As we become more reliant on AI, algorithms and automation to manage the complexity of our world - we also face the challenge of systems that not aligned with our values or even our intentions.
For Brian Christian, this is not unlike the position of Mickey Mouse in the Disney classic, ‘The Sorcerer’s Apprentice’, who suddenly finds himself overwhelmed by magic beyond his control. In his words, ‘we conjure a force, autonomous but totally compliant, give it a set of instructions, then scramble like mad to stop it once we realize our instructions are imprecise or incomplete—lest we get, in some clever, horrible way, precisely what we asked for’.
Brian Christian is one of my favorite writers on AI, with a unique perspective that is very much at the intersection of computer science and philosophy. He is the author of the acclaimed bestsellers ‘The Most Human Human’ and ‘Algorithms To Live By’, which have been translated into nineteen languages. A visiting scholar at the University of California, Berkeley, he lives in San Francisco. His latest book, and the subject of this interview, is ‘The Alignment Problem’.
In this episode you will learn:
02:30 Brian’s origin story
05:01 Convergence: social sciences & tech
08:20 The alignment problem defined
14:12 AI Safety & the black box problem
18:28 Addressing the issue of bias
27:18 Business vs technical questions
37:06 Reinforcement learning
45:40 Challenging the system
52:45 Future skills