Four short links: 28 February 2018

Winning Atari, Explaining Algorithms, Code for America Summit, and Historic AI Book

By Nat Torkington
February 28, 2018
  1. Benchmarking Canonical Evolution Strategies for Playing Atari — the AI found two clever strategies for succeeding at Qbert: one is to play a level endlessly, where jumping off a level causes an enemy to follow but you get enough points from killing the enemy that you get another life; and in the other the agent discovers an in-game bug. First, it completes the first level and then starts to jump from platform to platform in what seems to be a random manner. For a reason unknown to us, the game does not advance to the second round but the platforms start to blink and the agent quickly gains a huge number of points. There’s video too. (via Brendan Dolan-Gavitt)
  2. A Case Study of Facebook’s ExplanationsOur experiments demonstrated that Facebook’s ad explanations are often incomplete and sometimes misleading, and that Facebook’s data explanations are incomplete and often vague. These findings have important implications for users, as they may lead them to incorrectly conclude how they were targeted with ads. Moreover, these findings also suggest that malicious advertisers may be able to obfuscate their true targeting attributes by hiding rare (and potentially sensitive) attributes by also selecting very common ones. This won’t be good with GDPR, which demands a right to an explanation in machine-driven decision-making. (via Arvind Narayanan)
  3. Learn faster. Dig deeper. See farther.

    Join the O'Reilly online learning platform. Get a free trial today and find answers on the fly, or master something new and useful.

    Learn more
  4. Code For America Summit — May 30 to June 1, in Oakland.
  5. Paradigms of AI Programming in Lisp — Peter Norvig posted the PDF and code of his 1992 book on AI programming.
Post topics: Four Short Links