4 ms·
There's not much reasoning going on in those samples. It can stick a subject input you provide it into a sentence which is sometimes serendiptously funny (and
by notahacker 7y ago
There's not much reasoning going on in those samples. It can stick a subject input you provide it into a sentence which is sometimes serendiptously funny (and usually grammatically correct tbf, although it looks like that's mostly because it heavily reuses direct quotes from its source material), but it can't handle the basics of game construction like responding properly to a request to go north after telling you there's a path to the north
- malux85 7y agoThat’s why I said “the beginning” the first step is forming a representation of thought, so things like the embedding matrix inside word2vec form “thought vectors” - self organised representations of high level conceptual ideas, generalised to all knowledge. And it turns out these vectors work under linear algebra (e.g. They are commutative) I think these will form the basis of machine reasoning in future
- bluntfang 7y ago>There's not much reasoning going on in those samples. but you agree there is some? that's the significance.
- cmroanirgo 7y agoMy favorite pastime playing MUDs in the early 90s was to do dumb stuff. > Eat horse > YOU CAN'T DO THAT > Eat horse > YOU FEEL SICK What I read in this article doesn't show much I (of AI) to me, but highlights how 30yrs ago there was a lot of creative thinking to generate an open seeming environment.