
Separating Interesting vs Useful AI Trends

#16 - Responsible AI and Building Trust in AI - Liu Feng-Yuan
Topic 5 of 16
Published Nov 23, 2020 · Clip 20:15–23:42 (3:26)
Feng-Yuan urges separating what is interesting from what is useful in AI trends. Visual demos like reinforcement learning playing Atari or Dota rely on regenerate-and-learn dynamics that don't transfer to physical robots. He calls GPT-3's self-written articles cherry-picked Turing test red herrings, and argues the real challenge is making organizations genuinely data-driven rather than collecting point solutions.