Hanson AI Expert Survey

In a small informal survey running since 2012, AI researchers generally estimated that their subfields have moved less than ten percent of the way to human-level intelligence. Only one (in the slowest moving subfield) observed acceleration.

This suggests on a simple extrapolation that reaching human-level capability across subfields will take over a century (in contrast with many other predictions).

Details

Robin Hanson has asked experts in various social contexts to estimate how far we’ve come in their own subfield of AI research in the last twenty years, compared to how far we have to go to reach human level abilities. His results are listed in Table 1. He points out that on an outside view calculation, this suggests at least a century until human-level AI.

Year added to list  Person Subfield Distance in 20y  Acceleration
 2012  A few UAI attendees  5-10%  ~0
 2012  Melanie Mitchell  Analogical reasoning  5%  ~0
 2012  Murray Shanahan  Knowledge representation  10%  ~0
 2013  Wendy Hall  Computer-assisted training  1%
 2013  Claire Cardie (and Peter Norvig agrees in ’14)  Natural language processing  20%
 2013  Boi Faltings (and Peter Norvig agrees in ’14)  Constraint satisfaction  Past human-level 20 years ago
 2014  Aaron Dollar  robotic grasping manipulation  <1%  positive
 2014  Peter Norvig *
 2014  Timothy Meese  early human vision processing  5%  negative
 2015  Francesca Rossi constraint reasoning  10%  negative
 2015  Margret Boden  no particular subfield  5%
 2015  David Kelley  big data analysis  5%  positive
 2016  Henry Kautz  constraint satisfaction  >100%
 2016  Henry Kautz  language  10%  positive
 2016  Jeff Legault  robotics  5%  positive
 2017  Thore Husfeldt  human-understandable explanation  <0.5%

Table 1 : Results from Robin Hanson’s informal survey

*Hanson’s summary of Peter Norvig’s response seems hard to fit into this framework:

After coming to a talk of mine, Peter Norvig told me that he agrees with both Claire Cardie and Boi Faltings, that on speech recognition and machine translation we’ve gone from not usable to usable in 20 years, though we still have far to go on deeper question answering, and for retrieving a fact or page that is relevant to a search query we’ve far surpassed human ability in recall and do pretty well on precision.


We welcome suggestions for this page or anything on the site via our feedback box, though will not address all of them.

1 Trackback / Pingback

  1. AI Impacts – The range of human intelligence

Comments are closed.