SingularityNET: Building Decentralized AGI

Janet Adams, COO at SingularityNET (© SingularityNET)

Launched in November 2017, Amsterdam-based SingularityNET was already competing in the Artificial General Intelligence (AGI) race before it became trendy. Silicon Luxembourg spoke to SingularityNET’s COO Janet Adams to learn more about the motivations behind building a decentralized AGI. 

What topic did you come to discuss at TNW Conference this year?

For us, TNW Conference was a great opportunity to come and spread the word about SingularityNET and inspire people towards a beautiful future and the vision we have of Artificial General Intelligence (AGI). There’s so much fear and scaremongering about AI out there and SingularityNET brings a different perspective, one where AGI takes away from the stress and suffering and where it can help solve the world’s issues.

Together with blockchain and AI, we have the opportunity to fundamentally change what it means to be human and to take better care of ourselves and the planet and emerge as more evolved human beings. That’s our goal.

What problem does SingularityNET aim to address?

By building a decentralized AGI running on blockchain, we hope to reduce the concentration of power that big tech currently holds over AI development and democratise access to technology and technology-generated wealth. By having a large decentralised community with democratic voting rights we also hope to curtail the ethical concerns about AI.

It is estimated that the AI revolution will be bigger and faster than the internet and mobile combined. The amount of wealth that will be generated will also be bigger, so this is a unique moment in time in which it will be important that it’s not just big tech that dictates all the rules.

How do you plan on beating big tech in developing the first AGI model?

There are two major AI races going on in the world today. The first one is the race of the decade which is about who’s going to build the first AGI model, which is vastly different compared to today’s very simple, more narrow AI models. This race is going to be won by the company that has the best tech and we think we’ve built that. We’ve built dynamic knowledge graphs, we’ve built a world model, a functional programming language called meta as well as an AI domain-specific language which allows AIs to interact with each other on our two-sided AI marketplace.  

The second AI race is about seeing who can build a genuinely intelligent ChatGPT killer, a large language model (LLM) which overcomes the fata and dramatic limitations of today’s LLMs. Today, these models are based on deep neural networks and they have no way of knowing whether they are right or wrong. Because its accuracy rate is not 100%, its potential use cases are still severely limited, particularly in low error tolerance cases. So this is going to be the race of the year, who can build the first neuro-symbolic smart LLM. 

How do you intend on balancing your mission of building an ethical AI model and winning the “AI race”? 

Firstly, our AI algorithms are ethical by design, which means that they are intended to serve positive purposes that benefit humanity. Secondly, our neuro-symbolic methods are explainable by design, so they can be programmed to have the right ethics and they have a world model against which they can check themselves. Lastly, on top of this model, we’re adding a human training layer operated on the blockchain which allows millions of people around the world to train it, bringing a wide variety of ethical views into the model. 

What tips do you have for entrepreneurs to stay ahead of the AI race without falling victim to empty hype?

Firstly, I recommend them to stay educated on the topic. Andrew NG’s course on Coursera, AI For Everyone is a great place to start. You don’t need to learn data science or programming for it, but you get a basic understanding of what the tools are and what the limitations of AI are.

Secondly, I would recommend putting in place a plan for structured data management at your company. Make sure all your data is gathered digitally and that it’s structured because this will provide the foundation for the implementation of any AI solution that you will use down the line.

Related Posts