The AI Con

AI is being increasingly pushed as inevitable everywhere, but this book provides hope and recommendations for how to see through that hype. I had a slightly different perspective than most of the audience of this book, as a developer who has worked for years in language technology, but for the most part, the technical details matched my experience. I definitely recommend this book for anyone wanting to push back on the attempts by businesses and government agencies to automate away our jobs, knowledge, and vital services, which has become even more important in the last few months.

A much-needed overview of the shortcomings of AI - and, in particular, of the AI vendors. There's a lot here that's important for everyone to understand as the technology threatens to become omnipresent. I'm glad to recommend it as a primer; it's important to keep your eyes open. But the question is left open: okay, so now what?

I liked the suggestion that smaller, more ethically-trained, more specific software is better, and I completely agree. But in a world where the generalist models are being added to everything, where rampant surveillance is being incentivized by their use, and where businesses and non-profits are being held accountable by their shareholders and board members to investigate the technology because of the hype surrounding it, what are the real, immediate solutions? The problem is clear; the way forward is not.

A nice and approachable introduction to AI in general, and LLMs in particular, as hype, with lots of examples, and a very detailed chapter on what to do (as opposed to just griping about it).

I enjoyed reading this book, though as someone who watches Mystery AI Hype Theatre 3000 pretty regularly, the information contained within is something I was already aware of.

For those of you who haven't read it, give it a read - it covers a lot of the underlying lies being bandied about in the current AI Hype bubble, and the sheer amount of bullshit coming out of Silicon Valley as they desperately grasp for the trillion dollar use case.

Overall, pretty short, but extremely well cited. The only ding I'm giving the book is that I think the prose could have been structured a bit better, there are a bunch of "we'll cover this in {x} section" or "we covered this in {x} section" as you go through and I think you'd have needed less of that were the sections reorganized a bit to have a stronger "narrative" arc.