How Much Worse Could AI Censorship Be Than Social Media?

Marc Andreessen says censorship of AI has the potential to be a thousand times worse than social media censorship. Studies show the political orientation of the models, because questions get direct answers, and nine out of 10 of them are tremendously biased, with a handful that are not. He has been telling people in Washington that one threat is the government forcing startups into compliance, not only with trade rules but with their products: a censorship regime on AI exactly like the censorship regime on social media.

Social media is important, but at the end of the day it is just people talking to each other. AI is going to be the control layer on everything, including how kids learn in school, who gets loans, and whether a house opens at the front door. If AI gets wired into the political system the way the banks did and the way social media did, the result is a very bad future.

A big aim has been to keep AI from being wired into the political system that way. The Biden administration was explicitly on that path, and the direction was crystal clear. Another Biden administration would be even more emboldened, at 100 percent.