Re:Eli Lilly CEO pushes back on Dario Amodei's ludicrous med...
by disgruntledphd2 on Friday November 13, 2026 @10:20AM
(#50484840)
> in some ways more intelligent than you are, it becomes very difficult to predict its outcomes
I think that this depends on your definition of intelligence. The models definitely have better working memory than us, probably worse inhibition behaviour (just look at the thinking tokens), and absolutely no ability to convert their working memory into short or long term memory.
These tools are super, super weird, and I think that applying human definitions for intelligence doesn't make much sense.
(as an aside, this is also my issue with super-intelligence, as it's incredibly ill-defined).
Re:Eli Lilly CEO pushes back on Dario Amodei's ludicrous med...
by 01100011 on Friday November 13, 2026 @10:35AM
(#50485042)
Yeah that’s one of my subtle points. AI is smarter by some definitions of intelligence but certainly not all. That has a lot of implications. It will do things we can’t and even do things we can’t conceive of yet. It also can’t do everything.
The weird thing is that when you create something that is in some ways more intelligent than you are, it becomes very difficult to predict its outcomes. We face a wall of uncertainty and it is impossible to see beyond it.