News for hackers. Stuff that matters.
Re:GPT-Prompt-Engineer by fatso784 on Tuesday July 11, 2023 @07:45AM (#36679631)
This tool doesn’t benchmark based on how a model actually responds to the generated prompts. Instead, it trusts GPT4 to rank prompts simply in terms of how well it imagines they will perform head-to-head. Thus, there’s no way to tell if the chosen ‘best prompt’ actually is the best, because there’s no ground truth against actual responses.

Why is this so popular, then (more popular than promptfoo, which I think is a much better tool in the same vein)? AI devs seem enamored with the idea of LLMs evaluating LLMs —everything is ‘auto-‘ this and that. They’re in for a rude awakening. The truth is, there are no shortcuts to evaluating performance in real world applications.

Re:GPT-Prompt-Engineer by donkeyboy on Tuesday July 11, 2023 @07:48AM (#36679655)
There is a paper on arxiv saying that GPT4 correlation with human evaluators on a variety of tasks with strongly positive. I am also uncomfortable with it, but using GPT4 as a grader is not as bad as you think.
Re:GPT-Prompt-Engineer by fatso784 on Tuesday July 11, 2023 @07:54AM (#36679703)
You’re missing the point here. It’s not even getting the LLM’s opinion on evaluating the responses to the prompts (which itself is fraught for some tasks, and benchmarks are known to be limited —even OpenAI admits this, it’s why they made evals). It’s one level abstracted from that. It’s evaluating what the LLM thinks of how well the prompt will do, in purely hypothetical terms. That’s hogwash —different LLMs perform very differently even for the same prompts. Try any tool that lets you compare model responses side-by-side. Unless I see actual use cases, this is yet another iteration of overtrusting AI.

Here is what HN was talking about, nearly three months ago -the exact same type of ‘auto-prompt-gen’ tool: https://news.ycombinator.com/item?id=35660751

Re:GPT-Prompt-Engineer by duskwuff on Tuesday July 11, 2023 @03:08PM (#36685843)
> Here is what HN was talking about, nearly three months ago -the exact same type of ‘auto-prompt-gen’ tool.

I was reminded of the same thing. What a lot of it boils down to is that LLMs have no innate ability to self-reflect. They can pretend to do it, but no more effectively than an untrained human would.

Re:GPT-Prompt-Engineer by ChikkaChiChi on Tuesday July 11, 2023 @04:20PM (#36686705)
> They can pretend to do it, but no more effectively than an untrained human would.

Which is exactly as much as Generative AI should be trusted.

Re:GPT-Prompt-Engineer by jstanley on Tuesday July 11, 2023 @08:00AM (#36679746)
That's definitely begging the question.

If you're prepared to accept that GPT-4 can answer questions just as well as humans can, why do you even need to do prompt engineering?

Re:GPT-Prompt-Engineer by damascus on Tuesday July 11, 2023 @11:09AM (#36682102)
Humans still need 'prompt engineering' to answer questions more accurately though.

* What's the best way to get to Radio Shack from here?

is not the same as

* What's the easiest way to get to Radio Shack from memory when riding a bicycle from here?

Re:GPT-Prompt-Engineer by smogcutter on Tuesday July 11, 2023 @02:21PM (#36685151)
Easiest way to get to Radio Shack on a bicycle is to ride that bike down to Doc Brown’s house, charge the Delorean up to 1.21 gigawatts, and go back in time.
Re:GPT-Prompt-Engineer by QuantumGood on Tuesday July 11, 2023 @01:39PM (#36684462)
Humans benefit from good communication too. For example, annual U.S. deaths from medical errors is in the hundreds of thousands. Much of it is due to miscommunication. Is this akin to poor human-to-human prompt engineering? Of course, humans will rush and not attempt better communication, and you can take all the time you wish with an AI. And AI will continue to incorporate better prompt engineering that you won't have to write out. But there will always be a continuum from good to bad for communication, and communication outcomes.
Re:GPT-Prompt-Engineer by therein on Tuesday July 11, 2023 @02:27PM (#36685231)
You're forgetting what you may consider to be factual, self-evident and a priori is your opinion.

You may be under the impression that annual U.S. deaths from medical errors being in the hundreds of thousands miscommunicates but that is truly your opinion. You are merely jumping to conclusions at places another person might not.

And going on to rely on the LLM to validate your perspective is a lossy process. It may not lose your perspective but it loses someone else's and you don't even seem to notice or care.

Re:GPT-Prompt-Engineer by jstanley on Tuesday July 11, 2023 @02:37PM (#36685370)
This is an excellent example.

The post you replied to was saying that the deaths were caused by miscommunication, but you interpreted it to mean that stating the number of such deaths is somehow a miscommunication itself!

Re:GPT-Prompt-Engineer by therein on Tuesday July 11, 2023 @04:41PM (#36686935)
Doesn't help that I had just recently woken up but yes, most definitely.
Re:GPT-Prompt-Engineer by LASR on Tuesday July 11, 2023 @08:07AM (#36679825)
I think the parent poster is saying that it’s grading the prompts and not the output generated from the prompts.

Yeah I agree there. Unless you can check against the output, it’s not really telling much.

Re:GPT-Prompt-Engineer by DearAll on Tuesday July 11, 2023 @09:17AM (#36680543)
> There is a paper on arxiv saying that GPT4 correlation with human evaluators on a variety of tasks with strongly positive

Could you post the link please

Re:GPT-Prompt-Engineer by otikik on Tuesday July 11, 2023 @09:53AM (#36680995)
Of course there is strong correlation. That is literally what it was designed to do.

The problem is that it will simultaneously say that "cow eggs are bigger than chicken eggs", with the same confidence (and in a way that correlates well with human evaluators).

https://www.reddit.com/r/Funnymemes/comments/10ohd2n/chatgpt...

