[theSherwood]

The analogies to previous technologies always seem misguided to me. Maybe it allows us to make some predictions about the next few years, but not more than that. We do not know when/where we will hit the limits on AI capabilities. I think this is completely unlike any previous technology. AI is intentionally being developed to be able to make decisions in any domain humans work in. This is unlike any previous technology.

The more apt analogy is to other species. When was the last time there was something other than homo sapiens that could carry on an interesting conversation with homo sapiens. 40,000 years? And this new thing has been in development for what? 70 years? The rise in its capabilities has been absolutely meteoric and we don't know where the ceiling is. Analogies to industrial agriculture (a very big deal, historically) and other technologies completely miss the scope of what's happening.

[Imustaskforhelp]

Let me give my two cents. I remember when people used to think ai models are all the rage and one day we are gonna get super intelligence.

I am not sure If we can call the current sota models that. Maybe, maybe not. But a little disappointing.

Now everyones saying that ai agents are the hype and the productivity gains are in that, the Darwin Gödel paper which was recently released for example.

On the same day (yesterday), hn top page had an ai blog by some fly.io and the top comment was worried about ai excelling and that as devs we should do something as he was concerned what if companies reach the intelligence hype that they are saying.

On the same day, builder.ai turned out to be actually Indians.

The companies are most likely giving us hype because we are giving them valuation. The hype seems not worth it. Everyones saying that all models are really good and now all that matters are vibes.

So in all of this, I have taken this.

Trust noone. Or atleast not take things at face value of ai hype companies. I genuinely believe that ai is gonna reach a plateau of sorts at such moment like ours and as someone who tinkers with it. I am genuinely happy at its current scale and I kind of don't want it to grow more I guess, and I kind of think that a plateau might come soon. But maybe not.

I don't think that its analogous to species but maybe that's me being optimistic about future but I genuinely don't want to think too much as it stresses my brain too much and makes evene my present... Well not a present(gift)

[theSherwood]

LLMs have only really been around a handful of years and what they are capable of is shocking. Maybe LLMs hit a wall and plateau. Maybe it's a few years before there's another breakthrough that results in another step-change in capabilities. Maybe not. We can focus on the hype and the fraud and the marketing and all the nonsense, but it's missing the forest for the trees.

We genuinely have seen a shocking increase in reasoning abilities over the course of only a decade from things that aren't human. There may be bumps in the road, but we have very little idea how long this trajectory of capability increases will continue. I don't see any reason to think humans are near the ceiling of what is possible. We are in uncharted territory.

[Imustaskforhelp]

I may be wrong,I usually am but wasn't ai basically possible even in the 1970s but back then there were of course no gpus and basically alexnet showed that gpu are really effective for ai and that is what basically created the ai snowballing.

I am not sure,but in my opinion, a hardware limitation might be real. These models are training on 100k gpus and like the whole totality of internet. I am not sure but I wouldn't be too certain of ai.

Also,maybe I am biased.is it wrong that I want ai to just stay here,at the moment it is right now. It's genuinely good but anything more feels to me as if it might be terrifying (if the ai companies hype genuinely comes true)

[tptacek]

I've got no dog on this hunt at all, the idea that any give AI company could be a house of cards is not only plausible but is the bet I would place every time, but the whole "builder.com is all Indians" thing is something 'dang spent a half an hour ruefully looking into yesterday and it turned out not to be substantiated.

[Imustaskforhelp]

I am not sure but I read the hn post a little and didnt see that part I suppose.

But even then,people were defending it,saying so what,they never said that they aren't doing it or SMTH. So I of course assumed that people are defending what's true.

Maybe not,but such a rumour was quite a funny one to hear as an Indian myself.

[kypro]

While robotics are still relatively immature, I would think of AI as something akin to a remote worker.

Anything a human remote worker can do, a super human remote worker will be able to do better, faster and for a fraction of the cost – this includes work that humans currently do in offices but could theoretically done remotely.

We should therefore assume if (when) AI broadly surpasses the capabilities of a human remote worker it will not longer make economic sense to hire humans for these roles.

Should we assume this then what is the human's role in the labour market? It won't be their physical abilities (the industrial revolution replaced the human's role here), it won't be their reasoning abilities (AI will soon replace the human's here), but perhaps jobs which require both physical dexterity and human-level reasoning ability humans might still retain an edge? Perhaps at least for now we can assume jobs like roofing, plumbing, and gardening will continue to exist. While jobs like coding, graphic design and copy writing will almost certainly be replaced.

I think the only long-term question at the moment is how long it will take for robotics to catch up and provide something akin to human-level dexterity with super-human intelligence? At which point I'm not sure why anyone would hire a human except from the novelty of it – perhaps like the novelty of riding a horse into town.

AI is so obviously not like other technologies. Past technologies effectively just found ways to automate low-intelligence tasks and augment human strength via machinery. Advanced robotics and AI is fundamentally different in their ability to cut into human labour, and combined it's hard to see any edge to a human labourer.

But ether way, even if you subscribe to this notion that AI will not take all human jobs it seems very likely that AI will displace many more jobs than the industrial revolution did, and at a much much faster pace. Additionally, it will target those who are most educated, which isn't necessarily a bad thing, but it unlike the working class who are easy to ignore and tell to re-skill, my guess would be that demands will be made for UBI and large reorganisations of our existing economic and political systems. My point is, the likelihood any of this will end well is close to zero, even if you just believe AI will replace a bunch of inefficient jobs like software engineers.

[theSherwood]

This matches my expectations for the near term pretty closely.