This is what happens in thought-isolation. It isn’t better than educating yourself, whether that education involves AI or not.

Phillip Kitcher is known for epistemic monoculture, Dawkins and then Henrich popularized collective intelligence and cultural evolution.

The thing about these fear pieces is concepts like the hollowed mind are reductive and that reductionism is based on a reductive view of (usually other) people.

But what actually happens is we have formalized processes and can externalize them. This is a benefit if you can use your newfound capacity and free time for something better, which I think most people ultimately will.

I'm not sure why this is at the top of the page; it's not that it's wrong, it's just a sequence of truisms.

I also don't believe that everybody I know is idiosyncratic in the way they view the world. And even if they were, I'd probably just pay attention to the things that are directly relevant to me. So probably I'll misunderstand most of what they say anyway.

That is, of course, provided that you pay attention it actually does research. In their current state, LLMs are practically useless for this purpose for the vast majority of users, as no one knows how they work, what to watch out for, what the failure modes look like, and how to keep nonsense apart from facts when both are presented with an equal amount of conviction. That’s not a user problem, it’s an education problem.

Isn't this just inherent to any system that takes some time to update? E.g. if a country moves its capital to a different city, then textbooks, maps, etc. are going to contain incorrect information for a while until updated editions are published.

A lot of the complaints about AI are really about the drawbacks of information systems more generally, and the failure modes pointed out are rarely novel. The "Cognitive Inbreeding" effect attributed to AI would also have occurred with Google search would it not? Lots of people type the same question into google and read the top results, instead of searching a more diverse set of information sources. It's interesting that the author mentions web search as a way to ameliorate this, when it seems to me that web search is just as capable of causing cognitive inbreeding.

Nothing about the nature of evolution implies our current cognitive processing is ideal/sacred and shouldn't ever change.

The framing of questions massively affects the results you get from discussion with humans, and I'd argue it's even more pronounced with LLMs.

I do think there's a solution to this—kind of—which dramatically reduces the probability and allowing for broad inductive biases. And that's to ask question with narrower scopes, and to ensure you're the one driving conversation.

It's true with programming as well. When you clearly define what you need and how things should be done, the biases are less evident. When you ask broad questions and only define desired outcomes in ambiguous terms, biases will be more likely to take over.

When people ask LLMs to build the world, they will do it in extremely biased ways. This makes sense. When you ask it specifics about narrow topics, this is still be a problem, but greatly mitigated.

I suppose what's happening is an inversion of cognitive load, so the human is taking on more and selecting bias such that the LLM is less free to do so. This is roughly in line with the article's premise (maybe not the entire article, though), which is fine; I think I generally agree that these are cognitive muscles that need exercising, and allowing an LLM to do it all for you is potentially harmful. But I don't think we're trapped with the outcome, we do have agency, and with care it's a technology that can be quite beneficial.

Isn't this whole thesis negated by the fact that tool calling web search exists? This just feels like a whole lot of words to say, don't treat a LLM as an always up to date infallible statistical predictor.

Would you attempt to, for example, simultaneously modify for available ingredients, number of diners, and time-optimize the prep method for a recipe you've never cooked before if you were following an old-school cookbook? No. You'd have to be a pretty solid chef to try all that on at once.

Using AI, you might branch out confidently in to new areas, executing all of these modifications simultaneously, and even adapting the output for a specific audience or language.

This toy example shows an important property of AI as decision support systems, which are well studied in the military domain: using these systems, we build confidence to act in unfamiliar domains, thereby extending our reach. From this experience we can learn more. The fact that the learning may then occur through, ie. during or after the experience, rather than beforehand, is secondary. It's still there. The fact we didn't know the language the AI translated to for our chef is totally irrelevant.

Sitting comfortably at the effective apex of millions of years of human cognitive and technology development with the entire world's knowledge at our fingertips, every day we can extend confidence in novel domains through AI, and enjoy it. We should be feeling pretty damn "developed".

Rote formalism and fixed paths in pedagogy are gone: good riddance. This is the hacker age.

Slightly FTFY.