Pages – AI Impacts

Surveys on fractional progress towards HLAI

How long until human-level performance, if we naively extrapolate progress since researchers joined their subfields?

2019 recent trends in Geekbench score per CPU price

From 2006 – 2020, Geekbench score per CPU price has grown by around 16% a year, for rates that would yield an order of magnitude over roughly 16 years. Details We looked at Geekbench 5,

Precedents for economic n-year doubling before 4n-year doubling

Does the economy ever double without having first doubled four times slower? Yes, but not since 3000BC.

Resolutions of mathematical conjectures over time

Conditioned on being remembered as a notable conjecture, the time-to-proof for a mathematical problem appears to be exponentially distributed with a half-life of about 100 years. However, these observations are likely to be distorted by

Trends in DRAM price per gigabyte

The price of a gigabyte of DRAM has fallen by about a factor of ten every 5 years from 1957 to 2020. Since 2010, the price has fallen much more slowly, at a rate that

Preliminary survey of prescient actions

In a 10-20 hour exploration, we did not find clear examples of ‘prescient actions’—specific efforts to address severe and complex problems decades ahead of time and in the absence of broader scientific concern, experience with

Interviews on plausibility of AI safety by default

This is a list of interviews on the plausibility of AI safety by default. Background AI Impacts conducted interviews with several thinkers on AI safety in 2019 as part of a project exploring arguments for

2019 recent trends in GPU price per FLOPS

We estimate that in recent years, GPU prices have fallen at rates that would yield an order of magnitude over roughly: 17 years for single-precision FLOPS 10 years for half-precision FLOPS 5 years for half-precision

Incomplete case studies of discontinuous progress

This is a list of potential cases of discontinuous technological progress that we have partially investigated. List In the course of investigating cases of potentially discontinuous technological progress, we have collected around fifty suggested instances

Effect of AlexNet on historic trends in image recognition

AlexNet did not represent a greater than 10-year discontinuity in fraction of images labeled incorrectly, or log or inverse of this error rate, relative to progress in the past two years of competition data. Details