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

GPT-4 is not getting worse

coagulopath.com

Post reply on HN
Print view View original

161–170 of 186 posts

Re: GPT-4 is not getting worse

#161

Post by ilaksh »

thefurdrake wrote: post #147
jiggawatts wrote: post #67

The reason for these articles and the frustration people have is that GPT 4 is a black box and is being changed whether we like it or not. Nobody can choose to continue to use a specific snapshot with the alignment that suits them, because a third-party is enforcing their preferred alignment -- and not even specifying what that is! It's as-if a private company had invented portal technology and used it to establish t…

I haven't used OpenAI's API, but I might soon, and I've read tangentially that you can specify the checkpoint version in API calls to gpt3.5. Is this not the case with gpt4? Did they backport shitty alignment to older model versions?

Yes, you can specify the particular snapshot when using the API with gpt-4.

Re: GPT-4 is not getting worse

#162

Post by smcleod »

jw1224 wrote: post #35

There is an ongoing bug[1] with OpenAI's API, where it stops streaming responses after exactly 5 minutes. When I first came across the issue, I debugged it by writing a prompt along these lines: > Output every number from 1 to 10,000 in written form (e.g. "one", "two", etc.). Respond with one number per line in numeric order. As expected, the API would begin counting every number just as I asked. This would continue…

Yeah I’ve noticed similar, you can even explicitly say, do not truncate your output, or do not take any shortcuts. Make sure you haven’t missed anything ensure there are no placeholders etc…. And it ignores all of them.

Re: GPT-4 is not getting worse

#163

Post by nomel »

layer8 wrote: post #122
jw1224 wrote: post #35

There is an ongoing bug[1] with OpenAI's API, where it stops streaming responses after exactly 5 minutes. When I first came across the issue, I debugged it by writing a prompt along these lines: > Output every number from 1 to 10,000 in written form (e.g. "one", "two", etc.). Respond with one number per line in numeric order. As expected, the API would begin counting every number just as I asked. This would continue…

The inability to resolve issues like the mentioned 5-minute timeout with OpenAI support even after months should be a red flag to not build any products based on their services.

There’s no real alternatives for a product that would need their service to begin with.

Re: GPT-4 is not getting worse

#164

Post by ilaksh »

I think the answer to the problem of the OpenAI releases being out of our control and not necessarily consistent is to work on improving the open models.

I believe the big thing missing from open models is the advanced architecture and large amounts of human reinforcement. So it's actually not easy to replicate that with a volunteer effort. But I think the efforts of some great people working hard are gradually moving the open efforts forward.

Having said all that, it's funny how quickly people become entitled, demanding and critical towards this one company that provides a service with the ability to think for you and is smarter than any other such system in existence.

Re: GPT-4 is not getting worse

#165

Post by jiggawatts »

ilaksh wrote: post #161
thefurdrake wrote: post #147

Earlier quoted context omitted.

I haven't used OpenAI's API, but I might soon, and I've read tangentially that you can specify the checkpoint version in API calls to gpt3.5. Is this not the case with gpt4? Did they backport shitty alignment to older model versions?

Yes, you can specify the particular snapshot when using the API with gpt-4.

Yes, but they deprecate old checkpoints very fast.

You also can’t ever use the original version that wasn’t aligned to Sam Altman’s preferred vision.

By all accounts there is an “alignment tax” that reduces the capability of these systems.

For industrial or large scale public use I would also argue for the well-behaved models to be the only ones used.

For personal experimentation I prefer a sassy AI that tells me I’m wrong. Sam insists I’m too immature and I can’t be entrusted with making that choice for myself.

Re: GPT-4 is not getting worse

#166

Post by syndicatedjelly »

lairv wrote: post #87
rendall wrote: post #11

> I hate its craven, cringing, condescending, bitch-broken, mewling, groveling, wheedling, suck-ass tone. I hate its endless passive voice and weasel words. I hate its pompous It is important to remembers and it should be noteds.* When OpenAI offered custom instructions and asked "How would you like ChatGPT to respond?" I wrote: Be terse. Do not offer unprompted advice or clarifications. Remain neutral on all topics.…

The main pain point I still have with GPT4 is that it's hard to prevent it from being a yes-man "- I have problem A, what do you think of X ? - Yes X is great ... - But with X I'd encounter problem B, maybe I should go with Y ? - You're right X would cause problem B. With Y ..." One way to limit this issue is to not give ChatGPT any hint on your intent, instead of writing "I have problem A, what do you think of X ?",…

This gave me an idea to add the following to custom instructions:

* At the end of all answers, related to programming, provide a list of 4 keywords or topics that represent first principles for the subject so I can research them later.*

Re: GPT-4 is not getting worse

#167

Post by rendall »

syndicatedjelly wrote: post #166
lairv wrote: post #87

Earlier quoted context omitted.

The main pain point I still have with GPT4 is that it's hard to prevent it from being a yes-man "- I have problem A, what do you think of X ? - Yes X is great ... - But with X I'd encounter problem B, maybe I should go with Y ? - You're right X would cause problem B. With Y ..." One way to limit this issue is to not give ChatGPT any hint on your intent, instead of writing "I have problem A, what do you think of X ?",…

This gave me an idea to add the following to custom instructions: * At the end of all answers, related to programming, provide a list of 4 keywords or topics that represent first principles for the subject so I can research them later.*

How is this working?

Re: GPT-4 is not getting worse

#168

Post by pyinstallwoes »

agentgumshoe wrote: post #83
pyinstallwoes wrote: post #50

Earlier quoted context omitted.

It was still pompous and condescending to me. Glad other people are also perturbed. It was enough for me to cancel my subscription. Do not tell me how to think.

But yet you ask it to think for you?

No.

Re: GPT-4 is not getting worse

#169

Post by pyinstallwoes »

15457345234 wrote: post #56
pyinstallwoes wrote: post #49

Earlier quoted context omitted.

Aren't your sentences probabilistically based on the existing corpus of text within your mind?

No? If that's how sentence forming worked it would be impossible to count backwards, or count at all for that matter.

Why? Numbers have directionality relative to the species.

Re: GPT-4 is not getting worse

#170

Post by pyinstallwoes »

CamperBob2 wrote: post #135
joshuahedlund wrote: post #105

Earlier quoted context omitted.

Not completely. If you trained me on a thousand bad arguments for why the earth is flat and one really good debunking, I predict I would reason myself to the correct conclusion. I predict ChatGTP would not.

That seems like a compelling argument at first, but the churches are full of intelligent people who will fail the same test because of childhood indoctrination. They have been conditioned not to care about what's objectively true, while the LLM simply doesn't possess a concept of objective truth in the first place. I suspect it will be easier for ML researchers to fix that problem in their software models than it wou…

Objective truth doesn’t exist. Any system which defines truth can only point to objects within that systems language space.
Post reply on HN