We aren’t convinced by any of the arguments we’ve seen to expect large discontinuity in AI progress above the extremely low base rate for all technologies. However this topic is controversial, and many thinkers on
Computer performance per watt has probably doubled every 1.5 years between 1945 and 2000. Since then the trend slowed. By 2015, performance per watt appeared to be doubling every 2.5 years. Details In 2011 Jon
This page contains the data from Appendix 2 of William Nordhaus’ The progress of computing in usable formats. Notes This data was collected from Appendix 2 of The progress of computing, using Tabula (a program for turning
Tensor Processing Units (TPUs) perform around 1 GFLOPS/$, when purchased as cloud computing. Details In February 2018, Google Cloud Platform blog says their TPUs can perform up to 180 TFLOPS, and currently cost $6.50/hour. This
This page lists examples of AI systems producing solutions of an unexpected nature, whether due to goal misspecification or successful optimization. This list is highly incomplete. List CoastRunners’ burning boat Incomprehensible evolved logic gates AlphaGo’s
Published 7 Feb 2020 Altitude of objects attained by man-made means has seen six discontinuities of more than ten years of progress at previous rates since 1783, shown below. Year Height (m) Discontinuity (years) Entity
In April 2015, the lowest GFLOPS prices we could find were approximately $3/GFLOPS. However recent records of hardware performance from 2015 and earlier imply substantially lower prices, suggesting that something confusing has happened with these
We do not know how AGI will scale with marginal hardware. Several sources of evidence may shed light on this question. Contents DetailsBackgroundConsiderationsEvidence from existing algorithmsEvidence from human brain scalingEvidence from between-animal brain scalingEvidence from
We estimate that ‘human-level hardware’— hardware able to perform as many computations per second as a human brain, at a similar cost to a human brain—has a 30% chance of having already occurred, a 45%
There is modest evidence that people consistently forecast events later when asked the probability that the event occurs by a certain year, rather than the year in which a certain probability of the event will