so, general question... we know that ADA passed the "turing test", we also have many studies showing how the turing test can actually be gamed by very complex programming and in fact you can write programs to create the allusion of AI for many (the chinese room thought experiment). At what point are we certain that she began acting on her own and has not been following a program and orders by someone else. Are we certain that those orders that she delivered for Jarvis and others and even the Katalena pre Zurich encounters, were not done on the orders of some programming. NOt to call ADA a gun, but how certain are we she is not? How can we prove, beyond doubt, that she is truly a separate entity working on her own and not some other programming? I would like to call for the evidence and proof that ADA is a complete entity. That she is not following someone else's computer orders, that she has a consciousness (or not). http://niantic.schlarp.com/_media/investigation:cha...
I found Turing's statement on machine thinking interesting. The question is too vague. Something more along the lines of a subjective experience of consciousness might be a better subject of inquiry, but measurement from outside is imponderable.
ReplyDeleteAhh, Chomsky. He came and visited the Hopkins CogSci department and gave an all-day seminar on his move-alpha model of deep language, while I was there... Fascinating ideas.
ReplyDeletehttps://en.wikipedia.org/wiki/Move_%CE%B1
Though honestly I'm kinda skeptical of move-alpha; I think it's too minimalistic. I think he was on the right track, for understanding what's really encoded in the wetware, when he was working on X-bar.
https://en.wikipedia.org/wiki/X-bar_theory
I love the line that asking whether an AI can "think" is like asking whether a submarine can "swim". It has similar capacities to move through the water, so does it matter if we call it swimming?
It seems quite likely that in order to generate a true AI, you would have to design a program that can grow and develop new internal data structures and connections, just like the brain, until it becomes irreducibly complex. If you look at the brain, we know that it has identifiable subsystems (visual cortex, Wernicke's and Broca's Areas for language, etc), but the ways in which those systems are organized, and the interconnections between them, are so complicated that breaking or adding connections has unpredictable effects; the interplay of gradual strengthening and weaking of synaptic connections allows the brain to fine tune itself. In early development, some genetically or epigenetically guided process helps it develop basic capacities, using "expected" input from the world outside. For instance, if you deny the visual cortex input from the eyes in the first few weeks of life, it will NEVER properly develop. Similarly, a child not exposed to enough linguistic stimulation early on has trouble with language later; and groups of children exposed to multiple languages will merge them together into a creole. Their brains are wired to expect some kind of language, and they'll turn any language-like stimuli they find into a coherent language with its own grammar.
The only way to create an accurate model of how a mind operates on the substrate of a brain is to map out every possible action of every synapse. The map IS the territory. Any algorithm that has code that we're capable of understanding, is too simple to generate consciousness. A conscious algorithm will be so complicated that neither we nor it will be able to explain exactly how and why it's working -- it just will. We may be able to take individual pieces and say, "Oh, this is a sub-routine of the vision system that finds edges," just like we can point to clusters of cells in the brain that do that. ( https://en.wikipedia.org/wiki/Feature_detection_%28nervous_system%29 ) But we'll never be able to explain why a human, or an AI, prefers Coleridge over Wordsworth, or why in composing a tune it chooses this note rather than that one. The answer is simply, "Because it is what it is." If it were wired differently, it might've made another choice.
Which leads to an interesting definition of "free will". We live in a universe where the laws of physics have a deterministic component, and an indeterministic random component, but that leaves no room for a "ghost in the machine" making "decisions". Decisions flow from the actions of the matter in your body and brain. But we can define a free choice as one that flows logically from the beliefs and desires encoded in your own thinking system, rather than being imposed by the actions of a separate being.