When humans create real artificial intelligence, will it need sleep?
This was my thought just before I fell asleep. I was thinking about Lex Fridman’s podcasts I watched this weekend. I wanted to see people live, people I had only read about before, read their posts or articles about them or knew their work.
I watched Gary Marcus‘s speech on ML street talk channel. I gave him another chance but I’m keeping the same opinion as when I blocked his Twitter and Substack long time ago.
I like how he picks on LLM and GenAI mistakes but my problem is that he only does that. After criticism I want to hear a solution. He doesn’t have it. That’s why I consider his criticism as noise.
In his speech he stated that he almost hasn’t changed a thing in his slides from 3 years ago. Well that describes it all.
After Gary Marcus I also watched Yann Lecun, Elon Musk, Andrew Huberman. I’ve read an article about Huberman in Wired, so I could be biased about him. I think he mixed together meditation, sleep phases and task alterations (let’s call it that for short).
And that’s where I started thinking how I would explain the purpose of sleep. What an organism misses when it’s sleep deprived? Are dreams about repeating past experience and “archiving” them in the brain? How do we believe memory works?
What about different rhythms in sleep? Do they “reset” neurons? And thus, is it an important mechanism that is not even considered by computational biology?