Live data from Hacker News

CalcGPT

calcgpt.io

51–54 of 54 posts

Re: CalcGPT

#51
post #30

This is neat, but most people are going to miss "GPT-3 (babbage-002)". Using a rudimentary, outdated model seems disingenuous when making any kind of point about AI.

Yeah I would say it actually makes the contrary point. That pre hype version of the GPT is poor and if you have to use this one to prove a point it probably means there is a huge jump between GPT3 and GPT4. So to me it proves the contrary. And anybody going for that or believing it doesn't actually understand the performance of GPT4 or better if they are thinking that this is post hype LLM output.

Well, what if it just got better at covering up human-presentable cases?

See this comment [0] on this very post, showing how it makes quite problematic mistakes on larger numbers still.

It's still improvement, but only in the way of imitation. It shows that while clever within their constraints, these models still don't have the capabilities to truly perform computation or "thought". Chain of thought can help, but you there are some things you cannot split into atomic tasks; if the very world model isn't that stellar, no amount of elucidation will compensate for the inaccurate representations within. (i.e. "How would person X react to Y?" If your theory of mind is poor, no amount of further subtasks will help you give a better prediction.)

[0] https://news.ycombinator.com/item?id=41092987

Re: CalcGPT

#52
post #51

Earlier quoted context omitted.

Yeah I would say it actually makes the contrary point. That pre hype version of the GPT is poor and if you have to use this one to prove a point it probably means there is a huge jump between GPT3 and GPT4. So to me it proves the contrary. And anybody going for that or believing it doesn't actually understand the performance of GPT4 or better if they are thinking that this is post hype LLM output.

Well, what if it just got better at covering up human-presentable cases? See this comment [0] on this very post, showing how it makes quite problematic mistakes on larger numbers still. It's still improvement, but only in the way of imitation. It shows that while clever within their constraints, these models still don't have the capabilities to truly perform computation or "thought". Chain of thought can help, but yo…

For larger numbers it just needs to execute code. Most people also can't calculate such numbers in their head.

It shouldn't have to be able to do things it knows how to use code for. E.g. dumb thing slike how many Rs in a strawberry. It doesn't even see characters, so even if it was somehow possible, it couldn't count for sure.

It is like asking someone who only has ever seen hieroglyphs how many Rs are in a character by character version of strawberry.

Re: CalcGPT

#53
post #51

Earlier quoted context omitted.

Well, what if it just got better at covering up human-presentable cases? See this comment [0] on this very post, showing how it makes quite problematic mistakes on larger numbers still. It's still improvement, but only in the way of imitation. It shows that while clever within their constraints, these models still don't have the capabilities to truly perform computation or "thought". Chain of thought can help, but yo…

For larger numbers it just needs to execute code. Most people also can't calculate such numbers in their head. It shouldn't have to be able to do things it knows how to use code for. E.g. dumb thing slike how many Rs in a strawberry. It doesn't even see characters, so even if it was somehow possible, it couldn't count for sure. It is like asking someone who only has ever seen hieroglyphs how many Rs are in a characte…

Still, let's not anthropomorphize computational processes. It is a function approximate, which we'd expect to pick up on simple patterns like intersections or base10 arithmetic. When we see its predictions diverge from truth, that shouldn't be disregarded with a "just so" story, this is a sign we're pushing the architecture to its limits.

Re: CalcGPT

#54
post #14

Earlier quoted context omitted.

You can only get an 8 in the rightmost digit of the result by multiples of the rightmost digits, but 0 8 obviously gets you a 0, so fairly easy to see this is wrong. (10a+b) (10c+d) = 100ac+10(ad+bc)+bd

Well… I was joking. Even more generally, multiplication by b in base b gives a zero at the end.

… and, in fact, ‘b in base b’ always looks like 10 anyway!
Post reply on HN