Question 1

What is the neuro-symbolic approach to AGI behind OpenCog Hyperon?

Accepted Answer

_Goertzel's own bet replaces the neural net as the core data structure with a self-modifying knowledge graph that fuses neural, logical, and evolutionary learning._

> The basic data structure isn't a neural net, it's a huge dynamic self-modifying knowledge graph living in RAM across multiple machines. You can put neural nets in there, you can put logical propositions in there and do theorem proving, and you can put program code in there and evolve it with genetic programming that simulates evolution by natural selection. Then you wrap a goal-driven cognitive architecture around it.

Source: [Rethinking the AGI Race, with Benjamin Goertzel (eCornell Keynote)](https://ecornell.cornell.edu/keynotes/view/K052826/)

Question 2

Where do neuro-symbolic AI systems beat large language models today?

Accepted Answer

_The clearest near-term edge for non-LLM AI is messy data that is part numbers, part narrative._

> When you have data that is a mix of quantitative and qualitative in a confusing, tangled-up way, LLMs are not great at sifting through the structured part at scale. Medical data is like this: a lot of qualitative information plus hard clinical numbers. Many areas of finance are like this. I've worked on catastrophic risk estimation, like how to price the risk of a revolution in a developing country. That's where symbolic reasoning plus creative leaps, with an LLM there to crunch the text, has a big advantage.

Source: [Rethinking the AGI Race, with Benjamin Goertzel (eCornell Keynote)](https://ecornell.cornell.edu/keynotes/view/K052826/)

Q&A, neuro-symbolic

What is the neuro-symbolic approach to AGI behind OpenCog Hyperon?

Where do neuro-symbolic AI systems beat large language models today?