I popped off on Twitter a while back while reading What Computers Still Can’t Do by Hubert Dreyfus, a treatise on the implausibility (or potentially even impossibility) of AGI. He had an argument that really hooked into my brain and gave me doubts about the feasibility of AGI for the first time.
I felt really confused, so I did some thinking aloud of my confusion on Twitter. The variety of feedback was pretty crazy, ranging from people calling me an idiot for buying into the argument, all the way to people calling me an idiot for feeling any doubt about the argument at all (gotta love the internet).
Mostly the feedback was really positive and constructive, though. People sent me a lot of material to look through and study to address my confusions and doubts. I felt that my confusions were pretty well resolved, so I went back to Twitter to give an update on my thinking. Sadly, it did not gain nearly as much traction.
I actually feel some damage was done by my first thread. A lot of people took it as a pretty persuasive argument that AGI is impossible, and I now think the argument fails. I’m trying to do what I can to rectify that. Some people asked for me to write up my thoughts more formally, so here it is.
I’ve summed up all my current thinking in two pieces over at the website BuiltIn:
There’s no paywall or anything, so please enjoy. :)
The brain is an analog computer that also managed to instantiate digital computation. https://arxiv.org/abs/2012.05965