Re:ProofOfThought: LLM-based reasoning using Z3 theorem proving
by a3w on Sunday October 05, 2025 @12:43PM
(#45483044)
I cannot use wolframalpha most of the time since the syntax is not that natural. WolframAlpha is good AI, it never lies.
Calculators are good AI, they rarely lie (due to floating arithmetics rounding).
And yes, Wikipedia says calculators are AI tech, since a Computer was once a person, and not it is a tool that shows the intelligent trait of doing math with numbers or even functions/variables/equations.
Querying a calculator or wolfram alpha like symbolic AI system with LLMs seems like the only use for LLMs except for text refactoring that should be feasible.
Thinking LLMs know anything on their own is a huge fallacy.
Calculators are good AI, they rarely lie (due to floating arithmetics rounding). And yes, Wikipedia says calculators are AI tech, since a Computer was once a person, and not it is a tool that shows the intelligent trait of doing math with numbers or even functions/variables/equations.
Querying a calculator or wolfram alpha like symbolic AI system with LLMs seems like the only use for LLMs except for text refactoring that should be feasible.
Thinking LLMs know anything on their own is a huge fallacy.