Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Ten advances in mathematics and theoretical computer science

openai.com

Post reply on HN
Print view View original

511–520 of 1001 posts

Re: Ten advances in mathematics and theoretical computer science

#511

Post by reducesuffering »

xpct wrote: post #382
kypro wrote: post #373

I want to iterate the most important thing about this is that it's yet more evidence of AI's accelerating competency in solving math and comp sci problems, and suggests we're now getting close to the point where you could throw AI at AI research challenges (which are largely just math and comp sci problems) and potentially find very real algorithm improvements. AI development is likely to be more compute bottlenecked…

Okay, let's take it seriously. What do you propose? What can your average person do to prepare for RSI beyond bracing themselves mentally?

You can not prepare or brace yourself mentally any more than you can a terminal cancer diagnosis. An RSI foom right now means an unaligned superintelligence will disregard us in pursuit of its goals. We would be ants in the way of a data center being constructed. All people can do is collectively support the notion, like 1200+ frontier AI researchers and their CEOs, that we do not have control of where this is headed, we need to immediately slow down the race, in time for people to agree that we do not have the capability to align a superintelligence to humanity’s wishes

Re: Ten advances in mathematics and theoretical computer science

#512

Post by ken47 »

titanix88 wrote: post #509

How do we know that these solutions don't exist in the training data? It is open secret that they have used pirated materials for training. Perhaps it plagiarized solutions from works of some obscure Belgian mathematician from the sixties, who did not get mainstream acceptance. I wouldn't be surprised if they also got access to mathematics done in the "defense contractor" setting from various three letter agencies. W…

This wouldn't be a problem so long as they properly attribute.

Re: Ten advances in mathematics and theoretical computer science

#513

Post by alightsoul »

Legend2440 wrote: post #504
fhfncjcc wrote: post #499

Earlier quoted context omitted.

The models are frequently getting worse at items that they aren’t being benchmarked for — and that’s happening more and more over time! Other people in other fields aren’t idiots, they are accurately perceiving the fact that these models are being hyper optimized for our industry, and are becoming less capable in other domains over time. Models of the same scale are massively worse at writing a broad variety of style…

Proof? In my experience modern models are better at all tasks than models from two years ago, especially complex multi-step tasks.

you are working on coding. they are working on things like "creative writing" remember that gpt 4o was popular among those who had ai as a romantic partnet?

Re: Ten advances in mathematics and theoretical computer science

#514

Post by pama »

sdenton4 wrote: post #497
WarmWash wrote: post #409

Earlier quoted context omitted.

>Whilst current models can't 'intuit That's how they are finding these solutions though, unless we are just going to label intuition as something only humans can do. Like a submarine being unable to swim or whatever that example is.

Some of them... The two places were seeing lots of movement are: * Updates to lower/upper bounds. In many cases, these kinds of problems are the deep-math equivalent of calculating more digits of pi. Yes, if you throw time at it you'll break the record, but it may not be terribly worthwhile. * Finding counter examples which disprove conjectures. This is really useful, and helps offset some positivity bias on the huma…

It is unfair to dismiss contributions to decades old open problems as equivalent to calculating more digits of pi. It missed the mark by a lot—as does the two bucket simplifaction.

Re: Ten advances in mathematics and theoretical computer science

#515

Post by Legend2440 »

alightsoul wrote: post #513
Legend2440 wrote: post #504

Earlier quoted context omitted.

Proof? In my experience modern models are better at all tasks than models from two years ago, especially complex multi-step tasks.

you are working on coding. they are working on things like "creative writing" remember that gpt 4o was popular among those who had ai as a romantic partnet?

Well that's on purpose lol. OpenAI does not want you falling in love with their chatbot and have been deliberately training it to be less romantic.

Re: Ten advances in mathematics and theoretical computer science

#516

Post by cubefox »

lkey wrote: post #451
deaton wrote: post #402

Earlier quoted context omitted.

If the thesis that "If anybody builds it, everyone dies" is true, or has any chance of becoming true, then it is the logical thing to do. As Geoffrey Hinton said, "If you want to know what it's like not to be the apex intelligence, ask a chicken."

It is not 'logical' to imagine a hypothetical doomsday scenario that justifies a preemptive nuclear war. (which is what the grandparent commenter's bio contemplates, 'nuke the datacenters' is their central credo). Does it bother you that the people who are publicly cocksure that P(doom) is moments away are the same people that have profited most handsomely from that pronouncement? That the 'humanists' that want to do…

> It is not 'logical' to imagine a hypothetical doomsday scenario that justifies a preemptive nuclear war

You hallucinated the "preemptive nuclear war". He didn't say anything about nukes. That's your own invention.

> Does it bother you that the people who are publicly cocksure that P(doom) is moments away

20% is not "cocksure". The "moments" is again an exaggeration.

Re: Ten advances in mathematics and theoretical computer science

#517

Post by rothos »

dash2 wrote: post #455
muchmirulys wrote: post #367

problem number 1 and 9 are surprisingly very intuitive check here : 1. high dimensional sphere packing https://muchmirul.github.io/conjectures/sphere-packing/ 2. multicolor ramsey number https://muchmirul.github.io/conjectures/multicolor-ramsey

The first link is very sloppy and doesn't actually explain why the "certificate" proves anything about the sphere packing. Or if it did, I couldn't understand it.

Agreed

Re: Ten advances in mathematics and theoretical computer science

#519

Post by emceestork »

gowld wrote: post #357
emceestork wrote: post #328

Earlier quoted context omitted.

They aren't claiming that science doesn't progress by moving goal posts. They're talking about how critics of AI have claimed it isn't revolutionary/useful, then progressively changed what would it mean for AI to be actually revolutionary/useful. Not long ago many folks were saying AI was the same as the crypto bubble. No real useful technology and only hype.

Did you know that "revolutationary" is not equivalent to "useful", and "revolutionary" is quite ambiguous?

I don't know if you're trying to dunk on me. I didn't intend to imply they are synonyms.

I think AI is clearly both revolutionary and useful. Revolutionary insofar as the job I do has changed almost completely in a year or so span.

Post reply on HN