Summarizing Superintelligence: Past developments and present capabilities (chapter 1)

  • Posted on: 30 October 2017
  • By: Juho Vaiste

The first chapter of the book takes a look on the economical growth of the human history, presents a brief history of AI research and development and offers AI researchers' estimations of the timeline for developing human-level and superlevel artificial intelligence.

Growth modes

The human history involves different growth modes in which the economical development has followed a really different pace. 

Prehistory - Agriculture - Industrial 

Bostrom argues that it is possible that there will be a next-level growth mode affiliated the development of intelligence level. However, he takes a bit distance from the idea of "exponential growth & singularity", presented by Kurzweil for example.

The history of AI

Darmouth - Expert systems - Neural networks (backpropagation) and genetic algorithms, a common mathematical framework  

Achievements of AI:
- Game-playing AI ("Ai has by now succeeded in doing essentially everything that requires "thinking" but has failed to do most of what people and animals do "without thinking" - that, somehow, is much harder!" (D. Knuth)
- AI effect
- Progress in "without thinking field" too: analyzing visual scenes, recognizing objects, controlling behaviour
- The challenge of common sense and nautral language understanding

The estimations of AI researchers

This might be the most quoted part of the book. It includes estimations of 
1) timeline for developing human-level machine intelligence
2) timeline for development from human-level to superintelligence
3) overall long-term impact of human-level machine intelligence

For example, this is the estimation table for the probability that the human-level machine intelligence has been developed by (estimation 1):