it's so wild to me that in 2024 half of the #LLMs related posts on my timeline (after filtering a fair amount of people that annoyed me) is still: "these things are useless lying pieces of nonsense". Do you live under a rock? How does this happen? Did you ever try these things out?
Or are people happy skimming the foam at the surface and then complaining their thirst doesn't get stilled and that they now have milk foam on their chin? (lol where am I going with that analogy...)
@mnl I'm still not convinced by the things. There is something to do with LLNs, but they are being hyped. They bind us to companies in ways that we should not accept. I see people using them in inflationary quantitative ways, and we risk another cultural devaluation of work that we shouldn't ignore.
AI - yes, but if it causes further economic imbalance then we should really take care of that. Right now we don't - and that's a real issue.
@promovicz is that what you're seeing one year in? The local models are starting to shred (I'm not entirely sure what cultural devaluation of work means).
One of the consistent impact of LLMs I see is that people build things they couldn't have imagined building before (me included). And they build things, it's not just a dream. Excel macros, python webapps, ios apps, learning something. It's so weird to me that people would look at that and be like "nah, you're deluding yourself".
@promovicz I don't think you can discuss impact of LLMs meaningfully if you're starting position is that they're useless. That's actual the most frustrating thing about all of this. How can I have a meaningful conversation if you don't acknowledge the ground I'm standing on.
@mnl You're missing my point if you skip the "devaluation of work" argument. LLNs can be used productively, but we need to question whether they do so in a way that our society can sustain. Socialism dictates that we must maintain work perspectives for people to some degree. Are we doing that? Or are we just firing the overhead - which is the most likely outcome?
@promovicz i'm still not sure what you are getting at, and if we are not talking about different things. I think that any meaningful discussion of the impact of LLMs on work, and examining which agendas the hype is pushing, cannot happen if you don't even acknowledge that these things are actually useful.
@mnl "LLNs can be used productively" - that's what I said. If that's not "actually useful" then I don't know what is.
The problems are socio-economic. If we let these things replace us - even by raising individual productivity - then it seems likely that all we do is to eliminate jobs while leaving people hanging in mid-air - and that's also what the companies want. They want to raise worker productivity so they need less workers. That's bad.
@promovicz right but that's not my rant here. Don't turn this into a productive discussion. The thing you are bringing up is what I want to discuss. You need to at least be open to the premise that these are productive, in ways that are good.
So many developers complain about boilerplate. Obviously a problem we can't solve despite all our fancy haskells and lisps and transpilers. And here's a tool that sets boilerplate on fire, and people are like "nah this doesn't work...".
@mnl I simply don't think that "developer boilerplate whining" is a relevant problem - because I don't think that worker happiness relates to productivity very much.
In system software, we need better code rather than more code - and that's not a strength of the LLNs. In app software and product software, all this is going to do is to cost jobs - because the markets are already inflationary.
@mnl Summary question: can AI create the equitable jobs that we need to sustain our social system? I think that it can't (precisely because it is not GI).
@promovicz What I wonder is it's june 2024, out comes "figma to react 1.0" that legit works (and I know we'll get there). Now it is good engineering (not just business) to let a single designer and a single developer do what 4 teams used to do before. (good engineering because too many cooks and communication inefficiencies and what not, plus the increased iteration speed means you probably get better results).
and now what? Even if I wanted to keep the 3 other teams, what do I do?
@mnl Before rampant capitalism it was extremely common to maintain jobs for their "human usefulness" - just "sit a human face in the chair". This is how we managed to keep employment up - and that's something that we have to do in times like these. We need easy happy jobs - and increased productivity may only risk the transition. We need to redistribute - and that's way more important than AI.
@mnl ha yeah. i imagine if a wizard appeared and provided us with literal magic, those same people would argue, "but it doesn't do X or Y, and here's some other problems". like, can we not appreciate what we have?
@kellogh it's so weird to me. Like ok, I get being turned off by people pretending ChatGPT can make your business 100x better or something, but just put a programming question into perplexity or phind, be prepared for some stochasticity in the result (you've used google before, and everybody seems to also know how LLMs work), and then assess if that technology isn't just freaking amazing...
It OK to complain about google using it to fill their top result after that, but saying they're useless?
Add comment