Skip to content

Lecture 9

Jump to bottom
chris wiggins edited this page Mar 19, 2018 · 6 revisions

readings

We’ll be following the second major thread born at Bletchley park; in addition to the thread from Bletchley to Bayesianism, there’s the thread from Bletchley to artificial intelligence. Specifically, we’ll read 4 documents, spanning approx 25 years, defining the first spring, summer, and winter of artificial intelligence.

1950: Turing’s ‘Computing Machinery and Intelligence’ defines the Turing test (“The Imitation Game”) and the program for artificial intelligence which the next half-century followed.

1955: The proposal documents for the 1956 ‘Dartmouth Summer Research Project on Artificial Intelligence’, which gathered many who would go on to define the field. Shannon, from this week’s reading, makes an appearance here. This document coined the term AI.

1973: The ‘Lighthill Report’ begins the first AI winter by casting doubt on the AI research program. Don Mitchie, who worked with Turing and Good and is mentioned in this week’s readings, also appears as a AI defender, along with McCarthy, who coined the term in the first place.

1974: McCarthy reflects on the 1973 debate, and what was learned about AI 1950-1974.

BONUS/OPTIONAL (but fun!!!)

Watch the original video of the debate itself from 1973!!! Available either in 6 youtubes: https://www.youtube.com/watch?v=yReDbeY7ZMU or as 1 long video: http://www.aiai.ed.ac.uk/events/lighthill1973/1973-BBC-Lighthill-Controversy.mov

Features comments from Strachey and Lighthill's accent

discussion

Turing (1950)

Important historical note on Turing Test:

  • As Presented: highly gendered (man or woman?)
  • Modern Conception: free of gender (human or not?)

Turing test is operational, freed of philosophical concerns and social constructs

Chinese Room Argument (CRA) Caveat:

  • Someone rotely doing the instructions cannot be said to be thinking, but passes Turing test!
  • Conclusion: Turing test isn't what you think it is

Was Turing prescient or has everyone just been following Turing since 1950?

Lots of modern ideas: Randomness in Computation, Chaos Theory, etc.

Dartmouth Report (1955)

Invention of term Artificial Intelligence

Surprisingly economical for birthing of field! ($13,200 in 1955 US Dollars < $200,000 in 2017 US Dollars) (cf.: $20 million for machine translation by 1960)

Lighthill and Friends (not) (1973)

Lighthill is Lucasian Professor of Mathematics (others: Isaac Newton, Stephen Hawking)

  • Trying to keep money with his friends!
  • Intimidated by amount of money in the US going to this new-fangled thing

His categorization is pretty terrible:

  • A: Advanced Automation
  • B: Building Robots
  • C: Computer-based CNS research (Neuroscience)

Lighthill arguing that people only getting results out of machine learning what (human intuition) they're putting into it

Dissecting field by claiming A and C from control theory and psychology whereas B was just bad

McCarthy Response: My field isn't within those! General problem planning and goal seeking Michie Response: Come see my robots, they do decently

Clone this wiki locally