Artificially Unintelligent Hysteria
Updated: Jun 26
The recent open letter proposing a 6 month "time out" on AI research seems to me one of the most ridiculous ideas I've ever heard. It doesn't surprise me that posers like Elon Musk support it. But I'm disappointed that people like Steve Wozniak have signed it. Wozniak doesn't get as much notoriety as Steve Jobs, but he was the co-founder of Apple and the technical genius that complimented Jobs' business genius.
The advances of ChatGPT are rather amazing. I've used it a couple of times to find information that I couldn't find using Google. But the idea that ChatGPT or any other form of AI is near sentience is laughable. In the "first wave" of AI we had very similar claims by very bright people such as Marvin Minsky who in 1967 said: “Within a generation … the problem of creating artificial intelligence will substantially be solved.”
This may seem a bit inconsistent given my previous post where I argued against Searle's famous Chinese Room attempt to show that "Strong AI" was an impossibility. However, these are two different questions. On the one hand, unlike Searle, I see no theoretical justification for claiming that Strong AI is prima facia impossible. On the other hand, I recognize that while we've made great strides lately due to Machine Learning such systems are nowhere close to achieving the kind of autonomy that even a mammal has, let alone a human. Noam Chomsky has several articles and videos where he explains the difference between how ChatGPT learns and uses language and human language acquisition and use. Several are behind paywalls on the NY Times and Fortune. Here is one that is freely available on boing boing.
My favorite guru when it comes to Machine Learning is Andrew Ng from Stanford. He is one of those rare people who has a deep understanding of the technology but also a solid understanding of business issues. If anyone is looking for a good introductory course on Machine Learning I highly recommend Ng's class on Coursera. He also recently had a talk on Youtube where he addressed the question of the AI timeout that I thought was very insightful.