Why AI is more challenging than we judge
[Submitted on 26 Apr 2021]
Summary: Since its starting up within the 1950s, the field of man-made intelligence has
cycled just a few instances between sessions of optimistic predictions and broad
investment (“AI spring”) and sessions of disappointment, loss of self perception, and
diminished funding (“AI wintry weather”). Even with these days’s seemingly swiftly slouch of AI
breakthroughs, the come of long-promised applied sciences such as
self-driving autos, housekeeping robots, and conversational companions has
turned out to be powerful more challenging than many people anticipated. One cause within the abet of these
repeating cycles is our restricted working out of the persona and complexity of
intelligence itself. On this paper I report four fallacies in licensed
assumptions made by AI researchers, which will consequence in overconfident predictions
relating to the field. I slay by discussing the commence questions spurred by these
fallacies, including the age-former downside of imbuing machines with humanlike
licensed sense.
Submission history
From: Melanie Mitchell [view email]
[v1]
Mon, 26 Apr 2021 20: 39: 18 UTC (23 KB)