Review of Empire of AI

Book Link: https://smile.amazon.com/dp/0593657500

I listened to this audiobook over many drives into work (admittedly at an AI-based company). It took me longer than usual, not because it wasn't interesting or compelling, but because it was pretty depressing overall. On the spectrum of AI evangelist to AI skeptic, I am much closer to the skeptical side. It clearly has value is specific applications, but applying it everywhere and using it as a replacement for human judgement, creativity, connection, and just general humanity is the absolutely wrong approach. Implementing and growing AI without accounting for privacy, security, control, and the harms it is causing is just pure recklessness. Empire of AI really drives home the level of reckless abandon Open AI used to grow its product with overwhelming strain on human labor, data, compute, land, energy, water, and institutional power. It also makes clear that the early days of Open AI as public-interest mission was never going to succeed when the likes of Altman, Musk, and co. are around.

In addition to all the powerful evidence and lesson of "powerful narratives must remain subordinate to evidance", this book gave me a huge metaphorical shove towards the Responsible AI frameworks and organizations. Ultimately it will require oversight from a government and/or international body of some sort to properly balance AI progress against the environmental, social, and economic harms it introduces. Something like, but not necessarily, the UN that will enforce responsibility and accountability policies through economic impact and oversight. It won't be perfect and will likely slow us down from potential progress, but that seems to be a good thing at this point and kind of the point.