4 ms·

>The irony is that AI is best at replacing the work of the upper classes. Synthesizing different opinions, summarizing them, and producing outputs based on stat

>The irony is that AI is best at replacing the work of the upper classes. Synthesizing different opinions, summarizing them, and producing outputs based on statistics are things AI does well. AI just repeats whatever the prevailing opinion is at that time. I am a very heavy AI user (Claude, Gemini, ChatGPT) and have queried it on a variety of topics. AI is not thinking, it is repeating.

banannaise 5mo ago
Thus replacing the work of the upper classes.
ben_w 5mo ago
> AI just repeats whatever the prevailing opinion is at that time. That would be an improvement. They are generally far too sychophantic to just repeat the prevailing opinion, and instead synthesise the opinion that they think wants to be heard by the user.
jdw64 5mo ago
I agree with the view that AI does not truly think and cannot produce genuinely original opinions. It is difficult for AI to give answers that lie far outside the average distribution of its training data. In that sense, it is not very good at producing truly novel business insights. But that is not what most “work” usually means. Work is mostly repetitive. The actual moment of decision is brief. So what do I mean by work here? I mean the collection, organization, and synthesis of the materials needed before reaching that decision. For that part of the process, AI is extremely effective.
- curio_Pol_curio 5mo ago
  LLM platforms are not fundamentally collaborative (think chatGPT vs stack overflow, or google DeepMind vs Bell Labs) Otherwise they would 1. force humans to spend much more time on these decisions, and 2. Teach humans that the only way to save time on decision-making is to collaborate with other humans I had a reply to your Asian take https://news.ycombinator.com/item?id=47970518 https://news.ycombinator.com/item?id=47970518
goalieca 5mo ago
> AI is not thinking, it is repeating. This^. We aren’t talking about general intelligence. We are mostly using a large language model. It’s a model of languages and not the universe. Language encodes a lot, enough to be “unreasonably effective”.
make3 5mo ago
I'm sorry but AI generalizes on various takeaway sets of reasoning benchmarks about a variety of subjects, and if you ask it complex philosophical questions in an argumentative back and forth, it does really pretty well most of the time. I'm progressive, but this is just not helpful. Progressives should focus less on trying to argue that AI sucks actually (it obviously doesn't), we should focus on thinking about how to transform society to still protect human quality of life and dignity even when strong AI exists.

banannaise 5mo ago

Thus replacing the work of the upper classes.

ben_w 5mo ago

> AI just repeats whatever the prevailing opinion is at that time. That would be an improvement. They are generally far too sychophantic to just repeat the prevailing opinion, and instead synthesise the opinion that they think wants to be heard by the user.

jdw64 5mo ago

I agree with the view that AI does not truly think and cannot produce genuinely original opinions. It is difficult for AI to give answers that lie far outside the average distribution of its training data. In that sense, it is not very good at producing truly novel business insights. But that is not what most “work” usually means. Work is mostly repetitive. The actual moment of decision is brief. So what do I mean by work here? I mean the collection, organization, and synthesis of the materials needed before reaching that decision. For that part of the process, AI is extremely effective.

curio_Pol_curio 5mo ago
LLM platforms are not fundamentally collaborative (think chatGPT vs stack overflow, or google DeepMind vs Bell Labs) Otherwise they would 1. force humans to spend much more time on these decisions, and 2. Teach humans that the only way to save time on decision-making is to collaborate with other humans I had a reply to your Asian take https://news.ycombinator.com/item?id=47970518 https://news.ycombinator.com/item?id=47970518

goalieca 5mo ago

> AI is not thinking, it is repeating. This^. We aren’t talking about general intelligence. We are mostly using a large language model. It’s a model of languages and not the universe. Language encodes a lot, enough to be “unreasonably effective”.

make3 5mo ago

I'm sorry but AI generalizes on various takeaway sets of reasoning benchmarks about a variety of subjects, and if you ask it complex philosophical questions in an argumentative back and forth, it does really pretty well most of the time. I'm progressive, but this is just not helpful. Progressives should focus less on trying to argue that AI sucks actually (it obviously doesn't), we should focus on thinking about how to transform society to still protect human quality of life and dignity even when strong AI exists.