An advanced AI system is already ‘self-aware’ – ASI Alliance founder

ASI Alliance founder Ben Goertzel says the alpha version of OpenCog Hyperon – the artificial intelligence system he's been building for more than two decades – is already somewhat “self-aware”.

Goertzel also told the magazine that OpenAI may refrain from making its “fantastic” new o1 model an autonomous agent for fear that it could be seen as “dangerous and risky” and prompt action against regulators.

The Artificial Superintelligence Alliance was founded in March, bringing together Goertzel's SingularityNET project, Ocean Protocol, and DeepMind veteran Humayun Sheikh FechaAI.

This week, 96% of CUDOS voters approved the decentralized cloud hardware network's merger with ASI. The integration boosts Goertzel's plans to grow OpenCog Hyperon, which he's been working on since 2001 and started as an open-source AI framework.

OpenCog Hyperon and the future of artificial general intelligence

Three years ago, the project began a total rebuild of OpenCog “in pursuit of scalability, and we're a long way into that process,” he said. The alpha launched in April, and while it says it's very slow at the moment and “breaking changes” are expected, the team is working to “speed it up a lot.” I think it should be finished this fall. And then, that means next year, we'll try to build into AGI on top of the new Hyperloop infrastructure.

Goertzel said the system takes a different approach to large language models (LLMs) like GPT-4 and o1.

Also Read: Creating ‘Good' AGI That Won't Kill Us All: Crypto's Artificial Superintelligence Alliance

“The Hyperon system is not just a chatbot. He is portrayed as an autonomous agent with his own goals and his own understanding and trying to figure out who he is and who he is. So it's not just an autonomous, self-aware agent, but a question-answering system.

AGI systems require autonomous self-awareness

Wait, is Goertzel saying that the current model is self-aware?

“What I mean by self-awareness, is that the process, even the current versions, has a model of who it is. It has a model of who you are. He has specific goals that he is trying to achieve in the situation. He knows how to relate to that situation and what to try to do. And the ChatGPT type system isn't really working properly.

The system combines a logic engine, evolutionary programming learning, and deep neural networks in a dynamic knowledge graph that self-revises and adapts. (Below explains what GPT-4 means).

Incidentally, Goertzel's lack of a world model explains why none of the autonomous AI agents developed so far have worked. You can't just take a “question-and-answer system” and tell them, “You're an agent interacting with the world.”

OpenAI's o1 and regulatory risk

For all the hype around OpenAI and “Strawberry” about autonomous agent-like behavior, Goertzel believes that OpenAI has deliberately avoided taking that path with the released o1 model.

“It's trying to be good at logic and logic, and it's great for that. I love it. It's amazing. It's not trying to be an autonomous agent. It's something that's deliberately different. I think they don't want to do that, because it looks risky and dangerous, and people are going to be trampled by regulators.” The last thing you want to do is create an independent agent.

A decentralized open source system is one of the benefits. Regulators can't “kick” it in the same way.

“When we launch advanced AI systems, they run in 50 or 100 different countries on machines spread across continents,” he says. “So I mean, if a country decides OpenCog Hyperon is illegal, it's only a small fraction of the network.”

Since our interview at Token2049 last week, OpenAI CEO Sam Altman has released an essay depicting the future utopia of AGI's “Age of Intelligence.” Goertzel hopes that the AGI he is trying to build will be so useful that no one will want to stop him.

“I believe this can be smarter than what Big Tech is doing. So if we can get something smarter than OpenAI, O1 model using Opencog Hyperon and release that on a decentralized network, I mean they'll jump into that world into ChatGPT and then they, by the way, use a decentralized network because basically It is in the infrastructure.

Challenges and benefits of decentralized AI systems

The tension at the heart of decentralized AI projects is that it is much easier and cheaper to run large models using centralized tools.

Training neural networks and transformers on decentralized devices is currently a non-starter, though Goertzel says new research suggests.

But he says that logical thinking and evolutionary learning through algorithms “run naturally on a decentralized network of machines.”

The plan seems to be to advance the network with some additional centralized resources as the decentralized network adds to the computing. SingularityNET and Fetch have released a “large fraction” of 100 million ASI tokens (currently renamed FET) to buy GPUs to build supercomputers. After token values ​​recover, they spend the rest.

“We want to have a decentralized network as a starting point to start this,” he said. “We want to provide hosting for people who want to put AI agents on SingularityNET.” The vision is a Web3-friendly cloud solution with a hugger face.

“AI developers are looking for the fastest, cheapest and easiest way to deliver a given task,” he explains. “If you want it to be decentralized, you have to somehow make it more attractive to end users, unless it's a decentralized philosophy.”

Among the four projects in the alliance, Goertzel estimates they will soon have “a couple hundred million dollars worth of dedicated computer hardware.” Additive Computing from SingularityNET spin-off NuNet, a platform that connects idle CPU and GPU power to connected computers and Hypercycle AI services.

Renewable energy to power decentralized AI systems

LayerZero's Bryan Pellegrino has worked in AI for over a decade and told a magazine at Korea Blockchain Week that his experience with Bitcoin mining profits at the price of electricity made him a big fan of decentralized AI.

“It is very difficult to calculate economies of scale. [needed]or the underlying hardware value [down]When you're competing with Google and AWS and all these others in the world and how to cool down and do everything, from the price of electricity to the bottom. So I've always been indifferent to most different parts of the media.

But Goertzel says the cost of electricity pales in comparison to the cost of the equipment, so “I disagree that it's a non-starter.”

However, he added, SingularityNet and HyperCycle are exploring opportunities to benefit from cheap renewable energy, given AI's huge power needs.

“We have talked to people in the Ethiopian government about putting more servers near the dam,” he said. “Tufi (Saliba, CEO) from Hypercycle is in discussions with the Paraguayan government about placing several AI data centers and server farms next to the dam on the Brazil-Paraguay border. It's one thing to get several gigawatts of power directly to an AI compute center by placing your computer center next to the dam with a fat cable.

Also Read: Ben Goertzel's Profile — How To Prevent AI From ‘Destroying Humanity' Using Blockchain Indeed To highlight the amount of electricity needed, OpenAI CEO Sam Altman revealed this week that the White House plans to build massive data centers in several US states, each requiring 5 gigawatts of power. .

AI data centers are usually located in major metropolitan areas as the services require low latency to provide extremely fast responses. However, Goertzel points out that the o1 model prioritized the quality of responses over speed, and so does Opencog Hyperon.

“If you have an AI that's trying to do any kind of deep thinking, let's say the AI ​​is trying to predict the direction of the market the next day, right? Or is the AI ​​trying to find a new drug, right? Then it doesn't matter whether it's in Paraguay or Ethiopia or anywhere.

Subscribe

A very engaging read in Blockchain. It is given once a week.

Andrew Fenton

Based in Melbourne, Andrew Fenton is a journalist and editor covering cryptocurrency and blockchain. He has worked as a film journalist for News Corp Australia, SA Wind and national entertainment writer for Melbourne Weekly.

Follow the author @andrewfenton