I had a dinner conversation this evening with someone who is completely convinced that ChatGPT is going to become a fully intelligent entity sometime in the next 6 years.
Given that I know that the current path to machine learning cannot, intrinsically by its nature, lead to general intelligence, you can imagine that it was an awkward conversation.
I feel as though we are in a situation somewhat akin to the time way back when, at the dawn of the invention of photography, when people became convinced that artists would be put out of business by photographs.
I wonder how soon we will get out of this intermediate space, where people apply magical thinking to the wonderful, yet limited, capabilities of modern machine learning techniques.
At some point, these techniques will be just another tool that we will use in our everyday lives. But as long as people continue to believe that such things bring us into a world of Harry Potter or C-3PO, we still have some conversations to work through.
I tell fans of Generative ML techniques that the technology is Clever Hans, only with 100 million hooves.
I like that!
I have sympathy for those who believe these ML chatbots are or will become conscious entities. Alan Turing, Ray Kurzweil, and Westworld have all posited that “if you can’t tell, does it matter?”
(To be fair, Ray Kurzweil publicly denied that there was sentience in the recent Google flap.)
Whether these experiments can or will reach some arbitrary level of intelligence is largely irrelevant. What matters is the perception. I can’t prove this blog is written by a conscious, sentient human, but I treat it as such because it appears to be.
My brother asked ChatGPT to write a poem, and he sent me the result. I immediately wondered who holds the copyright in that poem. I can see lots of potential arguments for different answers. But I also imagine that kind of question may someday go to court, and it may be decided differently than the cases of painting elephants or selfie-snapping bonobos.
Then what happens to employment law? The factory worker who loses their job to a robot arm that mindlessly follows a script is laid off because the business no longer needs an employee in that role. But perhaps the advertising copywriter who loses their job to a chatbot that creates original material has been replaced by a different employee. Does the chatbot deserve other rights that employees enjoy? Can it join a union?
An ML tool that lays out circuit boards will always be perceived as a tool. But an ML tool that communicates in natural language and does useful work is going to be perceived by many as an intelligence. My prediction is that’s going to have all sorts of surprising repercussions on society regardless of whether experts can articulate why they are not intelligent in the sense that humans are.