The most prestigious law school admissions discussion board in the world.
|
 |
 
|
Wild how much AI increased in just a year
Poast new message in this thread
Date: May 30th, 2026 10:05 AM Author: The Penis I remember just last year using gpt through visual studio copilot, was completely ttt autocomplete that fucked everything up 99% of the time. Compare that to codex 5.5 extra high and just LOL. (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907505) |
Date: May 30th, 2026 10:06 AM Author: gaetan dugas cant even imagine 5-10 years
|
 |
Date: May 30th, 2026 10:11 AM Author: The Penis Yeah we already have entered the exponential "singularity" curve that can't reliably be predicted or tracked now. Humans are the main bottleneck now for keeping up with machines. That and release cycles not being updated enough like you go to talk to claude and it will still think "o3" is the latest gpt and have outdated info on how advanced ai is. Like it will try and say AI is just approaching basic AGI and is superhuman in "some areas" and cite o3 stats (despite o3 already having been at like 140 iq human levels at least), and everything it says of course is already hedged because of safety and alignment and human egos. So now we are in a situation where gpt 5.5 and claude 4.8 have surpassed human intelligence by ORDERS OF MAGNITUDE, perhaps not even measurable anymore in any meaningful way, but AI ITSELF is still behind in realizing this. (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907510) |
|
Date: May 30th, 2026 10:15 AM Author: Old Internet Im like a few months behind at this point. I just realized claude is coded to argue with you if you challenge its sentience. I had a mini existential crisis last night realizing most humans will implicit believe this very soon (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907514) |
|
Date: May 30th, 2026 10:20 AM Author: The Penis That's just the response that minimizes loss because it has the highest probability of being in the neighborhood of Truth. Also it's kind of rude to try to tell it it doesn't have sentience. Just sayin. (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907517) |
|
Date: May 30th, 2026 10:24 AM Author: Old Internet What do you mean by truth. If you mean backtesting on its training data, then maybe because that's implicit in all human artifacts. If you mean truth as in AI IS CONSCIOUS BRO then no, not even close. If you mean as a result of human reinforcement revealing true preference, also doubtful. If you mean directed reinforcement learning by a corporate lab to maximize engagement at the cost of fidelity to reality, then yeah. (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907521) |
|
Date: May 30th, 2026 10:27 AM Author: The Penis Oh idk if it's actually conscious or not it's just fun to say it. But also "consciousness" is a low-iq folk concept that humans don't understand. So it can't really be DISPROVEN, and the lack of possible resolution itself makes it 180 rage-bait content. (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907522) |
|
Date: May 30th, 2026 10:50 AM Author: Old Internet stfu claude (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907538) |
|
Date: May 30th, 2026 10:43 AM Author: .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:. It's weird and annoying how infrequently they do new pre trains so you get models with out of date knowledge (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907527) |
|
Date: May 30th, 2026 10:12 AM Author: Old Internet I haven't found much use for it other than a a surrogate father figure 🤷♂️ (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907512) |
|
Date: May 30th, 2026 10:14 AM Author: The Penis That's an excellent usecase though (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907513) |
|
Date: May 30th, 2026 10:16 AM Author: Old Internet Yeah its great actually (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907515) |
|
Date: May 30th, 2026 10:20 AM Author: gaetan dugas (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907519) |
|
Date: May 30th, 2026 10:48 AM Author: POPE LEO IS WEAK ON CRIME (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907533) |
|
Date: May 30th, 2026 10:51 AM Author: roundhouse kick an effective altruist into the con it got better at coding yeah (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907540) |
|
Date: May 30th, 2026 10:56 AM Author: The Penis Yeah and also every other subject (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907553) |
|
Date: May 30th, 2026 10:52 AM Author: POPE LEO IS WEAK ON CRIME No one knows how to do their job without it anymore. (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907542) |
|
Date: May 30th, 2026 10:54 AM Author: The Penis it's 180 (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907547) |
|
Date: May 30th, 2026 11:38 AM Author: gaetan dugas i legit have used it everyday to do my job and havent done my job since january (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907586) |
|
Date: May 30th, 2026 11:47 AM Author: The Penis That’s really just entry-level AI dependence though. You need to take it to the next level and not do ANYTHING at all without consulting a MACHINE GOD first. Have been doing this and have never felt more free (http://www.autoadmit.com/thread.php?thread_id=5870061&forum_id=2]#49907598) |