From: Ari Gesher To: Joscha Bach Cc: Sebastian Scun , Joi Ito , Kevin Slavin Greg Borenstein <[email protected]> Subject: Re: MDF Date: Thu, 24 Oct 2013 00:10:24 +0000 talcashi ikegami Martin Nowak , Jeffrey Epstein On Oct 23, 2013, at 8:09 AM, Joscha Bach <1 > wrote: That being said, AGI will have trouble succeeding because it is following the scruffy tradition. Perhaps the main failing of this tradition is its refusal to define objective (and preferably quantitative) measures of success. The question of good benchmark tasks is haunting Al since its inception. Usually, when we identify a task that requires intelligence in humans (playing chess or soccer or Jeopardy, driving a car etc.) we end up with a kind of very smart (chess-playing, car-driving) toaster. That being said, Al was always very fruitful in the sense that it arguably was the most productive and useful field of computer science, even if it fell short of its lofty goals. Without a commitment to understanding intelligence and mind itself, Al as a discipline may be doomed, because it will lose the cohesion and direction of a common goal. So now this gets interesting and starts to point us towards both MDF and the study of deception. The smart toasters emerge because they're being designed to solve well-bounded problems (like playing chess). There is no deception in chess (I would put feints in a different category), no hidden information, no adaptive adversary that can breach the bounds of the rules of the game. Given that, either brute force or ML/statistical approaches works well enough to build things like Deep Blue or the Google self-driving car. At Palantir (where I work), we have, to date, stayed away from heavy machine learning or algorithmic approaches to data analysis, focusing instead on building better and better tools to connect human minds to data in a way that's rigorous and interactive. This is the only current way to detect adaptive adversaries like so