The only thing that worries me is that we don’t actually know how the human brain works. There’s an entire school of psychological theory and practice - really its oldest truly scientific branch, that holds that human intelligence actually does work a lot like an LLM. The hard core behaviorists believed that literally all human behavior was just a really complex version of Pavlov’s dogs. It sounds absurd, but they had good arguments. In principle even very complex behaviors can be the result of reinforcement and conditioning.

https://en.wikipedia.org/wiki/Behaviorism

The 25% more figure includes all of these things. How else do you think they would make the comparison?

IDK. I get really uneasy about claims that a computer or AI can never be intelligent or self-aware. Sure, it’s “just” circuits, but your brain is “just” cells passing information between each other. An individual cell is no more intelligent or self-aware than an individual transistor is. It’s deeply unscientific to believe there is some magic voodoo involved in biology that can’t be reproduced in a machine.

Eh. It’s par for the course. 20 years ago, at the height of the frenzy of outsourcing things to China, I remember saying that this will just result in US companies creating their own competition. Anyone with a brain could see that Chinese companies weren’t going to be willing to serve as second-fiddle to their US masters. The idea that you could keep design and management, while sending production overseas, and that you can keep that arrangement stable long-term? Pure fantasy. Of course a country isn’t going to be content just doing the grunt work. They want the highly paying design and management jobs, not just the menial labor ones.

The real issue is that since any fingerprint that can be mandated for AI content must be algorithmically implemented, then that fingerprint can be algorithmically removed.

For example, let’s say companies voluntarily choose or are forced to integrate text fingerprinting into LLM output. Automated AI writing detection tools already exist, but they’re not reliable. But in principle we could make the output of LLMs easy to identify. Maybe we force them to adopt subtle but highly unique patterns of word choice, punctuation, sentence structure, etc. Then if any student attempted to upload an LLM-generated essay to their course website, the system could with high accuracy flag it as AI generated.

But…if those patterns are so clear and unambiguous, it also means they can be easily detected by third party tools. If one person can code ChatGPT to add special fingerprinting to the text ChatGPT creates, another person can create a program that you can paste ChatGPT text into that will remove that fingerprinting.