lecture 1 Flashcards
externalizing a piece of cognition
Abacus
o Earliest thinking tool
o Made by Sumerians to outsource mental processes
o This way we extend our mind and make it more powerful
Mechanical Turk
First idea of machine being able to think
it was an illusion (dwarf in the machine)
analytical engine
Allowed you to write a computer program that would tell the machine to do mechanical operations
Was never built bc they did not see practical use of the machine
Ada lovelace
First person to actually write a specification to generate Bernoulli numbers
turing machine
a machine that can handle all computable functions
the original idea of the turing machine is a machine that draws/erases 1s and 0s based on algorithms
Boolean algebra
mathematical implementation of logic
proposition has a value of 1 = true, or 0 = false
computational theory of mind
thinking consists of computations
reasoning can be thought of as realized (not caused by) in the brain
this leads to the computer metaphor “mental processes are to the brain what software is to hardware”
turing test
when a compute cannot be distinguished from a human being, it passes the turing test and you have to conclude that the computer has consciousness
strong AI thesis
we will eventually be able to build intelligent, conscious computers
WILL be conscious
the Chinese room
John Searle
Argues against the strong AI thesis
Just executing the right computer program in itself cannot create consciousness
The turing machine does not and cannot think because it doesn’t know what it’s doing
syntaxis, not semantics
strong AI
weak AI
artificial general intelligence (AGI)
- Strong AI
AI WILL be conscious - Weak AI
AI is useful to study the mind - Artificial general intelligence
AI that is generic in its operation
2 types of responses to Searle
- There is something in the room (or the room itself) that has a rudimentary form of conscious thought
a. There is something about the system/elements of the system that has conscious thought - The argument shows that the turing test is not reliable: a system can pass the turing test without conscious thought
The psychology of the Chinese room
All commentaries focus on cognition, separated from affect/motivation/goals.
The consequence of that would be that if you were to build a computer with mental states and consciousness, you would have to find a way to make it feel things and motivate it.
Without affect, is thinking even possible?
o To ask the question ‘will computers ever be conscious?’ you have to ask the question ‘will computers ever have affect states?’.
AI winter
Period where nobody believes in AI anymore
late 1990s AI spring and summer
- computing power and memory had increased
- Much more data became available
By training neural nets on these large amounts of data, AI suddenly became reality