Comment by Razengan

My two cents is there's some implicit arguments you're making here:

- LLMs give users more knowledge (debatable due to hallucinations and selection bias)

- having more data/knowledge means more power and therefore less authority / power imbalance (that was an argument for the Internet and social media, but it turned out exactly the other way around)

Both of these arguments is dubious, but somewhat reasonable, but requiring both to be true to achieve any form of net positive is a hard sell. That, plus as Al Khatib argues, AI is about taking away autonomy from the people. Beyond the cognitive studies about LLM impacts on our brains, having someone/thing else decide for you is decidedly incompatible with anti-authoritarian politics (autonomy vs heteronomy).

"Every human SHOULD have access to all human knowledge at their fingertips, and a personal assistant to help them navigate through it."

I agree.

I absolutely think that is not what all this compute power is for.

I do believe that all this statistical inference and generation, alongside the massive data collection apparatus we have built, is designed explicitly to completely lock down the possible actions for any person targeted by the folks who own these systems, and these systems aren't owned by "every human".

Quite the contrary- historically the folks who own those systems are pretty, uh, specific about what they man by "human". Those folks are so capricious about that concept of human they can't believe brown folks are human while their kids suspect the computers might be human.

How is it leveling the playing field if the big actirs have access to a million X (or something in that ballpark) the compute, and even more in terms of data?

Oh, and they are using consumers prompts to gather even more data. LLMs are creating even greater asymmetric information situations than we had before.

Also, the common person has no say in how these systems are developed, the big actors do, and these systems are already used to automate decisions. That's a totalitarian potential, not just authoritarian.

> Every human SHOULD have access to all human knowledge at their fingertips

So, your banking passwords?

Some technology carry politics as part of their essence. A knife or a hammer is a relatively neutral technology. An atomic bomb on the other hand is a political project in itself and has no charitable interpretation (game theory / mutually assured destruction aside).

I think AI falls into the second category. Even the most optimistic interpretations of AI, such as in Asimov, are very totalitarian. They're only positive because AI is automating away human work to create more resource/financial equality between people. In our real world though, AI is just as totalitarian, but there's not a single glimmer of hope that it will be used to better share resources. That sounds more political than technical.