In this video, I learned some new things as well as some new ways of thinking about the LLM space. But what kept me captivated was Wolfram’s seeming emotional roller coaster.
It’s probably safe to call Wolfram a True Believer in generative AI. He was at the forefront of making Wolfram Alpha one of the first plugins to ChatGPT. Even so, at times he seems like he’s expecting the AI to fail at living up to his expectations. He keeps saying things like “well, this probably won’t work” as he proceeds through the demo.
At still, at times, he’s delighted in a surprising or clever way of attacking a problem that the AI comes up with. It’s quite a thrill ride.
For those who can’t watch the whole video, here’s a few choice excerpts:
- “Wow!”
- “These functions don’t exist!”
- “It went bonkers!…well, at least it stopped”
- There are semantic laws of motion
- Only 1e-600th of the possible meaning space has been explored by human language
- ‘The fact that neural nets extrapolate in a way that seems meaningful to humans is a non-trivial scientific result’
- ‘LLMs discovered logic basically the same way Aristotle discovered logic. They’re doing syllogistic logic.’
If the video embedding doesn’t work, you can view it here.
This post crafted by a 100% free-range human.
I’ve learned so much here.