Comment by simonw
1 month ago
You think those billion people are all addicted to having a made-up friend?
I think it's being built to solve problems they have. OpenAI shut down GPT-4o because it leaned too hard into getting people addicted to it.
1 month ago
You think those billion people are all addicted to having a made-up friend?
I think it's being built to solve problems they have. OpenAI shut down GPT-4o because it leaned too hard into getting people addicted to it.
> You think those billion people are all addicted to having a made-up friend?
No, and I don't think you're really trying to understand what I think. I mean, do you actually think anyone could think ChatGPT user only use it in one way? But to briefly answer your off-base question: some fraction are certainly "addicted to having a made-up friend," others use it to cheat or slack, others use it trivially in easily replaceable ways (which my intuition says are the numerical bulk of users), many people use it because they're required to, and a relatively small fraction use it to build things or replace human labor in their employ, etc.
But to restate my point: AI systems like ChatGPT weren't built to solve any kind of problem most people actually have. Few people, if anyone, had a ChatGPT-shaped hole in their life in 2021. It was built because it could be built, and it continue to be built because some other people have a "problem" that they have to pay a lot of money to workers and they hope to do less of that in the future. And causing societal harm or damage doesn't seem to any kind of deterrent to many of them, if they think about it at all.
Thanks for clarifying, I genuinely didn't understand your point and I do now.
I agree that ChatGPT was launched without having an idea as to what it was for. OpenAI themselves debated internally as to whether it was worth releasing and were shocked at how popular it turned out to be.
Given what I know about what ChatGPT and similar tools can do, I firmly disagree that they don't solve problems for humans.
Computers are hard to use. People need to be able to solve problems using computers. I think LLMs are the biggest leap we have ever had in making the power or computers - previously reserved for programmers - available to everyone else.
> Computers are hard to use. People need to be able to solve problems using computers.
Solving problems is fun! And perversely, LLMs actually seem better at doing a good-enough job at the fun stuff and being worse at certain kinds of drudgery. So the people get more drudgery.
> I think LLMs are the biggest leap we have ever had in making the power or computers - previously reserved for programmers - available to everyone else.
That line sounds good, but it's not hard to understand how that could ultimately be a net-negative for most people.
We don't live in a communist society, where everyone is secure and running around solving problems for their own satisfaction and self-actualization. A lot of people survive and find the security they need by solving problems for others. Take that away and they have little need for or access to "the power [of] computers."
What that could mean in practice is the "everyone else" is actually some smallish elite that finds it can get away with sharing even less if its riches with others.
If you really want to advocate for AI, and address most of the concerns about it, advocate for literal communal ownership of the frontier models and data center infrastructure. Not some piddling 5% bribe, but 100% (or a super-majority stake, at the minimum). I'm sure there can be quite a bit of innovation in the area of communal ownership.
3 replies →