Ben Goertzel, Founder And CEO Of SingularityNET Discusses Decentralised AGI, Cybersecurity, And Humanoid Robots In Dinis Guarda YouTube Podcast

Ben Goertzel, founder and CEO of SingularityNET, shares insights on decentralised AI development, cybersecurity in generative models, inclusive AGI strategies, and advances in humanoid robotics, highlighting the MM2 compiler and the broader SingularityNET vision in the latest episode of the Dinis Guarda Podcast. The podcast is powered by Businessabc.net, Citiesabc.com, Wisdomia.ai, and Sportsabc.org.

Dinis Guarda Interviews Ben Goertzel, Founder And CEO Of SingularityNET

Ben Goertzel is a computer scientist, AI expert, author, speaker, and entrepreneur best known for introducing the term AGI (Artificial General Intelligence) in 2003 and making major contributions to its development. He is the founder and CEO of SingularityNET, a decentralised AI platform that uses blockchain to give open access to AI tools and services, enabling developers to earn through its native token, AGIX.

During the interview, Ben Goertzel discusses the importance of cybersecurity in generative AI:

“The way people deal with LLM security is to try to bolt on security after the LLM is done. If you have certain system prompts that you need to use over and over again, you would train a network for those system prompts and freeze that code so nobody could change it. Then that defends you against prompt injection attacks right away.

You need an intelligent sort of reasoning component that can try to detect when anyone is trying to overcome this wired-in, hard-coded system prompt you can use a sort of cybersecurity knowledge graph to do some symbolic reasoning about that.”

Decentralised superintelligence collaboration

Ben Goertzel discusses the potential of decentralised AI systems and the importance of inclusivity in the development of artificial general intelligence (AGI):

“A few big companies did not take over the internet right, but the fact that it hasn’t happened to a total degree is quite important. The internet is an open and decentralised protocol.

You can access research papers without paying Elsevier or $30 per paper which, while SciHub is illegal, if you’re a researcher in Liberia or something, how are you going to pay $30 to look at a research paper? You’re not going to do it. I mean, the takeover of the internet by a few big governments and big tech companies is not actually complete. You can still access knowledge and research through decentralised methods.

These indicate that the takeover of the internet by a few big governments and big tech companies is not actually complete. The power of decentralised protocols regarding AI. There is no moat. Big tech has a lead, but it’s not like the jet plane industry or something, where a small number of companies are geared up to build jets. Things are moving pretty fast in the AI space.

I think it’s entirely possible that a decentralised AI project can cobble together the data and compute power to make something smarter than anything big tech has done in the AI space.

We are the biggest AGI effort outside of big tech. The only serious AGI effort in the decentralised technology world. We’re exploring a broader variety of AI algorithms and cognitive algorithms.

The reality is a lot more open and heterogeneous than you would think from reading the leading tech blogs and newspapers. If you’re looking at the growth of a decentralised global brain, the best chance to get the first AGI to be beneficial to our species is if the AI has its value system and its knowledge base infused by a great variety and diversity of human perspectives.

If you look at it in a pro-diversity, pro-decentralisation way, like I am, then it’s good that there’s no moat. It’s good that things are more open than what you read in the papers.”

The Humanoid robots

As the interview continues, Ben Goertzel discusses his journey in humanoid robotics:

“Humanoid robotics is becoming popular post DeepSeek for mostly kind of shallow and commercial reasons, where people are like, well, if there’s no software, what can we invest in and defend ourselves against competitors? Well. Hardware is harder to replicate rapidly, so let’s invest in robots.

I led the software team behind the Sophia robot and Sophia’s little sister, Grace, who deals with medical robotics, and then Desdemona, who is the lead singer and poetess in Desdemona’s Dream, a rock band that I also play keyboards and sing in.

The main bottlenecks in humanoid robotics are actually software bottlenecks. We now have robots that can do every physical action a human can do, basically as well as a human. The issues are the cost of manufacturing of really capable hardware, and then the control software.

It’s great to have investment in promising advanced technology areas for any reason. I do think that, actually, the main bottlenecks in humanoid robotics are software bottlenecks anyway, not hardware bottlenecks.

We built this smaller robot, the Mindchild, to have something cheaper to manufacture and less likely to squash you if it falls down. It’s still a humanoid robot with an expressive face, different from Sophia and Desdemona, but can make a variety of facial expressions.

We’re developing them with a commercial perspective initially for the education space, helping kids in school and after-school programs. Then probably, after education, moving into the medical space, like the Grace robot.

Having a humanoid robot is good for building out the AGI knowledge base and value system. You need practical know-how of ethical values, just like raising human children. It’s not enough just to put knowledge about ethics into the AI’s knowledge base.

A humanoid robot allows shared meaningful activities with people in a physical environment, which is essential for AGI to learn human culture and values effectively.

We’re experimenting with interactive storytelling with the robot. A human kid and the robot act out stories together, helping the AI understand human interaction while being in the same physical space.

Having a human-like body is good for AGI development, but even within the limitations of the hardware, it allows the AI to learn how to act in the physical world with humans.

The speed with which we can do all this is quite remarkable compared to even a year ago. Big tech is moving fast, but I think we have clearer thinking in the decentralised world about how to really get to and beyond human-level AGI.

We do want to make products people can use and bring in revenue, but we are able to be more lateral and heterogeneous about it due to having a decentralised ecosystem. We can pull in more different parties around the world to help out in this super important quest.

I think we have less resources than big tech, but I think we have clearer thinking about how to really get to and beyond human-level AGI because of our thoughts not being clouded by the relentless pursuit of profit and corporate superiority.

We are able to be more lateral and heterogeneous about it due to having a bunch of different entities in an ecosystem working together, and it being open and decentralised, which allows us to pull in more and more different parties from around the world.”

Concluding the interview, Ben Goertzel discusses the launch of MM2 at the Hyperon AGI workshop:

“One of the things we launched called MM2, which is minimal meta 2, a new version of the compiler for our AGI language meta.

It’s around a million times faster than the very slow early research version we’ve been using before. It still needs some extra features to be able to deal with numerical data efficiently or text efficiently. Right now, it’s super fast at just mind graph processing. It’s not super fast at dealing with the outside world yet, but it will be within a couple of months.

We’re at a very interesting time for AI developers to jump in. You can start playing with our meta language and within a couple of months, the super fast version is going to be dealing with practical problems.

We can then take all these other AI algorithms that have been experimented with for decades and try them out at huge scale for the first time. It’s quite a fascinating time to be playing with all these new tools. There’s never been an easier time to jump into all this stuff.

There’s still a lot that still needs the human brain to do. You can make a web page or write a Python script by prompting, but you can’t yet build an AGI by prompting. There’s still this one big heroic task for humans to come together and cooperate in building, but now is the time to do it.”