told on the other hand and I I'm optimistic the next chapter of AI development will have a much more decentralized flavor now what has brought so much success to Big Tech in their recent AI Endeavors is a species of AI technology called Deep neural Nets right and some of these are convolutional neural Nets like used for computer vision or stable diffusion networks some are Transformer neural networks like underly large language models these are very valuable forms of AI technology but they're not good at everything deep neural nets for example are not good at complex multi-stage reasoning like you need to do to write a scientific paper or something they're also not good at being fun fundamentally creative like you need to do to make a break through new work of art or music unlike what was heard before so the the AI that big Tech has put out is great it doesn't do everything there's still a lot of opportunity left and what I've been aiming to do in Singularity net in the post chat gbt period is use our decentralized AI Network that lets different AIS running on different machines connect together to realize intelligent functions to connect together agents embodying different kinds of AI algorithms so take take a deep neural net that's good at generating text or images or analyzing data take another node on the decentralized singularity net Network running a logic engine that's good at doing formal scientific reasoning take another node on the singularity net network doing evolutionary learning which is a kind of AI That's good at inventing wild new stuff connect them together together on a decentralized network so that they can synergetically yield yield an intelligence where the whole is more than more than the sum of the parts of course all these things are easier said than done and big Tech is moving fast so to accelerate this further in the last couple weeks we've been working on what's I think the biggest toonomic merger in the crypto space so Singularity net fetch.ai and ocean protocol we're merging together our tokens agx ocean and fet into a single token ASI artificial super intelligence which will then have a combined market cap of of S billion or or something and the idea there is to join our communities join our underlying AI Technologies and our teams together in a a science to move faster toward decent decentralized intelligence which which which can you know fill in the gaps that what big Tech is doing can't do but do it on a decentralized platform right and the one thing we realized early in the singularity unet journey I mean we realized almost no one is going to use your AI because it's decentralized I mean they might think that's cool but pretty much they don't care people just want a product to do to to to do to do what it does so I think what we need to do for the decentralized AI world to win over big Tech big government I mean we need to launch the best AI in the world in ways that everyone can recognize when they're using it if it's a chatbot it needs to be smarter than chat GPT in meaningful ways like make less random up be better at reasoning make less boring and and derivative works of art right you roll out something that's smarter than what big Tech is doing but it just happens to be running on that decentralized Network then everyone will use the decentralized network just how most people who use stable diffusion didn't use it because it was open source I did cu I'm an open source guy right but most people use it because some kinds of pictures it generates better than than mid Journey does so the fact that it's open source fine so I think that's that's the challenge we have in the next few years we need to roll out AI That's better than anything big Tech is putting out in a way that Ordinary People can recognize when they use it the decentralized infrastructure needs to be good enough that it doesn't make it really slow really error prone or or or really expensive and then then you have a path to actually having decentralized AI take over the world in an economic sense hopefully not a military sense right so I mean imagine so open AI will launch GPT 5 and six meta belong to llama 3 llama 4 and so on so imagine we have something clearly more intelligent and creative than those things running on the decentralized network this would be really the first major crypto application outside of things that are crypto being used for crypto like like uh like Defi and and and and nfts it's hard which is why we're looking to merge together a number of projects to get some Synergy to to achieve it so the the merged entity we've created with the singularity net Fetch and ocean tokens we're calling ASI the artificial super intelligence Alliance and we may end up inviting other crypto projects in in in into the alliance as well to join forces because you know these big tech companies are multi-trillion dollar companies one big tech company who is now placing AI at the center of what they're doing has a bigger market cap than the entire entire crypto world right so I I mean we we need to join forces rather than than squabbling with each other and and I'm really happy we we've we've been able to to do that so how you know how I see things developing over the next few years I think we are probably not that far off from computers that think as well as people or even better than people so my friend Ray kwell who has been working for Google for the last decade or so but had a long AI entrepreneurship career before that Ray wrote a book in 2005 called The Singularity is near he forecast that AI would reach human level general intelligence in 2029 and then massively superhuman intelligence by 2045 and I think he was right roughly about the 2029 part and I think there's a a few pieces of evidence pointing there I mean one is all the evidence he cited in his book there's Mo's law there there's the the amount of memory that computers have is speeding up quantum computers are getting more and more capable networks are getting faster and faster so many different aspects of technological Advance are speeding up at an exponential or hyperexponential IAL rate The observed function of AI like chat GPT and mid journey and stable diffusion and so on llama that's another piece of evidence these things are just getting smarter and smarter we can all see it and feel it and the research that we're doing in singular net and in our other associate AI projects like the opencog project we're making more and more progress on getting reasoning and creativity working together with llm so we you can feel AI converging toward an amazing point in the next few years where I don't agree with Ray kerswell is I think once we get to human level AGI in 2029 or is it 2027 or 2030 it's all pretty close in the scheme of human history I think once we get there we're at most a couple years from having AI That's massively smarter than people because once you have an AI That's as smart as the people in this room that AI will take over AI development right that AI will take over design of AI Hardware that AI will Design new networks that AI will Design new blockchains and it will be able to copy itself over and over because computer programs can do that right so I think the path from Human level AI onto super intelligence will be a relatively rapid one which depending on your personality makeup you may find very frightening or or or very exciting or both right and and one thing I'm fairly confident of is the odds of this path from today's AI to AGI artificial general intelligence to Super intelligence the odds of this path coming out well I think are higher if the AGI we build is running on a decentralized infrastructure because what's the alternative the alternative is a big company working very very closely with military and intelligence agencies from whatever country it's sitting in having control of that of that AI right and that's not necessarily disaster but it's not I think what we really want so my my friend uh Alan comes who's a psychologist who was sort of a mentor to me gave a quote at a conference recently he said don't worry nothing is under control right and that I think uh we we we want to keep it that way and that's uh not a guarantee but it's it's it's the most likely condition under which beneficial superintelligence will will emerge so that's what we're working toward in Singularity net and the uh ASI Alliance so check out Singularity net. or superintelligence doai and uh find more information about the projects the software the the tokens and and and so forth also I wrote a book on this stuff recently it's called the the Consciousness explosion you can find that at the Consciousness explosion. all right thanks everyone