Comment on Devs gaining little (if anything) from AI coding assistants
FlorianSimon@sh.itjust.works 1 month agoI’m talking about people criticizing LLMs. I’m not a politician. But I’ve seen a few debates about LLMs on this platform, enough to know about the common complaints against ShitGPT. I’ve never seen anyone on this platform seriously arguing for a ban. We all know it’s stupid and that it will be ineffective, just like crackdowns on VPNs in authoritarian countries.
The reminder is the tech itself. It’s yet another tech pushed by techbros to save the world that fails to deliver and is costing the rest of the planet dearly in the form of ludicrous energy consumption.
And by activism, I mean stuff happening on Lemmy as well as outside (coworkers, friends, technical people at conferences/meetups). Like it or not, the consensus among techies in my big canadian city is that, while the tech sure is interesting, it’s regarded with a lot of mistrust.
You can take LLMs seriously if you’d like. But the proof that the tech is unsound for software engineering keep piling up. I’m fine with your skepticism. But I think the future will look bleaker and bleaker as times goes by. Not a week goes by without its lot AI fuckups being reported in the press. This article is one of many examples.
tee9000@lemmy.world 1 month ago
Theres no particular fuck up mentioned by this article.
The company that conducted the study which this article speculates on said these tools are getting rapidly better and that they arent suggesting to ban ai development assistants.
Also as quoted in the article, the use of these coding assistance is a process in and of itself. If you arent using ai carefully and iteratively then you wont get good results with current models. How we interact with models is as important as the model’s capability. The article quotes that if models are used well, a coder can be faster by 2x or 3x. Not sure about that personally… seems optimistic depending on whats being developed.
It seems like a good discussion with no obvious conclusion given the infancy of the tech. Yet the article headline and accompanying image suggest its wreaking havoc.
Reduction of complexity in this topic serves nobody. We should have the patience and impartiality to watch it develop and form opinions independently from commeter and headline sentiment. Groupthink has been paricularly dumb on this topic from what ive seen.
FlorianSimon@sh.itjust.works 1 month ago
Nobody talked about banning them, once again. I don’t want to do that. I want it to leave the mainstream, for environmental reasons first and foremost.
The fuckup is, IDK, the false impression of productivity, and the 41% more bugs? That seems like a huge deal to me, even though I’d like to see this study being reproduced to draw real conclusions.
This, with strawberrries, Air Canada’s chatbots, the 3 Miles Island stuff, the delaying of Google’s carbon neutrality efforts, the cursed Google results telling you to add glue to your pizza, the distrust of the general public about anything with an AI label on it, to mention just a few examples… It’s starting to become a lot.
Even if you omit the ethical aspects of cooking the planet for a toy, the technology is wildly unsound. You seem to think it can get better, and I can respect that. But I’m very skeptical, and there’s a lot of people with the same opinion, even in tech.