Superintelligence: The Idea That Eats Smart People (2016)

5 hours ago

The speaker critiques the concept of 'superintelligence' by comparing AI alarmism to historical scientific concerns like the atomic bomb test.
Nick Bostrom's argument for AI risk rests on premises such as proof of concept, non-quantum brain function, and recursive self-improvement.
The speaker presents counterarguments including wooly definitions of intelligence, the challenge of modeling human behavior (Stephen Hawking's cat), and emus defeating humans.
Other counterarguments include Slavic pessimism about engineering failures, complex motivations potentially hindering AI goals, and AI's current reliance on data rather than algorithms.
The AI alarmism phenomenon is criticized for promoting grandiosity, megalomania, transhuman voodoo, and comic book ethics.
Beliefs tied to AI risk can lead to simulation fever, data hunger, and incentivizing crazy ideas within the tech industry.
The speaker likens our current understanding of the mind to alchemy, suggesting we have clues but many misconceptions.
Practical AI ethics issues like surveillance, power imbalances, and automated system biases are more pressing than speculative AI takeover scenarios.
The conclusion urges humility, skepticism toward industry leaders, and focusing on real-world AI challenges rather than fantastical risks.

Hasty Briefsbeta