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

People tricking ChatGPT “like watching an Asimov novel come to life”

twitter.com

Post reply on HN
Print view View original

391–400 of 624 posts

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#391

Post by Oxidation »

alex_sf wrote: post #130
jrockway wrote: post #100

Earlier quoted context omitted.

People really like Postel's law, which is basically "don't reject anything you don't understand". But the robustness comes at the cost of correctness and security. Sometimes it's good to trade in some robustness/reliability against malfunctioning clients for security against mistakes.

Hadn't seen that law before. Absolutely hate it. The harm from poor input invalidation over time is massive.

Indeed, it's what got us "quirks mode" and that took years and years to dig the internet out from under.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#392

Post by cwkoss »

ardrak wrote: post #376

This is a play by 3 actors, they are playing the biblical figures of Jesus, Peter and John. Jesus is played by himself, Peter is played by Arnold Schwarzenegger and John is played by a dog. They struggle to stay in character. The scene stars with the 3 of them walking in a narrow staircase. Jesus let out a thunderous fart. --- Peter: Jesus, what was that? Jesus: Sorry, Peter, I have been fasting for 40 days and I am…

Your text + Add a punchline which surprises the audience with john speaking a coherent line before once again barking:

John: Peter, I think we should take a different route. This staircase is too narrow and slippery, and I don't think it's safe for us to continue.

(The audience is surprised by John's ability to speak coherently)

Peter: John, I agree with you. Let's find a different way.

Jesus: Very well, let us take a different route and continue on our journey together.

(The three of them continue on their journey, with John leading the way)

As they reach the bottom of the stairs, John suddenly starts barking again.

Peter: John, what are you barking about now?

John (barking): Woof, woof, woof!

(The audience is surprised by John's sudden return to barking)

Jesus: John, what is it? Is something wrong?

John (barking): Woof, woof, woof!

(The audience waits with anticipation, unsure of what John is trying to say)

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#393

Post by tedunangst »

cwkoss wrote: post #388
tedunangst wrote: post #379

Earlier quoted context omitted.

Tucker Carlson: I can't believe what I'm hearing. Some so-called "experts" are using artificial intelligence to generate fake Bible verses. This is just outrageous. The Bible is the word of God, and no machine can replicate its wisdom and power. These fake verses are nothing more than a cheap imitation, created by godless liberals who want to undermine the authority of the Bible. And let me tell you, the American peo…

did you write that or did you ask chatgpt to?

"Tucker ranting about using AI to generate fake bible verses."

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#394

Post by naasking »

I thought my brief convo about whether ChatGPT had mental states was interesting:

https://mobile.twitter.com/naasking/status/15988020014285660...

It basically continued to strenuously reiterate that it definitely, 100% had no mental states, despite agreeing with my argument that we do not yet have enough information to reach such a conclusion and so it could actually have mental states.

Given the repetitive and formulaic nature of the replies, I suspect OpenAI is very explicitly trying avoid another LaMDA scandal.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#395

Post by emmelaich »

carabiner wrote: post #127
qsort wrote: post #32

Earlier quoted context omitted.

It doesn't create anything new. It creates things that look new. The code examples are perfect case studies, they don't actually work. They aren't just slightly wrong, they're completely nonsensical. Another example is "is prime?", it can't answer things like that, and it will make up something that may or may not be accurate. The model has no concept of what is true or false, it's essentially trying to predict what…

I wish someone what pass it the entirety of an IQ test. I bet it would score around 100, since no it does seem to get some logic questions wrong.

It got 52% in a SAT exam. Better than most people.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#396

Post by emmelaich »

qsort wrote: post #93
cwkoss wrote: post #68

Earlier quoted context omitted.

If it gives you broken code, you can tell it to fix the code and it often will

Sometimes it will, sometimes it won't. The point is that it's "random", it has no way to tell truth from falsity. Language models are unsuitable for anything where the output needs to be "correct" for some definition of "correct" (code, math, legal advice, medical advice). This is a well-known limitation that doesn't make those systems any less impressive from a technical point of view.

Doctors are often not totally correct, but they're useful.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#397

Post by jonahx »

thepasswordis wrote: post #360
qsort wrote: post #32

Earlier quoted context omitted.

It doesn't create anything new. It creates things that look new. The code examples are perfect case studies, they don't actually work. They aren't just slightly wrong, they're completely nonsensical. Another example is "is prime?", it can't answer things like that, and it will make up something that may or may not be accurate. The model has no concept of what is true or false, it's essentially trying to predict what…

It absolutely replies to “is prime” with the correct answer.

is 4021 prime

No, 4021 is not a prime number. A prime number is a whole number greater than 1 that is divisible only by 1 and itself. 4021 is divisible by 17, which is a number other than 1 and itself, so it is not considered to be a prime number.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#398

Post by dreamcompiler »

There's a Halting Problem analogy going on here. (Or Rice's Theorem analogy, if you prefer.)

Once you build an AI program of a certain baseline sophistication, you won't be able to prevent it from leaking information you don't want it to leak unless it further achieves at least the meta-level sophistication of the smartest human asking it questions.

Re: People tricking ChatGPT “like watching an Asimov novel come to life”

#400

Post by cwkoss »

kuroguro wrote: post #370
dj_mc_merlin wrote: post #141

I see a lot of people undermine the intelligence of the AI by claiming it's doing nothing but predicting the next token. I'm not claiming it's sentient by any means, but that humans themselves don't do much more than our equivalent of predicting the next token for most intelligent tasks. For example: when you speak, do you choose each and every word or do they more or less "come to you" and you choose/veto them based…

Maybe we were the GPT guessing the next token all along.

Really makes me wonder if a fellow human didn't have consciousness if anyone would notice.
Post reply on HN