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)