Bruce Schneier is a fellow and lecturer and Harvard Kennedy School, sit on the boeard of the EFF and is generally recognised as an expert on digitsl security (see his bio here). His blog, “Schneier on Security” is a must go to learn about and keep abreast of tech security.
He recently published a post on a topic that is largely unexplored yet, but of growing importance. AI (let’s call it “AI” for now even though the word covers a wide range of different realities) makes mistakes. Despite the hype and the corporate narrative, anyone who has ever used ChatGPT or other LLMs will have been faced with this stark fact. However, a question that begs to be asked is: how does it make mistakes, and how is this mistake-making process similar or different from the human mistake-making process?
Well, that’s exactly the question Schneier asks in this post. The comments are worth reading as well.