Nobody says LLMs aren't impressive. But there is a subtle difference an impressive trick and being something worth throwing 1% of GDP at. Proponents say that if we keep throwing more money at it, it will improve, but this is far from certain.
> 1 % of GDP
LLMs are basically the only thing genuinely new in decades , that have someone excited in basically every dept in the entirely world, why is it so bad that we spend money on them? The alternative is going back to shovelling web3 crap.
There's definitely a new generation of bullshit merchants to go with LLMs but I think they (the models) target a very different part of the brain to normal tech so in some ways they're much more resilient to usual fad archetypes (this is also why some people who are a bit jittery socially hate them)
This money (and more importantly, this electricity and this amount of silicon) is diverted from other stuff.
It’s just a technology; one that will improve, sometimes stagnate, sometimes accelerate. Like anything else, right? I don’t see a time when we’ll just stop using AI because it “feels so trite.”
But if it is a half decent chatbot and has the label "AI", it gets another iteration with 10x the resources. This has happened a few times already.
It is a neat tool. It is very unreliable. Teams saying "just give us 10x as much resources so we can insist on this approach" is the hateful thing here.