Live data from Hacker News

So long “prompt engineering,” we hardly knew ya

medium.com

51–60 of 76 posts

Re: So long “prompt engineering,” we hardly knew ya

#51
post #4

i feel like people that get really into prompt engineering seem to forget that other people are working on building models that are even better . Do they really think that the tricks they developed to make gpt3.5 work well will still be necessary on gpt9 (or whatever new model replaces gpt)? if you want a long-lasting career in AI you need to work on the actual AI stuff, not just using the AI stuff.

I feel like people who write stuff like this don't understand the difference between software and software that delivers value.

I mean, you think those graphics optimizations we pour millions of dollars into before releasing AAA games will matter when the GTX 8020 outperforms a 4090?

Hint: Delivering value for actual people is rarely the result of sitting on your hands and waiting for the next big platform, or even rolling up your sleeves and trying to learn how to build the next big thing.

You have this idea that by investing energy in something that will be obsoleted you're losing out, but spoiler, that's how 99% of software that delivers actual value works.

The cutting edge of tech almost always ends up being PaaS/SaaS serving itself:

Your mail gets to you because someone is working on software with limitations we solved decades ago.

Your paycheck ends up in your bank account because people invest a ton of time in codebases subject to problems we solved long ago.

Your anti-lock brakes aren't built on a Rust codebase, but some horrible memory unsafe mess running on a processor that's a decade out of date.

The reality is: 99% of the effort that goes into trying to build the next big thing goes nowhere. The expected value of you trying to learn "the actual AI stuff" to the greater world is near 0 compared to you "just prompt engineering" and putting out something that solves a pain-point nicely with GPT 3.5.

At the end of the day most of the value that gets delivered to actual users comes from engineers who went deeper into extracting value from the current thing.

We still need people to work on the next big thing so that the 1% of effort that isn't wasted can actually materialize... but in my experience the most successful engineers in that regard are still able to realize the delusion it requires, without being paralyzed by the cognitive dissonance that realization invites.

Re: So long “prompt engineering,” we hardly knew ya

#52
post #18

Earlier quoted context omitted.

The culture war surrounding prompt engineering is so dumb. My eyes roll hard into the back of my head every time I read a comment like this. Prompt engineering is not the same as programming. In some regards, it is “better”. In other regards, it is “worse”. They’re two different yet similar disciplines, each with their own strengths and weaknesses, but both equally legitimate. Programming is simply the skill of being…

It’s more of a No True Scotsman thing about “prompt engineering is just inputting text, it’s not actual engineering” https://www.reddit.com/r/datascience/comments/14nbwfv/where_...

Anyone using a computer is "just" clicking on specific pixels and triggering keyboard events.. it still could be "actual engineering" imo if one is building a system following engineering methods, solving problems etc.

Re: So long “prompt engineering,” we hardly knew ya

#53
This article is garbage with a capital G

I'm going to link my own snarky titled by actually okay-ish take on prompt engineering, emphasis on the actual engineering, to show that the author doesn't even know what prompt engineering is - https://gist.github.com/Hellisotherpeople/45c619ee22aac6865c...

Prompt engineering is not designing a cool prompt. It's when you start applying genuine techniques to do things that are not possible with tokens alone. For example:

"What's the definition of {apple|orange}" where {apple|orange} is the mathematical average of those two words. This is prompt engineering. Right now, Prompt Engineering is basically in Stable Diffusion through Automatic1111, it's in libraries like microsoft Guidance or LMQL, and not a whole lot else.

Re: So long “prompt engineering,” we hardly knew ya

#54
This is a weird article not understanding what prompt engineering is. Basically with GPT-3.5 one can say bye bye to fine-tuning in most cases as natural language directives (prompting) combined with chain of thought can now beat most fine-tunings. That only works with extremely large models though. One can use prompt engineering for things like "the person I like talks this way: ... Please talk like this person". Or one can use prompt engineering to simulate one-shot NER and collect all the data appearing in the conversation without any super complex ML model pipelines. I fail to see how this goes away once GPT-274745 can write its own prompts.

Re: So long “prompt engineering,” we hardly knew ya

#55
I don't think prompt engineering is the problem. I find myself with minimal education but a lot of trial and error being able to "engineer" a good prompt eventually if I know the problem and can validate answers.

