Creating ‘good’ AGI that won’t kill us all: Crypto’s Artificial Superintelligence Alliance

A year after dire warnings of humanity's imminent demise at the hands of super-intelligent artificial intelligence (AI), magazine is on hand to hear the other side of the story at the Beneficial AGI conference in Panama. Crypto Folk, Scientist authors, futurists and academics including David Brin.

We're at a conference led by SingularityNET, a key member of the new Artificial Superintelligence Alliance, which plans to explore what would happen if everything went right by creating Artificial General Intelligence (AGI) – human-level, artificial general intelligence.

But how do we bring about that future, not one where Skynet goes rogue and kills us all?

One of the best insights into why those questions are so important comes from the author of The Coming Death, José Luis Cordero, who believes that humanity will be cured of all disease and aging with AGI.

Arthur C., author of 2001: A Space Odyssey.

“He said, ‘We have to be positive about the future because the future – the possible – begins in our minds. If we think we're going to destroy ourselves, we probably will. But if we think we're going to survive, [that] We are going to a better world. [then we] We will work towards it and achieve it. So it starts with our mind.

Useful AGI: The Search for Artificial General Intelligence

People are hardwired to focus more on the existential threats from AGI than the benefits.

Evolutionarily speaking, it would be better for our species to worry nine times that the wind blowing through the bush might be a tiger, rather than being indifferent about rust and getting eaten by a tiger on the 10th occurrence.

In a study of nearly 3,000 AI researchers, the worst-case scenario was 5% to 10%. So while that's a big concern, the odds are still in our favor.

Also read: Ben Goertzel's profile — How to prevent AI from ‘dehumanizing' using blockchain

Opening the conference, Dr. Ben Goertzel, founder of SingularityNET and father of “AGI”, paid tribute to Ethereum founder Vitalik Buterin's concept of defensive acceleration. That's the midpoint between the effective speed techno-optimists and the “move fast and break things” ethic and the decelerationists, which slow or stop AI development.

Goertzel believes a slowdown is unlikely, but admits there's a small chance things could go horribly wrong at AGI. Therefore, it supports monitoring AGI while keeping in mind potential risks. Like many in the AI/crypto field, he believes the solution lies in keeping the technology open and decentralizing hardware and management.

What is the Artificial Superintelligence Alliance?

This week, SingularityNET announced the creation of the Artificial Superintelligence Alliance (ASI), in partnership with decentralized multi-agent platform FetchAI – developed by DeepMind veteran Humayun Shaikh – and data exchange platform Ocean Protocol.

It will be the largest open source independent player in AI research and development and proposes to merge the existing tokens of SingularityNET, FetchAI and Ocean Protocol into a new one called ASI. It will have a market value of around $7.5 billion – a vote of approval is expected in the next two weeks. The three forums will continue to operate as separate entities under the leadership of Sheikh Goertzel, chairman.

According to the Alliance, the goal is to “create a powerful and compelling alternative to Big Tech's control over AI development, use, and monetization” by creating a decentralized AI infrastructure at scale and accelerating investments into blockchain-based AGI.

What are the benefits of AGI?

Perhaps the most obvious important impact is the ability of AGI to analyze massive data to help solve our most difficult scientific, environmental, social, and medical problems.

MIT researchers have seen some amazing medical breakthroughs by using AI models to evaluate tens of thousands of potential chemical compounds and have discovered the first new class of antibiotics in 60 years that is effective against the so-far drug-resistant MRSA bacteria. It is the kind of study improvement that human beings cannot achieve.

And that's all, before we get to the immortality and mind-bending stuff that transhumanists are so excited about, but most people wonder about.

This ability to analyze vast amounts of data suggests the technology could provide early warning of epidemics, natural disasters and environmental issues. AI and AGI have the potential to free humans from drudgery and repetitive work, from coding to customer service help desks.

While this creates a huge upheaval in the workforce, the invention of washing machines and Amazon's online businesses have had a major impact on the professions. The hope is that many new jobs will be created instead.

Economics professor Robin Hanson says this has happened over the past two decades, even though people were more concerned about automation replacing workers at the turn of the century.

Hansen's data on how automation has affected wages and employment across industries between 1999 and 2019 shows that despite major changes, most people still have jobs and are paid the same.

Between 1999 and 2018, he says, “automation has had no effect on average wages or U.S. jobs.”

AGI can make better decisions than we can: Janet Adams, CEO of SingularityNET

Janet Adams, CEO of SingularityNET Optimism, described AGI as having the potential to be “extraordinarily positive for all of humanity.”

“I see a time when our future AGIs will make more ethical decisions than humans do. And they'll be able to do that because they don't have emotions or jealousy or greed or hidden agendas,” she says.

Adams suggested that 25,000 people starved to death every day. She says it's a problem on the planet that can be solved by “smarter distribution of resources on Earth.”

But Adams cautioned that AGI must be trained on data sets that reflect the entire population of the world, and not just the top 1%. For the benefit of the wider civilization, the wider humanity.

AI Security and Ethics: Addressing the Concerns

Anyone who has looked at the utopian dreams of the past as a decentralized internet collapses into a corporate ad-laden landscape of addictive design and engagement farming can doubt the existence of this exciting future.

Building high-level AI requires computing and other resources that are currently unattainable from all but a few of the usual suspects: Nvidia, Google, Meta and Microsoft. So the default assumption is that one of these tech giants will end up controlling AGI.

