AI's impact on the open web: manton.org
@manton Really good post.
Personally, I agree with your last three stated beliefs, but not the first. I believe AI, (but not LLMs) will one day be an incredible resource and will transform education and how we work.
But chatbots make me more pessimistic about even this weaker version. I hate ‘em.
@njr We'll have to disagree on the first point, but I get it. It is also still very early in all of this.
@manton This may seem like a lateral, but I should point out that robotics has improved significantly over the past two decades, and with the state of capitalism as it is today, AI should not be seen as a positive, but a threat. Capitalists are flogging the hell out of early AI to spur development in hopes of accelerating advancements in two areas, AI models and autonomous robotics. Both of these will likely be used to replace the one thing capitalism cannot control: humans.
It's dark and dystopian and bleak, but taken in context with all of the crazy talk around Mars, the rise of authoritarianism and accelerationists developing a larger following, and humanities increasing dependence on the "algorithm", I think I'm right. I don't want to be right. I really don't.
@manton nice article.
#Kagi search with AI actually spits out citations and sources. I believe they are looking into how they can give back to sources too.
But I agree, AI slop is getting worse. What happens when LLMs learn from LLMs learning from LLMs with no human content?
AI content should be clearly labelled as such also, so that LLMs don't further loop their content.
@tylerknowsnothing I hope you're wrong! I think AI in robots is a mistake. AI can be great but needs human oversight.
@manton I hope so, too. Just look at how far it's come in the last few years, and now we're in the firing chamber of the Trump administration and he's just pulled the trigger. Now we're going to be heading down the barrel for four years. We're just 53 days in now and they haven't even warmed up yet. Trump & Co hate regulation and they effectively control all three legs of the American stool which are supposed to provide checks and balances to the others.
I'm not that hopeful, after all.
@manton We may need a web filled with human-generated content, but I don't see how we can get one. The cost of generating slop appears to be falling asymptotically to zero, and the incentives for generating it are rising as international criminals move to the web.
@manton @tylerknowsnothing I have other “theoretical questions” about language models (vis-a-vis knowledge and education), but this here is what my practical concern is.
Mars ambition, and beyond that, TESCREAL accounts give an idea where these peoples’ heads are. It is very dark. The political futures they imagine require heavy if not absolute population control. (Including Mars colonization.) Whatever other uses well-meaning people might find for it, the language models and algorithms increasingly appear designed, and marketed, to suit those dystopian ends.
@lzbth Indeed. Or, in other words, the strong shall live and the weak shall die. In today's context being weak is the same as being moral and good and not a scheming bastard willing to defraud people, or being "different" or an "other" or "sick". Basic eugenics. After so many decades of the evisceration of America's public education systems, I guess the TESCREALists think they finally have enough leverage to make their dreams come true. Fun...