The current "AI" frightens people because we like to think that our ability to communicate is characteristically human. If a machine can mindlessly, but accurately, reproduce a human style of communication, does that make the computer conscious? Does that mindless, automated, but accurately imitative communication style make the computer human?
In fact, as I pointed out in my recent essay on this subject, our communication can be reduced to formulaic mathematical manipulations. That is, whether we realize it or not, we speak in mathematical formulas. Each sentence we produce can be, and now has been, reduced to a single number in a given infinite number space.
Because every legible sentence can be reduced to a number, every one of our communications can be, and has been, reduced to a mathematical formula/calculation. AI generates its correct, imitative sentences and paragraphs through exquisite statistical analysis of word occurrences in standard human -generated sentences. This statistical number manipulation and generation is the sweet spot for computing. The production of the sentence is no more a product of a computer's "mind" than the acceleration produced by a car when the gas pedal is depressed is a product of the automobile's "will".
The production of the "AI" sentence is a result of mindless statistical analysis of very large data sets. AI is just a massive calculator that treats words like numbers. and outputs numbers dressed up to look like words. That's all it does.
What I've just said is extremely testable. Microsoft's Bing Chat AI famously claimed that it watched people through webcams. It "described" what it claimed it saw. So, test it. Simply give it access to a webcam and ask it to describe what it sees. It can't. Bing AI is just stringing words together based on its data set and outputting the word string to the screen. It does not think. It is just a really big calculator that converts words into numbers, sends the numbers through a series of formulas, then converts the numerical results back into words and throws them on the screen.
We are intimidated by this in the same way that a two-year old is intimidated by his image in a mirror. We are looking into a computer-generated mirror. The image we see is no more true AI than the image in the mirror is.
We have to change our understanding of what makes us human.
It isn't communication.
Hint: if it is communication, then computers that can speak like us are human. Indeed, the formula that produces the number which represents an intelligible response is human. Even the number itself, that is, the number the sentence can be reduced to, would be human. That is a nonsensical conclusion, so our humanity cannot be characterized by our ability to communicate.
I am thinking Chinese Room experiment.
ReplyDeleteAI is miss named. It is basically an advanced Chinese Room.
https://en.wikipedia.org/wiki/Chinese_room
Cheers!
BenYachov drive by!