That's actually intriguing because inductive reasoning is what creativity (including scientific discovery) relies on. Deduction, by comparison, is less tricky and challenging. Just let an 80s-style production system in Prolog handle deduction and you'll be fine.
Yes, and we already have plenty of robust systems that can do deductive reasoning at a fraction of the cost of LLMs. It's important to know what are the strengths and weaknesses of each tool at your disposal.
That's actually intriguing because inductive reasoning is what creativity (including scientific discovery) relies on. Deduction, by comparison, is less tricky and challenging. Just let an 80s-style production system in Prolog handle deduction and you'll be fine.
Yes, and we already have plenty of robust systems that can do deductive reasoning at a fraction of the cost of LLMs. It's important to know what are the strengths and weaknesses of each tool at your disposal.
This makes sense given chatbots are trained. They need to recognize when they need to use tools rather than produce tokens.