Emotion Machine: Difference between revisions

From P2P Foundation
Jump to navigation Jump to search
No edit summary
m (Mbauwens moved page CSAIL. Marvin Minsky. Emotion Machine to Emotion Machine: no compound titles in this wiki)
(No difference)

Revision as of 08:35, 2 June 2019

CSAIL. Marvin Minsky. Emotion Machine
http://docs.google.com/Doc?docid=0ATsf-IaJ1pQxZGcyanpkZnRfMTgwZnNrZDVuZzU&hl=en
Read 'Emotion Machine' by Marvin Minsky
Watch the video linked below as you read the notes.
Notes are based on this video
http://mitworld.mit.edu/video/484

2.17. Why We Will Need Them

9.14. Students who hear that "GOFAI" failed should learn more about AI's history!

  • 1957 Arthur Samuel's machine plays master-level Checkers.
  • 1957 NEwell, Shadw and Simon. Proving logical theorems.
  • 1960 Herbert Gelernter: Proving theorems in Geometry
  • 1960 James Slagle: Symbolic Integral Calculus
  • 1963 Lawrence G. Roberts: 3-D Visual Perception
  • 1964 Thomas G. Evans: Solving Geometry Analogy problems
  • 1965 Daniel Bobrow: Solving word problems in Algebra
  • 1969 Engelmann, Martin and Moses: the MACSYMA project.
  • 1969 Minsky-Papert-Winston robot builds structures
  • 1970 Terry Winograd: A program understands sentences
  • 1972 Sussman: A program recognizes some bugs in programs
  • 1974 Eugene Charniak: A program understands some stories.

Computers and thought

Terry Winograd 1970 Controlling a Robot with Verbal Commands

"is the red block taller than the green block?" answered

Book on AI. Patrick Winston Jerry Sussman - Fixing of Bugs - 1972

29.40. Fads in AI

  • Reinforcement Learning = Deficient in Credit Assignment
  • Rule-Based Systems = Limited Representation
  • Neural Networks = Opague to Reflective Thinking
  • Statistical Inference = Fails to use Causal Reasoning
  • Formal Logic = Cannot exploit analogies
  • Genetic Programs = Fail to remember the causes of failures
  • Insect-like Robots = These projects just re-create previous ones.
  • Baby Machines = All have failed to develop high level thinking

But each method works only in certain domains. We need to know both when each method works and why it fails.

30 & 31. Jerry Sussman and Hal Ablenson

One program to use an AI working to solve the problem.

43.17. A "Critic-Selector" Theory of Brain

Suppose you fail to achieve a goal or your thinking gets stuck in some other way. Then your brain uses special machinery to (1) diagnose what went wrong and (2) activate some more appropriate Way to Think.

Recognize a Problem-Type >> Activate a "Way to Think"
CRITICS / SELECTORS

If a problem seems familier, try reasoning by Analogy. If it seems unfamilier, change how your describing it. If it still seems too difficult, divide it into several parts. If a problem is too complex, replace it by a simpler one. If your methods do not work, ask someone else for help.

51.23. Old vs. new views of emotion

Old view of Emotion: Emotional states augment our mind the way artists add colors to black and white drawings.

New view of Emotions : Each emotional state suppresses certain features of regular thinking.