User Profile

This is a great summary of everything that's wrong with the current hype around AI and especially LLMs. It doesn't expect any prior knowledge to the field and is a very good introduction to a lay audience. The authors (@emilymbender@dair-community.social and @alex@dair-community.social) also have a great podcast that I'd recommend: "Mystery AI Hype Theater 3000".

The book has so many quotable passages that I can't possibly list them all here but to give you an idea: * In the vast majority of cases, AI is not going to replace your job. But it will make your job a lot shittier. * As with AI "art" discussed above, AI boosters think that science is only about ideas, rather than communities of practice. * The point of talk therapy is not to exchange text strings, but rather human connection, which furthermore is guided by the expertise of the therapist. And so many more!

The book does a great job of showing how ridiculous the claims of AI boosters are when taken at face-value but also what impoverished views of the world and notions of truth and expertise AI carries with it. I particularly liked what they say about information retrieval, which is supposed to remove "friction":

friction in information access is actually not only beneficial, but critically important.

This is crucial, and something I've noticed time and time again. People using LLMs for search are ready to trust the output simply because it appears to be the one answer to their query; they don't have to go through a list of links, evaluate them and make a final judgement call about their relative merits. But this exercise is essential! A document is not just a collection of text strings, it has an author, an origin and a context, and it's only when all those things are taken together that one can really gauge how much trust they should assign to it.

(To French readers: this also ties nicely with "Éloge du bug" by Marcello Vitali-Rosati @monterosato@mamot.fr, where friction is a central element.)