Sure, llms don't do discreet math or logic directly. On the other hand, they write surprisingly good code. I'm guessing we'll see llms that do input>program(s)

Sure, llms don't do discreet math or logic directly. On the other hand, they write surprisingly good code. I'm guessing we'll see llms that do input>program(s)>run>summarize>output I'm not really disagreeing - I just think llms will do "more of the work" themselves by way of writing and running prolog programs, symbolic math (Julia etc) and running theorem provers.