News for hackers. Stuff that matters.
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by Uptrenda on Tuesday April 11, 2023 @09:25AM (#35524806)
"Prompt engineering" is the most pretentious name. Lad, your typing words into a little box. You ain't "engineering" shit, lmao. The next time I read a book maybe I'll tell people that I'm a phonetic diffusionary or how about making a burger at Mcdonalds: gastronomical entrainer. I hate bullshitters.
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by olliecornelia on Tuesday April 11, 2023 @09:37AM (#35524937)
I imagine this is how professional engineers feel about HN folk calling themselves software engineers.
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by rcme on Tuesday April 11, 2023 @10:28AM (#35525652)
I don't think railroad workers have much of an opinion on the title of software engineers.
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by bboylen on Tuesday April 11, 2023 @10:40AM (#35525825)
As a former chemical engineer, I think software engineering is way harder.

Understanding with complex man made abstractions is much more difficult than plugging data into a thermodynamics calculator

Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by akiselev on Tuesday April 11, 2023 @01:03PM (#35527936)
> professional engineers

When the GP says "professional engineer", I suspect they mean the big boy official kind [1] that goes to prison if they sign off on a negligent design. It's not a question of difficulty but responsibility and qualifications (though to be clear, the PE is considered more time consuming to prepare for than the BAR exam and it's definitely much harder than plugging numbers into formulas).

[1] https://www.nspe.org/resources/licensure/what-pe

Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by Uptrenda on Wednesday April 12, 2023 @01:16AM (#35535293)
I think it's pretty low stalking someones HN comments and using them to form a character assassination to dismiss someone's ideas.
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by capableweb on Tuesday April 11, 2023 @09:58AM (#35525235)
> "Prompt engineering" is the most pretentious name. Lad, your typing words into a little box. You ain't "engineering" shit, lmao

Ever called yourself "software engineer"?

Edit: https://i.imgur.com/V1kgSNt.png

Ah, even better, you call yourself a "blockchain engineer"... Is that any more "engineering" than "prompt engineering"?

Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by rcme on Tuesday April 11, 2023 @10:28AM (#35525658)
> Is that any more "engineering" than "prompt engineering"?

Yes?

Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by capableweb on Tuesday April 11, 2023 @10:47AM (#35525931)
According to what arguments? What makes writing letters to create/modify a blockchain/blockchain application different than writing letters for a LLM to generate the correct text?
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by rcme on Tuesday April 11, 2023 @11:15AM (#35526342)
Predictable outputs for a given input. Do you think a civil engineer just hopes for a sound structure? Or do they know the structure is sound before construction begins?
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by rtontic on Wednesday April 12, 2023 @08:45AM (#35538755)
Civil engineers actually have liability, they can go to prison. Software "engineers", not so much.
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by Uptrenda on Tuesday April 11, 2023 @11:16PM (#35534616)
Yes, I call myself a software engineer and a blockchain engineer because that's what I am. I don't have impostor syndrome and don't feel ashamed to use these titles. They are simply what I do.

- To me: software engineering is a structured, semi-rigorous profession, that seeks to use common methodologies to solve problems. It is technical in nature and can be approached abstractly or practically.

- I think 'prompt engineering' is misleading because it implies the user is doing something more complicated than they are. Since engineering requires technical knowledge to do and the key breakthrough of LLMs is a system that can respond to human-language, the technical-sounding title actually undermines the utility of what LLMs provide.

- Given the above: you wouldn't really want users to think of ChatGPT as a tool that only nerds can use. I wasn't just dunking on ChatGPT tinkers.

- I have 10 years of engineering experience where I specialise in arcane systems (p2p networking, smart contracts, asset security, trading, etc.) I have more experience than a junior dev but definitely not as much experience as some HNs. I do think crafting text templates is not 'engineering' and using ChatGPT is not meant to require such a skillset.

Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by spacebanana7 on Tuesday April 11, 2023 @10:00AM (#35525262)
> Lad, your typing words into a little box.

It’s more than just using ChatGPT.

High value prompts are used in API calls where performance, security and variable substitution all come into play. These queries can reach thousands of words/tokens.

A developer who spends all day writing prompts is at least as respectable as one who spends theirs writing SQL.

Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by qaq on Tuesday April 11, 2023 @10:02AM (#35525296)
gastronomical entrainer - don't sell em short Senior Principal Gastronomical Entrainer
Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by 0xDEF on Tuesday April 11, 2023 @10:17AM (#35525514)
I would have agreed before reading this:

https://lilianweng.github.io/posts/2023-03-15-prompt-enginee...

There is more depth and rigor to Prompt Engineering than what the AI snake oil merchants on twitter would have you believe.

Re:Maximizing the Potential of LLMs: A Guide to Prompt Engin... by ericmcer on Tuesday April 11, 2023 @11:03AM (#35526148)
Chatting with GPT is definitely not engineering.

Writing a prompt that runs thousands of times a day in a production setting, where extra words result in unnecessary spending? Closer to engineering.