A Proposal for the Dartmouth Research Project on Artificial Intelligence (1955)
www-formal.stanford.edu
www-formal.stanford.edu
> 6. machine methods of forming abstractions from sensory and other data.
Still unsolved after 66 years, the problem indeed "needs more theoretical work".
Working on these at the same time with designing computer and compiler (1 & 2) probably was a big distraction.
Another distraction is introducing 5 & 7 ("self-improvement" and "randomness & creativity") before answering more fundamental questions of knowledge representation (3 & 6).
Kind of like building algorithms before figuring out the details of data structures.
Those influential 1950s seminars contributed to the current treatment of AI as a subset of CS instead of independent discipline with roots in cognition and neuroscience research.
but with all due respect you can’t really talk about artificial intelligence before understanding intelligence
http://mechanism.ucsd.edu/~bill/teaching/w07/philpsych/becht...
John Kemeny was recently in Dartmouth's math department at the time but AFAIK there was very little interaction with the early AI work.
Archived copy: https://web.archive.org/web/20210520010546/https://www-forma...