Ask Slashdot: Do You Use AI - and Is It Actually Helpful? 248

I use ChatGPT daily, not for work, but to ask questions about topics I'm curious about that I'm sure experts don't have time for.

I also discuss topics that I find overly hyped, skewed, oversimplified, tribal, or full of agendas or propaganda in modern discourse -- and that's pretty much everything you read about online. For instance, I recently discussed the film Anora, which won five Oscars. It was horrible -- a poor B-movie at best. I have no idea what happened to the Academy, and ChatGPT agreed with me.

I use tools like Lovable all day long now, 100% of the code I produce is generated. I do not even type anymore I use speech to text. I read the commits and suggest refactorings. The future of development is no more human teams just one senior dev who leads a team of sotware agents.

Sometimes I use AI to write quick functions with bounds checking, it does a decent job, but it always needs tweaking.

AI shines for debugging, when I have something that "should" work, I copy/paste the errors or problem description into AI and ask "How to Fix ELI5"

Once, AI even suggested there was a bug in a library, and it was right.

Note: I pay for Google Gemini 2.5 Pro and find it the best for coding compared to Chat-GPT.

Gemini does a good job of breaking everything down.

Here is an example from a mysql ses

I once asked chatgpt to write a sonnet about constipation. I remember being abused by the result.

Mostly, though, I try to avoid the brain atrophe device.

Let's start with some disclaimer: this is about LLM. Not AI, as it is a very large field of stuff that existed way before LLM became the latest craze, and hopefully will keep existing until we get something impressive out of it.

Also, some issues with LLM are not stemming from their output, but with economic models around them, privacy issues, licensing issues, etc. To address some of those, most of our daily stuff is done on locally running models on cheap hardware, so no 400B parameters stuff.

There's four

The one thing I found it useful for is searching something when I di not know the specific term. Then I can ask and do a real search afterwards with the term and get context, references, limitations. etc. For anything else, AI is a waste of time. The one time I found something I had searched conventionally without results, the AI was unable to provide an original reference, i.e. the result was worthless.

Lets face it, this AI hype is a dud. Much the same as all previous ones. Something will come out of it,

I use LLMs - primarily ChatGPT - in programming all the time.

It's a tool. I don't care if it "really" understands anything, that's not what I use tools for.

Yes, it really saves me time overall. I can't help it if it somehow doesn't suit your problem space (possible), or if you don't know how to or refuse to learn how to use it properly (likely).

Even if we had full-on human level AGI (which we don't), you'd still need to iterate and correct.

You wouldn't expect to give a non-trivial programming task to another human without having to iterate on understanding of requirements and corner cases, making changes after code review, bugs needing to be caught by unit/system tests, etc.

If you hired a graphic artist to create you a graphic to illustrate an article, or book, or some advertising campaign, then you also wouldn't expect to get it right first time.

What I find current LLMs the most useful at is to review my work to give a different and useful perspective.

The LLM doesn't generate any work that goes into my output, it augments my own work to make it better, while I determine what to use (I am in the drivers seat).

The LLM has been trained in a huge quantity of good quality human-generated text and it surprises me how good it can be at offering associations that I had not considered.

My #1 use for ChatGPT is "show me an example of some C code that implements functionality (X)".

Then I can read that example, research the APIs it is calling (to make sure they actually exist and are appropriate for what I'm trying to accomplish), and use it to write my own function that does something similar. This is often much faster than my previous approach (googling, asking for advice on StackOverflow, trial and error).

I use LLMs for my own amusement, they are useful for that.

I have little to no visual memory so I struggle to draw even simple things. I can do drafting-style sketches OK because they are logical, but just remembering the shape of a curve even between looking at the thing and looking at the paper is difficult. So I use AI to generate images and feel zero remorse about it, since it lets me do something I cannot otherwise do — envision a concept I can imagine, but cannot picture.

