Re: Neural networks and stuff (it is not about unboxed structures anymore)
"Tim Bradshaw (as tfb at tfeb dot org)" <[email protected]>
| Newsgroups | gmane.lisp.lispworks.general |
|---|---|
| Message-ID | <[email protected]> |
On 14 Jan 2026, at 11:36, Yuri Davidovsky (as work at disclosure dot ie) <[email protected]> wrote: > > I do not get this obsession with neural networks having to be a Turing machine, as in if they are not, then they are not useful (referring back to the famous argument that a single layer network cannot do XOR). Someone (not you I think, but I'm not following this in detail) made a claim which I read as 'neural networks cannot do x' (or equivalently 'NNs can only do y'. That claim (if it was made at all) is misleading because it is, in fact, the claim that 'no computer program can do x'. If we believe that physics is computable that in turn is the claim 'no physical system can do x'. Such a claim is interesting, especially if x is a thing that humans can do. --tim