Prof. Brian Cantwell Smith on Theory of Computation:
is it a thing? and is it interesting?
TL;DR: No, not a thing; but YES, interesting!
... my notes to follow.
is it a thing? and is it interesting?
TL;DR: No, not a thing; but YES, interesting!
... my notes to follow.
To think about AI, we relate the mind to computation.
Except we don't: we relate our theory of mind to our theory of computing.
Except we don't: we relate our theory of mind to our theory of computing.
At age 16, he set out to find a useful theory of computing. He thought it would take 6 weeks. 45 years later...
All our theories of computation are wrong.
All our theories of computation are wrong.
1. Computing as symbol manipulation.
Wrong because: If there's one thing 20th century logic taught us, it's that you cannot produce semantics with syntax.
Wrong because: If there's one thing 20th century logic taught us, it's that you cannot produce semantics with syntax.
2. Theory of Effective Computability
It's a mathematical theory of causality: as bumping and shoving.
Doesn't address semantics at all.
It's a mathematical theory of causality: as bumping and shoving.
Doesn't address semantics at all.
3. Information Processing
two unreconciled notions of information: semantic, and quantitative (thermodynamics). This theory works in the latter.
But computers straddle both: they don't just process information. They participate in our practices.
two unreconciled notions of information: semantic, and quantitative (thermodynamics). This theory works in the latter.
But computers straddle both: they don't just process information. They participate in our practices.
4. Digital philosophy
Discrete, with local determinism.
Digital implementations are super useful ... to build continuity on top of.
Discrete, with local determinism.
Digital implementations are super useful ... to build continuity on top of.
Computation is fundamentally an intentional phenomenon.
Our theories are too narrow.
Symbols are used to *specify* computations. They are not constituents of computations.
Our theories are too narrow.
Symbols are used to *specify* computations. They are not constituents of computations.
A theory of computing must
deal with the *interplay* between meaning and mechanism
and say how computers are special.
There isn't one. Computers aren't special.
deal with the *interplay* between meaning and mechanism
and say how computers are special.
There isn't one. Computers aren't special.
Computer Science is dealing with meaning⇄mechanism
at a level of complexity between the physical (physics, neuroscience)
and the social (anthropology, ethics).
For the first time, we have a laboratory to study and understand this space!
at a level of complexity between the physical (physics, neuroscience)
and the social (anthropology, ethics).
For the first time, we have a laboratory to study and understand this space!
