There's a viral dynamic to LLM use on programming teams that has profound consequences. All it takes is one guy who is all-in on code gen to start pushing for its use, and the entire team has to react to that in one way or another. In many cases the default social reflex is to live and let live, which in practice means letting LLM use proliferate and completely transform the team, its processes, and its output for the worse.
4 Boosts
0 Likes
@KuhllervOh There's only nuance if you're willing to disregard the massive externalities of just the model training, even before any inference is ran against it.
It's also a product that is massively subsidized by a handful of US companies vying for the ultimate monopoly on absolutely all intellectual work. There can be no nuance until the upcoming market readjustment which will reveal the true price of inference tokens.
Then we can talk about "what it's good at" given its real cost for businesses and consumers alike.
Thing is, as currently designed, it's not a tool.
It's a weapon.
The damage is the point.
@jplebreton a huge issue with LLMs is manifestation of status quo in lockstep with loss of sensible doubt. Quoting Jurassic park:
Dr. Ian Malcolm:
Yeah, but your scientists were so preoccupied with whether or not they could, they didn't stop to think if they should.
We have managers excited by LLM crunching numbers that were tedious to get.
Why is it tedious? Improving the upstream process? Simplification? Human conversation? Do we need this? What happens if we depend on this?
not anymore
@jplebreton In IT, there is no way to avoid LLM at this point. It's in every f*cking product, in every damn app, in the OSS projects or their contributions, on the platforms building and hosting your stuff, in the comments, on every website. Not using LLMs is now either denying reality or living under a rock somewhere far far away.
IMHO, there is no right living in the wrong. Push back, yes! But my main goal is this technology being in the hands of the people, democratized.
As it should.
@mrkaiser Right, let's put asbestos or radium in the hands of the people. Democratized.
@jplebreton That's what is happening currently at my company. They want to push for AI, and even if you say no in your whole team, other teams start vibecoding stuff left and right, things that touch your own team and that have a big impact on your work...
It's so freaking icky and disgusting, how it blurs and fucks up everything, responsibilities, ownership, skills, ethics, morals, motivation, morale, trust, respect for peers, etc etc...
Team members who push back on that, no matter how politely and principled they do so, get called luddites, mellow-harshers, bad teammates - they're the villains, not the guy who brought in the slop.
If the team's LLM use becomes a matter of public record, resistors speaking out would be seen as a betrayal. So they sit and suffer quietly as their teammates turn their brains to mush and make everyone's jobs worse, with likely few to no clear benefits to the work being done.
If you are one of these people, at innumerable thousands of companies or open source projects right now, engaging in principled resistance, doing your professional best and advocating for what you believe is right, then I want you to know that I and many other people respect what you're doing. It matters that you do this, and it's unfair that you've been placed in this position. I'm sorry that you have to work with people who are less conscientious than you.
@jplebreton It's taken over more than half my team at work and it's getting very tough to avoid it. It started with the VP getting AI pilled a few months back and it's rapidly taken over everything. And quality has gone to shit while they push all of us to use AI for everything now.
@jplebreton So far in my team, adoption is mixed. Some are full-on using it to produce code, some are using it in a supporting capacity, either for planning or analysing logs, and I'm here, politely refusing to use it myself. So far my boss has been receptive, but he's warned me than LLM usage monitoring could be coming.
@jplebreton There are 3 main paths with AI development tools.
1. Live and let be, leading to AI tools proliferating over time.
2. Strict anti-AI policy, often with little to no nuance with which the effects have yet to be studied.
3. Going all the way in the opposite direction and facing extreme public backlash and questionable quality.
I think there’s more nuance than those 3 paths offer. Ultimately it should be treated as a tool, used for what it’s good at like pen testing.