Hacker News Reader

Comment by datakan | Hacker News Reader

If the AI is digesting all the philosophy material ever published then why do they need philosophers?

knowing all the philosophy every published is not being a philosopher

there was literature about 15 years or so ago stating Philosophy as being an uncommonly lucrative course of study, in part citing Reid Hoffman

it is a way of thinking

Philosopher vs MBA. Everyone dogs on MBAs.

Philosophy can have strong mid career earnings especially if you go into law. Or get lucky like Reid did.

> knowing all the philosophy every published is not being a philosopher

Debatable. We may need to ask a philosopher.

Philosophy is a living process of integrating ideas. Classical materials are the whetstone upon which the mind is sharpened. Unlike history, where literal established accounts are ideal, in philosophy one is expected to view today (or the future) through the lens of contextual discourse.

While there is “no right answer” understanding what the issues are and how the discussion plays out is relevant.

I believe that you mean “whetstone”.

That is not what AI is. AI is a powerful tool, a semiautonomous set of wood working tools that still need a master craftsperson to use. You need the tool+genius to drive it. Everyone wants to shoot down AI but they think AI will do everything. Being proud of a creation where someone did style transfer between spongebob and Rembrandt and they think they made art. About as responsible for actual art as just downloading images from google.

I'm not seeing any evidence of this. Precision tools raise the ceiling. AI mostly just raises the floor. Ease of use is a focus point for all AI labs and it's what they're constantly trying to improve. Yes, an expert can juice these models for all they've got, but an average Joe today is probably getting better results than the best power users had a year ago. Extrapolate this a bit and ask yourself if businesses will ever want to pay your geniuses and craftspeople a professional's wage if they could get 'good enough' results from any desperate minimum wage worker, or even by doing the work themselves.

It’s actually pretty bad at it. I think there just isn’t enough literature to get a good effect from the LLM approach. Good luck with verifiable rewards when the target discourse is effectively pure self-criticism. Maybe ‘disputable rewards’ …

I have found that with proper framing I can get good help from Claude and ChatGPT on questions of translation of haute German philosophy and, to my amazement, Ancient Greek. An immediate ‘translate this passage’ request is a cataclysmic disaster. The nexus of sentences differs from other forms of discourse.

Reply on news.ycombinator.com