Science

Experts Predict When A.I. Will Beat Humans In Everything

We had a good run.

by Gus Lubin
"Ex Machina"

After three years, A.I. will likely beat humans at Angry Birds, according to a major new survey of A.I. experts.

After four years, bots will likely be able to win the World Series of Poker.

After six years, they’ll likely fold laundry better than the median clothing store employee — and beat us at Starcraft.

And that’s just the start.

Researchers from Oxford, Yale, and the Future of Life Institute got 352 experts (who had published at recent A.I. conferences) to fill out surveys on the future of A.I. The results represent a major new perspective on a notoriously hard-to-predict subject. While there was a wide range of forecasts, the researchers highlighted when the median response put a 50 percent chance that A.I. would reach a milestone. After that point, the wisdom of this crowd suggests that A.I. dominance in a given task is likely — with the probability of dominance rising over time.

Looking ahead, after A.I. will likely be able to fake a Taylor Swift song (10 years) and work as a surgeon (37 years), we’ve got the point when unaided machines will accomplish every task better and more cheaply than human workers (45 years); when A.I. will match A.I. researchers (86 years); and when it will be able to take over all human jobs (123 years).

Here’s the timeline:

Again, this is a hard-to-predict subject. Below is a chart showing the wide range of estimates for “high-level machine intelligence,” with at least a few experts saying it’s not likely in the next century:

For a little more on what beating humans looks like in various categories, we’ve pulled some details and specific median 50 percent estimates from the appendix:

  • Angry Birds: Play new levels of Angry Birds better than the best human players. Angry Birds is a game where players try to efficiently destroy 2D block towers with a catapult. 3 years.
  • Answer Factoid Questions via Internet: Answer any “easily Googleable” factoid questions posed in natural language better than an expert on the relevant topic (with internet access), having found the answers on the internet. 7.2 years.
  • Transcribe Speech: Transcribe human speech with a variety of accents in a noisy environment as well as a typical human can. 7.8 years.
  • Translate (vs. amateur human): Perform translation about as good as a human who is fluent in both languages but unskilled at translation, for most types of text, and for most popular languages (including languages that are known to be difficult, like Czech, Chinese and Arabic). 8 years.
  • Telephone Banking Operator: Provide phone banking services as well as human operators can, without annoying customers more than humans. This includes many one-off tasks, such as helping to order a replacement bank card or clarifying how to use part of the bank website to a customer. 8.2 years.
  • Assemble any LEGO: Physically assemble any LEGO set given the pieces and instructions, using non- specialized robotics hardware. 8.4 years.
  • Read Text Aloud (text-to-speech): Take a written passage and output a recording that can’t be distinguished from a voice actor, by an expert listener. 9 years.
  • High School Essay: Write an essay for a high school history class that would receive high grades and pass plagiarism detectors. 9.6 years.
  • Answer Open-Ended Factual Questions via Internet: Answer any “easily Googleable” factual but open-ended question posed in natural language better than an expert on the relevant topic (with internet access), having found the answers on the internet. 9.8 years.
  • Answer Questions Without Definite Answers: Give good answers in natural language to factual questions posed in natural language for which there are no definite correct answers. 10 years.
  • Produce a Song Indistinguishable from One by a Specific Artist: Produce a song that is indistinguishable from a new song by a particular artist, e.g., a song that experienced listeners can’t distinguish from a new song by Taylor Swift. 10.8 years.
  • Generate Top 40 Pop Song: Compose a song that is good enough to reach the U.S. Top 40. The system should output the complete song as an audio file. 11.4 years.
  • 5km Race in City (bipedal robot vs. human): Beat the fastest human runners in a 5-kilometer race through city streets using a bipedal robot body. 11.8 years.
  • Translate New Language with “Rosetta Stone”: Translate a text written in a newly discovered language into English as well as a team of human experts, using a single other document in both languages (like a Rosetta stone). Suppose all of the words in the text can be found in the translated document, and that the language is a difficult one. 16.6 years.
  • Go (same training as human): Defeat the best Go players, training only on as many games as the best Go players have played. For reference, DeepMind’s AlphaGo has probably played a hundred million games of self-play, while Lee Sedol has probably played 50,000 games in his life. 17.6 years.
  • Write New York Times Best Seller: Write a novel or short story good enough to make it to the New York Times best seller list. 33 years.
  • Putnam Math Competition: Perform as well as the best human entrants in the Putnam competition — a math contest whose questions have known solutions, but which are difficult for the best young mathematicians. 33.8 years.
  • Math Research: Routinely and autonomously prove mathematical theorems that are publishable in top mathematics journals today, including generating the theorems to prove. 43.4 years.
Related Tags