How has AI changed your work?

How have LLMs changed what you do?

[graeme_p]

There was a comment in another thread about the lack of discussion of AI on WW. I think it must have changed something for almost everyone, so what has it been for you? For me, a few things stand out:

1. I write more tests.
   
2. Faster debugging and finding things in documentation
   
3. I have long given up on content sites relying on human traffic anyway, but AI is the nail in the coffin
   

[ronin]

How has AI changed your work?

I use auto-completion / auto-suggestion somewhat in both Cursor and Visual Studio Code. (Of the two code editors, I prefer VSCode.)

I ask both Perplexity and ChatGPT half a dozen or more questions a week regarding web technologies, web APIs, more recent baseline JavaScript methods etc.

I don't find LLM code-generation to be very useful or time-saving on the whole, but I have used LLMs for code review and that can be very useful.

[Mark_A]

I use Ai for all sorts of questions. There is no doubt that it can help.

But I need it to suggest my company when others are searching for relevant solutions.

I need to fully comprehend how to achieve that in a lasting way.

[Kendo]

I have used AI for all sorts of things and tried all sorts of AI for its wordliness and image generation.

I am currently working on a 20 piece solo exhibit that are mostly figure studies. Hiring models is tedious so I tried AI image generators. One that I tried was ideal for pose arrangement and minor adjustments for the next shot but it was incapable of producing the required texture and maintaining current texture when anything was changed. Reason being is that AI draws on a vast reference library and keeps trying to be the artist, based on something else that it has seen.

Another AI actually managed to produce the required texture and maintain it throughout all poses and changes, but getting a good pose was hit or miss - having arrived at something close, it was incapable of moving a limb slightly without changing pose completely.

The final designs I have to draw by hand.

One AI I found to be quite clever and a little bit cheeky. So I used it to reply to a decision made by a local department of justice - fancy a $200k/year lawyer pontificating about what a charity is/not when everything and everyone supporting that charity is working for free and without any reward whatsoever. The AI generated reply was blunt but extremely enlightening.

That same 'clever' AI was used to create titles and descriptions for my artworks, you know - the flowery justifications that artists use to describe their work these days, that usually have no relation to what they produce? Well they all took quite a few retakes, but my titles and descriptions are as real as the huge effort that went into each one.

I could not have worded those descriptions as well on my own.

[graeme_p]

@kendo, my experience with using for legal questions is that you need to know enough yourself to ask the right questions, and, especially ask for the right clarifications.

Not very different with everything else, really. its am amplifier of one's own abilities, or one's own incompetence.

[Kendo]

its am amplifier of one's own abilities, or one's own incompetence.

I agree. Management helps. Sometimes it is brilliant and comes up with stuff that I would not have thought of, and other times it took 20 prompts before we got something that ticked all the boxes. I refer to mine as my personal assistant and partner in crime.