EFTA02345009.pdf

From: Joscha Bach Sent: Wednesday, August 20, 2014 7:48 AM To: Jeffrey Epstein Subject: Alex Wissner-Gross Attachments: signature.asc Dear Jeffrey, the recent AGI conference in Quebec was neat (I presented some =esearch sponsored by the Jeffrey Epstein foundation), and I was treated =o Alex' idea in the form of a keynote. (you sent this: =ttp://www.ted.com/talks/alex_wissner_gross_a_new_equation_for_intelligenc= ) The attempt of selling F = T V St as the new E = mc2=is amicably bold, but might be a bit contrived. He presents his "causal =ntropic force" it not only as a formula for intelligence, but also as a =aw for the development of the universe in general, as core ingredient =o an algorithm to beat the stock market etc. The main idea is not stupid: using maximization of future options is =omewhat similar to Marcus Hutter's idea of AIXI (a theoretically =ptimal intelligent agent with unlimited resources). In practice, it is =robably not what humans do; evidence suggests that people abhor =ncertainty and try to avoid branching expectation horizons, i.e. they =re willing to accept a smaller payoff if there are fewer possible =utcomes involved. However, people are not so smart, of course. In my view, Alex' theory has the more serious drawback of being not =alsifiable: if we find that an agent or a system does not maximize =uture options, we can always change the frame of reference until they =o. I imagine a chess playing program: you would not want to maximize =uture options, but go for the win in as few turns as possible. But =hen, winning the game might give you more options to advance in the =ournament later on. In practice, it can be used as a learning/decision making rule in some =ases; in others, there are better heuristics. In my view, it is not a =ilver bullet. Alex is way too smart to not know that a non-falsifiable theory spells =rouble, and he must have read the objections by Gary Marcus and others. = got a strong suspicion that he is deliberately selling snake oil. I hope that you are well! I certainly am. Cheers! Joscha 1 EFTA_R1_01314412 EFTA02345009