[gglon]

In my opinion the biggest near term danger is that AI will make people overly trust it. And they will start treating it as a truth oracle. Oracle that will be controlled by some small group of people. Namely, it will become a perfect tool for propaganda and control.

[cypherpunks01]

Agreed. I am worried that AI bots will become an object of "worship" in some sense, culturally. Like you mention, having an increased reliance on it for facts, but also for creativity, companionship, career advice, psychotherapy, etc. will have many unknown effects but definitely give frightening amounts of power to whatever corporate entities are controlling the most popular AI bots.

We will go through the whole decentralization debate again, with people saying that open-source AI will be the most transparent and fair, but with centralized corporate systems winning all market share with the amount of resources they have available to train and maintain these systems.

[goatlover]

That's a legitimate concern. I also don't like how a few corporations get to decide what appropriate content is for everyone else on the planet.

But it could also be worse in the hands of other groups, like certain governments. I imagine that's just a matter of time.

[precompute]

Seeing how this corporation had it for six months, the governments around the world have probably had this working for a few years, if under a slower / less stable beta. It's been probably been wargamed to death on the internet. And maybe even in real life conditions.

[mark_l_watson]

+1 a very good point!

At a birthday party yesterday I was talking with a friend who is a film director/producer (smart, somewhat tech savvy) and he said much the same thing as you did, adding that the whole field of generative models scares him. He was not talking about losing work to AIs, but rather he is concerned about harm to society. I agreed with him on possible risks but I argued that the potential benefits outweigh the risks. I am biased because of my age (early 70s) because I want as many AI tools available as possible to keep playing the ‘infinite game’: I want the best self driving cars to make me safer driving; I love being able to get so much more coding done now (Emacs tricked out with Copilot, and embedded ChatGPT and Dalle consoles); I have been working on information processing systems for over 40 years, and now with OpenAI, Hugging Face, LangChain, LlamaIndex, etc., this all becomes so much easier. I want more! More!

That said, I understand that people younger than myself may reasonably be more risk adverse than I am.

[ChatGTP]

I am biased because of my age (early 70s) because I want as many AI tools available as possible to keep playing the ‘infinite game’

Imo this is what it’s all about, gaining immortality and unlimited power under the guise of “good for society” through solutionism and technology.

Silicon Valley is getting older so the efforts to keep going wind up.

That’s why there’s so much anxiety around alignment. The Elixir of an eternal life might be the thing that takes it away, is this any different than in the past ? Sadly yes, but this time it might take all the young people with it.

[mark_l_watson]

I was not talking about achieving immortality via tech. We all die, turn to dust, etc. My goal is to keep playing the ‘infinite game’ every day I am alive, and enjoy each day. It is because we will die that the time we have has such high value.

BTW, I don’t take the life extension crowd seriously.

[ChatGTP]

Sorry I misunderstood what you meant!

[mark_l_watson]

Not a problem!

[precompute]

Just wait until you can type in a paragraph of text and tell the LLM to respond in a manner that reflects the way you write. Forget language translation, we'll have mood / writeprint translation because people will be unable to understand anyone else's viewpoint.

For example, if I'd like the LLM to tell everything to me in pig latin and with excessive "cool"s and "yo"s, I could do that and it'd accept. Now, for people that don't know how to read very well or understand a language well, this will be catered to their level and they will lose whatever modicum of familiarity they had with the system.