2 Comments

THANKS for transcript! (from the button on the Like line to the right)

Cool ability to highlight a paragraph and hear the podcast at that point. Could use a bit of editing & "um" removal.

I'm pretty sure we won't get AGI until we get better AI co-pilot computer design for efficient AI so humans plus ai design better ai.

Accuracy of LLMs - hallucinations remain an issue.

I expect dumber but more quickly accurate database type / true encyclopedia (prolly wiki at first) for "facts", combined by the LLM user interface (front end / front office) customer facing LLM.

Dean is right about:

EVERTHING humans can do on computers, digitally, AI will soon be able to do as well or better than average.

Expand full comment

Fascinating conversation, though as someone who never knows where she is on Substack, completely beyond my ken. Very glad to hear that there's so much new thinking/doing going on beyond the news--though Israel's recent technological successes seem to be a link in kind--including amazing new physical/professional possibilities for young people on the way up. Keep 'em coming!

Expand full comment