For answering questions,

I use it much like a community to get help with programming, asking questions such as "what does this statement do?" or "what are ways to do X?" or "What is wrong with this code ...?" or " Can I accomplish X by doing this using this code...? I don't use it to simply 'vibe code' but to help me get over a hurdle when I can't figure something out; much like how a community works but with instant response.

I view it as an adjunct and learning tool; not to simply produce cut and paste code. If it generates code

It will think outside the box, and maybe even come up with novel methods.
But none of this is of any use unless you are senior / experienced enough to ask the right questions
and guide your junior/intern towards optimal solutions.

In short its a tool but its a useless tool if the operator has little or no
domain knowledge in the areas for which the tech is being used.

Using Warp terminal, it actually nice for a non-admin to ask questions to Claude and get some really helpful work.

I do not know every in and out of Linux server config, my day job doesn't depend on that I do. So I can connect up, ask Claude, "is this service running?" or " My plex server isn't responding, can we run some diagnostics?"

Is it perfect? No, is it better than me? Oh god yes. Is my system a mission critical server? Not in the slightest.

But its fun, I actually can get a working docker serve

Yeah, needs a ton of oversight, etc, I have to read it carefully...

So I wanted to do a thing, and I could easily have made a working prototype of this thing in a couple of days, maybe a week tops. So I had claude do it and spent nearly an entire afternoon fixing things or making claude fix them because I understood the problem better.

But at the end of the day, I'd spent an afternoon doing something I would have expected to take a week.

I've used AI for two tasks where I found it very useful, and some very minor ones as well.

1) I had to write some code to invert a matrix in C. I knew the code was out there, but Google's search is so polluted today I could not find it. ChatGPT immediately returned working code. I noticed it did not calculate the determinant, so I asked it for that, and it modified the code to do so. As I say, I know that code is out there somewhere in a book, probably a dozen books, but I can no longer find older topics bec

It is completely worthless for anything that isn't completely trivial.

Even the glorified pay-for-use versions.

Software, electronic hardware, history questions - no matter what you ask it, the "AI" is worthless.

There is limited utility in some machine learning methods when sifting through bad data from physics experiments, but even there the applications need triple checking.

The only thing where the "AI" excels is creating trolling posts for maga-like readers.

Like the title says, its effectively an outsourced junior dev. It has no experience of its own and is just parroting what it read, imperfectly. Plus you probably aren't generating any institutional knowledge.

Source code + compiler =reliable output , prompts + LLM-du-jour = inconsistency. Vibe-coding is going to leave a lot of people with zero actual knowledge of their product as if they had out-sourced the development.

Because that's what they did.

So with that out of the way...its really good at first-dra

I don't use AI in cases of creative expression. Using AI for that feels downright dishonest, lazy, and unsatisfying.

However, when I need some information which is not straightforward, I use it. However, only if search engines don't give relevant information from which I can extrapolate things on my own.

I prefer to use my own brains.

ChatGPT Pro Deep Research has been very useful when researching my own genome. I use chatgpt to write the prompts for Deep Research, and AFAICT the results are accurate. From time to time, I ask Grok to check the output for correctness.

I read a great deal, particularly speculative fiction. I've tried repeatedly to use various AI tools to track down a short story or book where I can remember details of the story, and perhaps the rough publishing date, but not the title, author, or publisher.

AI appears to be utterly useless for this. It will either come up empty and make vague suggestions, like "Look at fantasy recommendations for the date range you've provided". Or worse, it will focus on the wrong book (say, one published in the 2020s, wh

Is there anything better for TTS, STT, and translation?

Problem I recently worked on:

Here is 20,000 hours of audio. Make it queryable.

Back in the 90's when I was doing some grad classes in Information Retrieval this would have been considered nearly insurmountable.

On the other hand I had 16MB then, now this takes 128GB of RAM.

That's mostly Python being obscene with RAM.

"Everything required some form of correction or editing before it was usable."

So just like a human assistant.