Robert, Hurdle: The current LLM’s can only synthesize. They are effectively very

Robert,
Hurdle: The current LLM’s can only synthesize. They are effectively very sophisticated search engines.

Until these models can perform basic logic, they will run out of ‘legs’ (market opportunity) rather quickly as other than fancy copy past macros. All the impressive processing is still in tech where the vocabulary is small, grammar is consistent and the sources are numerous.

It knows mywork. It can imitate my style. If prompted it can enumerate a set of logical criteria I’ve specified – vaguel (and hallucinating the rest). But it cannot reason at all.

I’ve been in the business a long time. AI since the 80s. Been through the winters. We cannot train an LLM to perform basic logic.

We have three different problems requiring AI solutions. LLMS help with the ismple one. The other two require a more conventional response.

Reply addressees: @Scobleizer