I think the big problem is actually finding problems you could use LLMs in. I think anyone that's played with them tends to have a good guess at whether they could or couldn't do something (although you need to really test to be sure, do some prompt engineering etc), but actually finding problems to work on that are within the realm of being solvable and useful is the hardest part imo.

Re: So long “prompt engineering,” we hardly knew ya

#56

So, the new thing is that we're going to call everyone an engineer? Carpenter -> Wood engineer Tailor -> Fabric engineer Ceramist -> Clay engineer See how dumb it looks? And maybe the field is too new to have a specific word for it, but calling it "engineer" diminishes the level of actual engineers.

This is just how language evolves, stop gatekeeping it. Most engineers these days have never been near an actual engine, and nor do they have any need to.

Re: So long “prompt engineering,” we hardly knew ya

#57
post #25

Am i the only one that tried using gpt4 for serious development work but stopped pretty fast because of its obvious limitations ? Reading that article makes me wonder if we're even talking about the same thing.

I'm using it for hobby development, which is more or less serious. Most useful cases I found:

1. API, concepts, etc docs with explanation. Much faster then googling then scrolling through tons of texts with blinking adds

2. Write simple things that I don't want to think about. Like in Python process all files in a directory, follow the links. Saves time.

3. Try things I don't know how to do. The recent was checking if user pressed a key without blocking in Python. Nontrivial, but possible. We went through several options till found the one which works on Ubuntu.

So, it's useful, no regrets about subscribing. Funny thing I'm using it working on toy GPT

Re: So long “prompt engineering,” we hardly knew ya

#58
post #18

Earlier quoted context omitted.

The culture war surrounding prompt engineering is so dumb. My eyes roll hard into the back of my head every time I read a comment like this. Prompt engineering is not the same as programming. In some regards, it is “better”. In other regards, it is “worse”. They’re two different yet similar disciplines, each with their own strengths and weaknesses, but both equally legitimate. Programming is simply the skill of being…

Working on new LLM model is like doing genetic engineering trying to come up with a new animal species. Prompt engineering is like feeding said animal and shoveling its manure.

Somewhere in this metaphor is a disdain for both prompt engineers and farmers.

Re: So long “prompt engineering,” we hardly knew ya

#59
post #18
post #4

i feel like people that get really into prompt engineering seem to forget that other people are working on building models that are even better . Do they really think that the tricks they developed to make gpt3.5 work well will still be necessary on gpt9 (or whatever new model replaces gpt)? if you want a long-lasting career in AI you need to work on the actual AI stuff, not just using the AI stuff.

The culture war surrounding prompt engineering is so dumb. My eyes roll hard into the back of my head every time I read a comment like this. Prompt engineering is not the same as programming. In some regards, it is “better”. In other regards, it is “worse”. They’re two different yet similar disciplines, each with their own strengths and weaknesses, but both equally legitimate. Programming is simply the skill of being…

Prompt engineering is a reality. No reasons denying. Better prompt is needed for better results.

But, even today prompts and rules of writing are not transferable even between existing LLMs. Future LLMs will have different architectures and requirements. I suspect today's prompt will be split into data and prompt. Probably just references, keywords to actual data.

In other words current state is transitory, next will be very different.

Re: So long “prompt engineering,” we hardly knew ya

#60
post #4

i feel like people that get really into prompt engineering seem to forget that other people are working on building models that are even better . Do they really think that the tricks they developed to make gpt3.5 work well will still be necessary on gpt9 (or whatever new model replaces gpt)? if you want a long-lasting career in AI you need to work on the actual AI stuff, not just using the AI stuff.

If you want a long lasting career in programming, work on building the languages, not just using the languages. Is probably what someone said many decades ago.

learning a programming language confers useful skills that you can generalize to new versions of that language or different languages. we can make these generalization becasue we know how the langauges work and people can draw connections from one to the other (i think people here would all agree that learning your second language is far easier than leraning your first).

this is different from LLMs because they don't have explicitly constructed abilities that we can compare across models. so we need to step back and approach every model as brand new and figure out what they're capable of. just because you have a prompt that works amazingly in one model, there's no guarantee that that prompt will continue to work in bigger "better" models.

so rather than working to devise way to trick a particular model into doing your task, it would probably be a better use of your time to learn how to train/modify models to explicitly solve the problem you care about.

Post reply on HN