Post

this technology that i cant even get my own employees to use without literally threatening them is definitely going to be profitable one day

3.1K reposts9.3K likes

They should use it the bare minimum. "I used it! I used it to translate this one sentence and then threw away the answer and did it myself because I don't trust your piece of shit AI garbage."

Heh. If only I could see it being dropped from harmful use. Unfortunately though, it's already caught on with big corporate, and the related big politics. Google being one notable example.

That’s not fair. The austerity and funneling money into fascist police state infrastructure is the second half of the end goal, because they know constant capital does not realize profit and the whole thing is going to implode and take civilization down with it.

Considering how Microsoft makes their own OS miserable to use by forcing Bing searches into every nook and cranny of the software to juice the numbers, who the fuck even knows.

That's not the thought process. It might be trying to standardize output but it also seems likely to me that they're trying to use the rest of their company to accelerate their AI push as an additional unpaid labor aspect - everyone is frantically trying to hit AGI first

As a dev with 29 years under the belt, I have been trying Cursor on an example microservice project. It does dumb things, confidently. If you want the code equivalent of 3 arms and seven fingers, they got ya covered.

I graduated as a systems analyst last year and, therefore, signed up for job rejections and whatever this is. I find it hard to believe a LLM knows how to develop, let alone know what it's developing, and it only makes sense that the less one knows the more correct the output looks to them.

I’ve added a “rule” to my Cursor settings: Don’t let me give you a prompt that asks for too large of a chunk of work. Between that and asking for smaller, more directly actionable, and more concise stepwise changes, I’ve seen many fewer dumb things come back. It’s an intern, not an engineer.

"My boss paid a lot of money for this machine that hurts my fingers, has a high pitched whine that gives everyone in the office a headache, and makes the coffee taste like fish."

I work at a (non-AI specialized) software company and our CTO said AI proficiency will be part of our reviews. And as it turns out, I'm the only engineer on my team who's not already gung-ho about it. These guys love ChatGPT and Copilot and I feel crazy for lacking the same enthusiasm.

I'm so sorry. You're not alone. Making it part of the review process is absurd. "[gen]AI proficiency" must include knowing when *not* to use it, which can definitely be most or all of the time. I know how to use recursion perfectly well, which includes not using it.

I’m a junior. Not even a senior, and even I find it more difficult to understand what the model produces, fix any issues etc, than it is to just do it. A decent example of a similar problem (ie a good Substack page) is a 10x better reference than trying to debug a non-functional AI response.

Been in various jobs, including wildlife conservation and aircraft maint (so ya know, a range) and one of the biggest hurdles is unpicking a job half done by someone else who isn't able to explain it to you. You invariably have to go back to square one.

One place I do find models useful is unknown unknowns. Solve a problem, get it working, THEN feed it to a model and ask for some methods to improve it. Was especially useful learning python coming from C, to learn pythonic methods I either didn’t know or forgot about.

Lol I've gotten the "model their use on our team" talk because the senior engineers at our company also don't like it They seem to think senior engineers will magically fix every issue these things have instead of seeing them as disqualifying

When I needed help figuring out how to fix something in Power Automate, consulting the stuffed crow I keep at my desk was more effective than using fucking Copilot

I have been /tempted/ to use AI to help solve a coding problem but once... And that is because the comments in the file itself say (essentially) "we were *very* drunk when we wrote this, so don't even bother asking us how it works, good luck!" But no, I'll figure out how to get it to work someday.

I just retired from working as a software engineer for 30 years. I had been using GitHub Copilot (voluntarily) for the previous 8 months, and was amazed at how it improved my productivity. IMHO, any software engineer *not* using AI tools is living in the past.

Everyone in silicon valley is all-in on AI because the allure of never having to pay employees again is too great. But they can't fix the models that are poisoning themselves with the junk data that LLMs are polluting the internet with. The crash is going to be horrific and spectacular.

"but it can help you write an email faster!" Bitch, I have 4 qualifications and have been doing this job 13 years, if I'm taking a while to write an email it's because it's important enough for me to really check details and tone and I wouldn't trust that to AI anyway

Before I left twitter I saw someone share 2 paragraphs describing an Aussie football game asking people to work out which was AI and which was human written with the smug implication that the AI one was more professional and better written. But it was very clear the AI hadn't watched the game.

The human one used the slang properly, referred to the MCG and the 'Pies, described the game as being an imperfect game. You could tell it was written by a sports writer for fans. The AI used the cliched descriptions that the game was exciting and thrilling, and wrote out the whole club names.

Yeah, the most time AI would save me is not having to constantly check to see if I slipped and put "You stupid piece of shit" in a message reiterating something for the third time. But I have faith that AI will one day get that tired of human idiocy.

imo it's more of a referendum on the entire package of "defining 'ai' in whatever way makes that claim possible while also laying everyone off and redirecting business objectives to the dumbest shit imaginable while still making everyone accountable on paper for security and shit"

I'm a software engineer. We love good tools. When we find a good tool, we play with it, figure out what it can do, and talk a lot with other software engineers about why they should use it too. If there is a tool that SWEs know about, but grassroots adoption isn't happening, it's not a good tool.

I can't think of a reason to force employees to use ai other than maybe they are hemorrhaging money and they want to at least feel like they are getting something out of it instead of it just sitting there bleeding money. I could be wrong but I just don't see how ai is profitable right now

It feels to me like how they'd force their employees to work on-site when they proved to be much more productive when they worked from home. The quality of the work is not important. What's important is to show everyone that their expensive building is actually being used.

For the CEOs who aren't *making* the program: it's because they paid a lot of money for a lot of corporate licenses after getting wooed by a powerpoint, as they tend to do.