Goertzel, a long-haired hippie who plays in a surprisingly good band fronted by a robot, wants to challenge that assumption.

Goertzel points out that there was already a default assumption that companies like IBM would dominate the computing industry and Yahoo would dominate search.

“The reason these things change is because people are actually fighting to change them in each case,” he said. Instead, Bill Gates, Steve Jobs and the Googlers came along.

The founder of SingularityNET, has been thinking about the Singularity (the theorized period of exponential technological progress) since the early 1970s when he read an early book on the subject called The Prometheus Project.

Since then, he has been working on AGI for a long time, popularizing the term AGI and launching the OpenCog AI framework in 2008.

Adams Goertzel says it's a key factor in SingularityNET's credibility.

“We are the largest non-profit crypto-funded AI science and research group on the planet,” says Adams, whose competitors have focused on “narrow AIs” like ChatGPT and are now shifting their strategy to AGI.

“We have years left,” she says. “With Dr. Ben Goertzel, we have three decades of research on neuroimaging methods.”

But opening up the platform to developers from around the world and rewarding them for their contributions, she says, would give a leg up on the mega-corporations that currently dominate the space.

“Because we have such a strong vision and strong commitment to democratize building the most advanced, most intelligent AGI, it's hard to imagine Big Tech or any other player coming in and competing, especially when you're open. source”

“[We will] Check out the massive influx of people on the SingularityNET Marketplace and the speed it's leading to AGI. It's more likely to be us.”

Decentralized AI: Opening the door to open source development

The Prometheus Project proposes that AI is such an earth-shattering development that everyone in the world should have a democratic voice in its development.

Therefore, when blockchain emerged, implementing a decentralized infrastructure and token-based management of AI was the next most practical option.

Tufi Saliba, CEO of Hypercycle, told the magazine that this would reduce the risk of a centralized company or sovereign nation developing AGI first and gaining superpowers, which would be “the worst thing to ever happen to humanity.”

Also read: Real AI use cases in crypto, No. 1: The best financial use for AI is crypto.

This is not the only solution to the problem. Meta's chief AI scientist Yan Le Cun is a big supporter of open source AI models and a thousand flowers to bloom, X owner Elon Musk recently opened the model for Grok.

But blockchain is arguably a big step forward. SingularityNET aims to network the technology around the world using different entities controlled by different communities, thus spreading the risk of any single company, group or government controlling AGI.

“So you can use these infrastructures to implement decentralized deep neural networks, to implement massive logic engines, to implement simulated ecosystems to implement artificial life, and to create lots of small artificial animals interacting. It's trying to move to intelligence,” Goertzel explained.

“I want to encourage creative contributions from everywhere, and as you probably know, a 12-year-old genius from Tajikistan comes up with a new artificial life invention that will make a big difference to AGI.”

What is a hypercycle? What is OpenCog Hyperon?

Hypercycle is a fast ledgerless blockchain that allows AI components to communicate, coordinate, and end-to-end in less than 300 milliseconds. The idea is to give AIs a way to summon resources from other AIs, paid for via microtransactions.

For now, the startup network is being used for small-scale applications, like an AI app calling another AI service to complete a task. But eventually, as the network scales, in theory AGI could be the emergent property of different parts of AI working together in one distributed brain.

“So, with that approach, the whole world is more likely to move into AGI as one entity,” says Saliba.

Goertzel didn't develop Hypercycle for that reason—he wanted something faster than existing blockchains to enable AIs to work with.

The project he's most excited about is OpenCog Hyperon, which launches in alpha this month. “It combines deep neural networks, logic engines, evolutionary learning and other AI paradigms in the same software framework to update a highly decentralized knowledge graph.

The idea is to improve the METTA AGI programming language and keep it open to anyone who wants to work on it. “We will have a complete set of tools to build the baby AGI,” he says. “To get something I'd like to call Baby AGI, we need to speed up the METTA interpreter a million times,” he says.

“My best guess is that Opencog can be the system to make Hyperon [AGI] success”

AI management: the role of voting systems

Of course, decentralization doesn't guarantee that things will go smoothly with AGI. According to Goertzel, in the 1990s Somalia's government was decentralized under many warlords and militants, but at the time preferred to live under the centralized Finnish government.

In addition, token-based management is a long way from being suitable for prime time. In projects like Uniswap and Maker, big owners like a16z and the core team have so many tokens that it doesn't matter if anyone votes. Many other decentralized autonomous organizations are mired in politics and infighting.

The rise in value of crypto/AI projects has attracted a lot of speculation. Are these the people we want to control at AGI?

Goertzel argues that while blockchain projects are currently primarily of interest to people looking to make money, the use case will change as it evolves.

“If we deploy the world's smartest AI on decentralized networks, you'll find a lot of other people who aren't primarily thinking about financial speculation. And then it's going to be a different culture.

But if the Artificial Superintelligence Alliance achieves AGI, won't the tokens be ridiculously expensive and largely inaccessible to those interested in useful AGI?

Goetzel suggests that a weighted voting system that prioritizes those who contributed to the project might be needed:

I think we need to have a sophisticated, decentralized reputation system to guide the brains of AGI, and have something closer to one person, one voice, but people with some experience in contributing to and creating an AI network. In some sense, get a higher weight.

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