So when you get an evaluation you are playing the russian roulette - you may get a decent result, or you may get cow eggs.

Re:GPT-Prompt-Engineer by flangola7 on Tuesday July 11, 2023 @10:00AM (#36681089)
I just asked and it told me cows are mammals and do not lay eggs. That Reddit post is not even GPT-4 and is 5 months old, which may as well be the 19th century on AI tech timescales.
Re:GPT-Prompt-Engineer by albrewer on Tuesday July 11, 2023 @10:36AM (#36681583)
The post you're replying to is a case in point. This time it's cow eggs; what next?
Re:GPT-Prompt-Engineer by oefnak on Tuesday July 11, 2023 @12:07PM (#36683007)
Some people believe the earth is flat, but they can still provide useful work.
Re:GPT-Prompt-Engineer by amelius on Tuesday July 11, 2023 @02:33PM (#36685308)
These people typically subscribe to a very limited number of conspiracy theories.
Re:GPT-Prompt-Engineer by otikik on Thursday July 13, 2023 @01:19PM (#36712715)
You are concentrating on the details and avoiding the point.

The point is that the tool fails, and it is known to fail, so much that we even have a name for the times when it fails - hallucinations. I have been calling them cow eggs because that's a nice mental image and I didn't want to have to remember for the proper English term. I will continue calling them cow eggs.

Re:GPT-Prompt-Engineer by martincmartin on Tuesday July 11, 2023 @11:02AM (#36682018)
correlation ... strongly positive.

A positive correlation just means better than chance. "Strongly" is vague, and might not be much better than chance.

Re:GPT-Prompt-Engineer by dragonwriter on Tuesday July 11, 2023 @11:04AM (#36682044)
> A positive correlation just means better than chance. "Strongly" is vague, and might not be much better than chance.

No, adverbs like “strongly” modify adjectives (or verbs, but that’s not relevant here) not nouns; “strongly” is an intensifier that modifies “positive”, its not a separate adjective that modifies the noun “correlation”.

Re:GPT-Prompt-Engineer by nomel on Tuesday July 11, 2023 @04:33PM (#36686861)
If we could use GPT-4 to grade prompts, we wouldn't need to talking about grading prompts to use for GPT-4, since this solution requires that the problem doesn't exist.

You can't use the thing you're testing to evaluate its own performance. This applies to rulers, speedometers, and AI. It's the difference between a "subjective" and "objective" metrics. If you want an objective metric, you need to have it based on something external, based on reality. Otherwise, you have ideas holding themselves up: nonsense.

Re:GPT-Prompt-Engineer by toxicFork on Tuesday July 11, 2023 @08:00AM (#36679757)
Well, you could keep everything else in the project and put yourself or a human as the "does this result feel better than the other one" decision maker
Re:GPT-Prompt-Engineer by larodi on Tuesday July 11, 2023 @07:59AM (#36679741)
it seems to me also, that this is very much some sort of snake oil for the llm era. prompt generation varies from llm-to-llm and I doubt gpt4 can do reasonable evaluation, provided that it does not know at all about other models.
Re:GPT-Prompt-Engineer by londons_explore on Tuesday July 11, 2023 @08:03AM (#36679788)
The various leaderboards show that, in aggregate, LLM's acting as evaluators match other LLM's and humans remarkably closely.
Re:GPT-Prompt-Engineer by jasonlotito on Tuesday July 11, 2023 @09:13AM (#36680515)
> Why is this so popular

Grifters. I won't say that the person working on this is a grifter. Instead, it's so popular right now because of grifters. The same type of NFT grifters and crypto grifters who are mostly silent now. They've moved on.

How ethical would it be to sell things to these grifters, to sell the shovels they will use? But I'm always hung up on the idea that they will use those shovels on others and exploit them.

Re:GPT-Prompt-Engineer by immibis on Tuesday July 11, 2023 @10:27AM (#36681447)
Because this is the upside if not the peak of the hype bubble. All you have to do is use GPT for a task, and nobody cares whether it actually works, you still get whatever VC funding is left after the interest rate hikes.
Re:GPT-Prompt-Engineer by __loam on Tuesday July 11, 2023 @03:49PM (#36686366)
The market correction is gonna be gnarly.
Re:GPT-Prompt-Engineer by wilg on Tuesday July 11, 2023 @04:57PM (#36687117)
This seems to be some guy who made a MIT licensed project, so I'm not sure there's any reason to complain about VCs.
Re:GPT-Prompt-Engineer by typpo on Tuesday July 11, 2023 @12:09PM (#36683052)
Thanks for mentioning promptfoo. For anyone else who might prefer deterministic, programmatic evaluation of LLM outputs, I've been building this for evaluating prompts and models: https://github.com/typpo/promptfoo

Example asserts include basic string checks, regex, is-json, cosine similarity, etc. (and LLM self-eval is an option if you'd like).

Re:GPT-Prompt-Engineer by fatso784 on Tuesday July 11, 2023 @02:33PM (#36685317)
No problem! I guess I will make a plug myself --we've been working on a similar 'prompt engineering' tool, ChainForge (https://github.com/ianarawjo/ChainForge). It's targeted towards slightly different users and use cases than promptfoo --geared more towards early-stage, 'quick-and-dirty' explorations of differences between prompts and models for less experienced programmers, versus the kind of continuous benchmarking and verification testing power that promptfoo offers.

I particularly like promptfoo's support for CI, which I haven't seen anywhere else, and is very important for developers pushing prompts into production (esp since OpenAI keeps updating their models every few months...).

Re:GPT-Prompt-Engineer by mistymountains on Tuesday July 11, 2023 @07:43PM (#36688666)
Agreed. This is a pretty terrible idea.