The A.I. long-term view: hopes and concerns from an A.I. veteran
Episode title: Stuart Russell: Long-Term Future of AI (December 9, 2018)
Source : Artificial Intelligence | Lex | MIT AI
Guests: Lex Friedman, MIT researcher and host, Stuart Russell, Professor of Computer Science at UC Berkeley and best selling author of AI books
Lex Friedman interviewed Prof. Russell, one of the fathers of Artificial Intelligence with his seminal book. Prof. Rusell shares his concerns about AI and some recipes to avoid the ultimate disaster! AI taking over the world..
The 3 main points
AI should be regulated. Prof. Russel takes a strong view on the application of AI in technologies and applications that impact on global population like Facebook. We need the equivalent of Food and Drug Administration for AI & Machine Learning.
AI research community is in denial when it comes to risks related to AI super-intelligence. While a lot of effort and resources are devoted to the next breakthrough in AI and Deep Learning, very few think about how such innovations will impact society and culture.
Game theory might help AI to better interact with humans. By letting the system figuring out the solution, it can invent smarter behaviours that are more adapted to deal with humans.
The three main issues about AI today:
AI cannot always respond. This is a BIG issue in certain use cases such as Autonomous vehicles where if the AI system disengages, it might cause a fatal crash.
AI pursues goals that are not aligned with humans objectives. AI systems are machines built to optimise: we put in there some objective and then it optimises. ****(e.g. by maximising to the extreme the engagement on social media, AI algorithms propagates unhelpful content like fake news)
Data is not the new oil, it is “snake oil”. Amassing a pile of data without thinking about the problem you wan to solve or the higher level goal is just a waste of money, there's no value in it.
Check out the vintage but still highly impactful book of Prof Russel - Artificial Intelligence : A modern approach