Harvard University recently sponsored a worldwide weekend long AI "hackathon" to test out some security strategies being employed in an AI implementation. Yes, the title is a play on the book and film Ready Player One since CS50 is the introductory Computer Science undergraduate series at Harvard. The event was structured like a game with increasing difficulty in its levels.
Over 3,100 players out of 5,000 registrants participated and nearly all of them (2,945) solved all 7 security levels.
This post mortem analysis includes a statistical analysis of the event and discusses the strategies and effectiveness of the different layers of security as well as some of the unique (and sometimes humorous) strategies employed to defeat the AI's security at each level of the "game".
What is important to note is how the security programmers talk about the AI and its lack of understanding. The truth is, and experts in the field at all of the major AI creators have admitted, they're not exactly sure how it is doing some of what is doing so well.
