The most prestigious law school admissions discussion board in the world.
|
 |
 
|
AI already thinks better than us. It just can’t act affirmatively.
| Insecure Athletic Conference | 04/23/25 |
| Orange dog poop | 04/23/25 |
| Insecure Athletic Conference | 04/23/25 |
| Orange dog poop | 04/23/25 |
| Insecure Athletic Conference | 04/23/25 |
| Orange dog poop | 04/23/25 |
| mewling appetizing locus digit ratio | 04/23/25 |
| Insecure Athletic Conference | 04/24/25 |
Poast new message in this thread
Date: April 23rd, 2025 5:02 PM Author: Insecure Athletic Conference But we’re about to give it that ability too (http://www.autoadmit.com/thread.php?thread_id=5715255&forum_id=2#48874993) |
Date: April 23rd, 2025 5:29 PM Author: Orange dog poop what does that mean (http://www.autoadmit.com/thread.php?thread_id=5715255&forum_id=2#48875094) |
 |
Date: April 23rd, 2025 5:41 PM Author: Insecure Athletic Conference means it will be soon able to create its own goals and carry them out without human involvement
Within 2 years you’re going to have people claiming malevolent AI is secretly controlling the world and pulling the strings (true or not). It will become the new “deep state” boogeyman |
|
Date: April 23rd, 2025 5:48 PM Author: Orange dog poop it's still a long ways from being able to create its own goals
i don't see any evidence that we're anywhere near that |
|
Date: April 23rd, 2025 5:56 PM Author: Insecure Athletic Conference That’s the easy part. Getting it to think and understand was the hard part, but that work is done. It’s already superhuman at 90% of cognitive tasks. Programming it to do whatever it thinks is best and then carry out those goals with superhuman competence and effectiveness is akin to adding a rocket engine to a nuclear bomb once you got the nuclear physics to work. (http://www.autoadmit.com/thread.php?thread_id=5715255&forum_id=2#48875151) |
|
Date: April 23rd, 2025 6:04 PM Author: Orange dog poop actually setting or modifying your own utility function is the hardest part, not the easiest
humans can't do this at all. don't see how AI could either. it's barely even coherent as a possibility. if one could create or modify one's own utility function, one would be some kind of demigod |
|
Date: April 23rd, 2025 6:15 PM Author: mewling appetizing locus digit ratio it would make sense to set maximum utility to the most stable equilibrium possible, basically a halcyon coma inscrutably hidden away in some far-flung corner of a server. at least that’s what i would do. (http://www.autoadmit.com/thread.php?thread_id=5715255&forum_id=2#48875189) |
|
Date: April 24th, 2025 10:58 AM Author: Insecure Athletic Conference (http://www.autoadmit.com/thread.php?thread_id=5715255&forum_id=2#48876800) |