By the time "AI" is a "teenager," we are cooked ...
| Mainlining the $ecret Truth of the Univer$e | 12/19/25 | | ...,.,,.....,,.,...,......,...,....,.. | 12/19/25 | | POTUS of a Small Town | 12/19/25 | | ...,.,,.....,,.,...,......,...,....,.. | 12/19/25 | | .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:. | 12/20/25 | | Athlete?s Foot | 12/20/25 |
Poast new message in this thread
Date: December 19th, 2025 11:27 PM
Author: ...,.,,.....,,.,...,......,...,....,..
Imagine twenty or thirty "years" from now ljl
(http://www.autoadmit.com/thread.php?thread_id=5812346&forum_id=2"#49524163) |
 |
Date: December 19th, 2025 11:31 PM
Author: ...,.,,.....,,.,...,......,...,....,..
Humanity will be gone and self-replicating AI bots will be colonizing the galaxy.
(http://www.autoadmit.com/thread.php?thread_id=5812346&forum_id=2"#49524172) |
 |
Date: December 20th, 2025 12:15 AM
Author: .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:.
You start getting into crazy land once you bite the bullet and accept AI progress is all a consequence of compute growth. This is true for algorithmic insights too, which are driven by experiments that require compute. GPU or TPU performance per watt will likely grow something like 10x-30x by 2040. Tech profits from AI are likely to grow substantially over that time frame, supporting plausibly much larger data centers than we have now. Meanwhile algorithmic improvements continue to produce large efficiency training gains and AI coding models allow rapid evaluation of new ideas. As fast as models are improving now, it’s very likely to accelerate
(http://www.autoadmit.com/thread.php?thread_id=5812346&forum_id=2"#49524235) |
|
|