Comment on Why Are LLMs So Gullible?parentComments−blackbear_2yIn principle you could attach a reasoning module to the neural network and only use the LLM part of the network for input/output.The design of such a reasoning module is an exercise left to the reader, obviously :p−daveguy2yYeah, LLM + Reasoning Module is only as good as its weakest part. And we had more than one AI winter around the development of reasoning modules. I think you're 2x correct -- we need both and the exercise to create it is worthy of :p
Comments
In principle you could attach a reasoning module to the neural network and only use the LLM part of the network for input/output.
The design of such a reasoning module is an exercise left to the reader, obviously :p
Yeah, LLM + Reasoning Module is only as good as its weakest part. And we had more than one AI winter around the development of reasoning modules. I think you're 2x correct -- we need both and the exercise to create it is worthy of :p