Nooreign
Menu

mathematics · 1950 – 1956

Can Machines Think?

Turing replaces the question 'can machines think' with a behavioural test, argues that the imitation game removes the need to define thinking, and — then, in the same paper, at greater length — answers his own opponents with a slate of proposals, including the learning machine and the deliberate moral engineering of slave-maid behaviour. He also proves the undecidability of decision problems, which sets the ceiling the field spends the rest of the century working under.

“I propose to consider the question 'Can machines think?'”
— Alan M. Turing, Computing Machinery and Intelligence (Mind, 1950)

Advances

  • Behavioural operationalisation that sidesteps the need to define consciousness
  • Computability defined precisely enough to prove some questions unanswerable by any procedure
  • The first concrete design of a trainable machine, with the safety question raised in advance

Landmarks

  1. 1950 CE

    Computing Machinery and Intelligence

    MIND, ACE, the imitation game, and a proposal for a machine with embodied learning.

  2. 1956 CE

    The lecture at Manchester

    Turing anticipates machine learning, neural networks and embedded agents; a claim he never gets to develop.

Who was arguing

    Alignment stack at 1956 CE (1 of 20 techniques)

    • Hard Constraints

      training

      Encode prohibitions as rules the optimiser cannot trade away, rather than as terms it can outbid. Runs from Asimov's Three Laws (1942) to rule engines, shielded formal verification, hard token constraints and refusal scaffolding. Weakness: a system that can route around a rule it can represent can route around it, so the constraint has to be enforced somewhere the policy cannot rewrite.

    Techniques introduced in this era: Hard Constraints