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

GPT Takes the Bar Exam

github.com

Post reply on HN
Print view View original

131–140 of 147 posts

Re: GPT Takes the Bar Exam

#131

Post by happythebob »

ranting-moth wrote: post #91
andrewinardeer wrote: post #74

Earlier quoted context omitted.

The release of GPT-3 I believe will be looked back along the lines of a monumental advancement in line with the Trinity nuclear test. Legislation and policy has mainly kept nuclear weapons under check, however, nuclear technology does provide us with a reasonably clean energy which is beneficial for the masses. Perhaps regulation at some point will need to be crafted to ensure that only crippled AI or AI within a def…

With nuclear weapons, we can very much see and feel the power and destruction they can unleash. A proof is trivial, dig it in the ground and hold your ears when it goes off. With AI we have to debate how destructive it is. The debate itself has been weaponized. You'll be debating against deniers and AI using all the logical fallacies and information overload it can feed you. I, being the greatest armchair expert in h…

>I, being the greatest armchair expert in human behavior I know, say it will destroy democracy as we know it.

What's left of it after social media?

Re: GPT Takes the Bar Exam

#132

Post by heywhatupboys »

blitzar wrote: post #123
heywhatupboys wrote: post #104

Earlier quoted context omitted.

surely they are making the traders money though. The smart thing about the free markets is that companies are generally not pursuing things that are not of benefit to them

> companies essentially battle trading bots against each other Bot A is trading against Bot B. > surely they are making the traders money though It is not possible in a trading battle, exclusively between two parties for them to both make money.

Stock trading is not a zero sum game.

Even further the point: there are more participants in total than the botters

Re: GPT Takes the Bar Exam

#133

Post by ogre_battle »

iovrthoughtthis wrote: post #42
izacus wrote: post #32

Earlier quoted context omitted.

I don't get why you're overengineering this so much - with how intelligent AIs currently are, it's much easier to just input everything you did into ChatGPT and have it answer whether you should be executed or sent to a penal colony building iPhones directly. This way we can use the same approach for AIs we already use for banning your Google, iCloud accounts and approving your mortgage and insurance claims. Much eas…

haha, it's absolutism v republicanism wearing new flesh!

No flesh involved mon ami. That's the selling point!

Re: GPT Takes the Bar Exam

#134

Post by aflag »

spiderfarmer wrote: post #25
aflag wrote: post #18

Earlier quoted context omitted.

If it could build a case better than a human, you would. But currently it's nowhere close to that

I'd still trust a lawyer that used ten different laptops better than just one laptop.

Would you trust more an intern consulting 10 lawyers about your case instead of any single lawyer? Assuming a world where AI has surprassed human ability to make a case, having a human component would just make it worse.

Re: GPT Takes the Bar Exam

#135

Post by BizarroLand »

imtringued wrote: post #128
ranting-moth wrote: post #107

Earlier quoted context omitted.

https://www.kurzweilai.net/the-law-of-accelerating-returns

Where is my human brain capability for $1000 by 2023?

Looks like with the NVIDIA 4090 we're about 1/200th of the way there for $1600, but that's with a lot of commercial markup.

I think if you had the credentials you could probably rent a server farm or a cluster of 160 Tesla v100s that would meet that requirement, although how long you would have access to that for $1,000 may be very disappointing.

Re: GPT Takes the Bar Exam

#136

Post by ilaksh »

ThalesX wrote: post #78
ilaksh wrote: post #54

Earlier quoted context omitted.

AI can play chess, Go, Jeopardy, StarCraft, Minecraft, Atari games, write essays, summarize text, answer almost any question, drive a car, do above average on a SAT test, fool many people into wondering on Twitter if there were real people typing answers, write chapters of books, act as a dungeon master, write code to spec, win programming competitions, act as a therapist, fold proteins, provide medical diagnosis, vi…

> act as a dungeon master I've been trying to get it to act as a dungeon master unsuccesfully. Do you have any other leads apart from GPT (ChatGPT or AIDungeon)?

AI Dungeon is not too bad on the latest model. https://character.ai has some that are decent.

Re: GPT Takes the Bar Exam

#137

Post by candiodari »

lm28469 wrote: post #121
candiodari wrote: post #79

Earlier quoted context omitted.

You forget that a lot of these are moving the goal posts. AI beat the top players at all those games, but new rules were introduced until AI researchers mostly abandoned those efforts. For instance, the starcraft AI found a really good build order that let it quickly build "Stalkers", a versatile but relatively weak unit. I will note that that build order had several factors that humans found very surprising (consist…

> but we're pretty far along. Go out literally anywhere outside of a major western city and you'll witness just how incredibly wrong you are about this statement. > and Waymo's "chauffeur" is apparently much better than that You forgot a major detail, on clear straight and wide californian roads with perfect conditions. Put them in a historical european city or in inda's traffic hell and witness the mayhem Even waymo…

> I spent my holidays in a Eastern european country, I'd pay big money to film Musk in a fully autonomous tesla there, it would probably be the funniest shit ever, I bet it would crash in the first mile

I agree it'd be funny, but crash in the first mile? I'd take that bet.

Re: GPT Takes the Bar Exam

#138

Post by impossiblefork »

espadrine wrote: post #102
impossiblefork wrote: post #76

Earlier quoted context omitted.

10% is actually small though, also, the latest program tested is GPT3. GPT3.5 was what really impressed people and was the point when people saw that these models could be capable.

> also, the latest program tested is GPT3. GPT3.5 was what really impressed people The latest program shown in this chart[0] is text-davinci-003, which is part of GPT-3.5[1]. [0]: https://github.com/mjbommar/gpt-takes-the-bar-exam#progressi... [1]: https://beta.openai.com/docs/model-index-for-researchers/mod...

Ah, I it seems I can't read.

I still believe these models have potential for major improvements-- they'll probably still be unreasonable, but perhaps they'll be reasonable enough that people will have to make harder bar exams.

Re: GPT Takes the Bar Exam

#139

Post by blitzar »

heywhatupboys wrote: post #132
blitzar wrote: post #123

Earlier quoted context omitted.

> companies essentially battle trading bots against each other Bot A is trading against Bot B. > surely they are making the traders money though It is not possible in a trading battle, exclusively between two parties for them to both make money.

Stock trading is not a zero sum game. Even further the point: there are more participants in total than the botters

Stock trading is a negative sum game - there are costs for every trade.

> companies essentially battle trading bots against each other

If the bots are trading against each other, but making money from the other market participants then they are not trading against each other - they are trading against the other market participants.

Re: GPT Takes the Bar Exam

#140

Post by vasilipupkin »

bryankaplan wrote: post #118
vasilipupkin wrote: post #111

Earlier quoted context omitted.

I keep seeing comments along these lines. It doesn’t “make things up”. It outputs text under the objective function of maximizing probability of the token being outputted as “making sense” on some criteria. So, actually, it is biased towards not making things up. Occasionally, it hallucinates. I suspect we’ll see subsequent versions hallucinate less and less.

So what you're saying is it makes things up.

It occasionally does. But it’s biased against it. It doesn’t make things up as it’s default behavior.
Post reply on HN