We continue from Twitter and Elon Musk Part 1.
In 1950, the computers we know were science fiction. This did not stop mathematics, which tends to run 50 – 100 years ahead of actual need. So 40 years before Tim Berners-Lee invented the Web, pioneering computer scientist Alan Turing considered a hot problem of today: How do you detect a bot? His answer: Talk with it.
Since computer speech was neither anticipated or essential, the Turing Test involves the use of a teleprinter by a person to converse with two hidden entities, one computer, and a second person. If the hidden computer cannot be distinguished in conversation from the hidden person, the computer has passed the test; it is intelligent. This edges towards sentience, self-aware machine life. It may be the same thing.
As of 2022, though claims have been made, (BBC – Computer AI passes Turing test in ‘world first’), prevailing opinion is that no computer has passed the Turing test. Yet there is strong motivation to deny the event. (CNBC) Google fires software engineer who claimed its AI chatbot is sentient. Google’s quest for perfection, with a chatbot powered by neural networks, may be at a level where the distinction is fuzzy.
Noam Chomsky’s deep structure linguistic theory argues that Google’s LaMDA actually has to know things. And how much could it know without knowing about itself? Some AI researchers deprecate the Test, arguing it’s a test of deception, not intelligence. But Turing’s intellect was great. His work, along with that of John von Neumann, is the bedrock of everything we do. Is it wise to be so dismissive?
To solve Twitter’s bot problem without expensive human employees, Musk’s software interrogator must run the Turing test. It must embody the intelligence of Turing’s human. If it does, Turing implies it’s possible to thwart any bot that lacks equivalent computing resources. Something like LaMDA might fit the bill. But if the human brain has some secret sauce unavailable to artificial neural networks, then Musk’s LaMDA-esque bot interrogator lacks the unfoolable human advantage of Turing’s human tester — unless he can find the secret sauce.
The secret sauce is the possibility that human brain is a quantum computer. The speculation is that a quantum brain has as yet unknown advantages against any silicon you could throw at it. To summarize:
- If the human brain is not a quantum computer, an efficient artificial interrogator for Turing’s Test applied to bots is within reach.
- If the brain is a quantum computer, an artificial interrogator that fits Turing’s specification is not possible — unless Musk goes quantum himself. The technology for this does not exist.
The LaMDA chatbot and the Tesla autopilot use state-of-the art silicon neurons; neither use quantum computation. A Twitter interrogator based solely on silicon would be vulnerable to bots that share this technology. A bot war is implied. It might have entertainment value, like Saturday Night Wrestling.
To be continued shortly.