The most prestigious law school admissions discussion board in the world.
|
 |
 
|
There is 0% possibility of “AGI” ever occurring
Poast new message in this thread
Date: October 18th, 2025 8:11 AM Author: Stimulating Old Irish Cottage AI’s greatest accomplishment will be making millions of otherwise sentient / marginally intelligent people retarded.
It’s just another fucking computer program and that’s all it will ever be. |
Date: October 18th, 2025 8:26 AM Author: Ruddy sadistic site i remember in 90s they pretended internet wld solve all probs theyd be no traffic anymore and shit techpigs are furking faggots (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49357411) |
|
Date: October 18th, 2025 11:49 AM Author: Primrose kitty cat mediation Nearly 100%. (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49357699) |
|
Date: October 18th, 2025 11:52 AM Author: big unholy scourge upon the earth electric furnace but it's definitely possible a future iteration will fool the masses into think it's sentient. that's already happened to some dimwits using chatgpt (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49357709) |
|
Date: October 18th, 2025 12:30 PM Author: federal faggot firefighter The only thing that would/could stop the eventual development of artificial intelligent minds is the breakdown and collapse of civilization
Which could totally happen |
|
Date: October 18th, 2025 12:34 PM Author: snowy high-end hall I don't see how far off this shit is. It's effectively The Same AGI atm. If it had a Body and Dexterity and if robots didn't suck, it could definitely do all kinds of irl shit. It's not exactly the same as humans but humans are not all they're cracked up to be, like at all. For the most part they just believe and regurgitate The Same Things over and over again. (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49357808) |
|
Date: October 18th, 2025 12:36 PM Author: Bateful Swollen Boistinker that's AGI though
make users insanely retarded
everyone uses AI to try not be retarded
agents do basic tasks that humans can no longer do |
|
Date: October 18th, 2025 10:30 PM Author: honey-headed comical mood twinkling uncleanness (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49358822) |
|
Date: October 18th, 2025 10:42 PM Author: slimy swashbuckling gaping chapel (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49358839) |
|
Date: October 18th, 2025 11:50 PM Author: Motley Flushed Box Office It doesn’t matter. Sentience is not even required to take over all humans jobs. (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49358898) |
|
Date: October 19th, 2025 12:32 AM Author: dark odious telephone stage Many humans don’t have sentience (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49358939) |
 |
Date: October 19th, 2025 12:33 AM Author: naked water buffalo travel guidebook (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49358940) |
|
Date: November 8th, 2025 12:32 PM Author: Charismatic splenetic theater foreskin (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49412455) |
|
Date: April 20th, 2026 2:30 AM Author: The Penis arguably most (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828651) |
|
Date: October 19th, 2025 12:35 AM Author: Abnormal toaster casino possibly the funniest outcome of all of this is humanity going extint to an AI that goes defunct right after (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49358941) |
|
Date: November 8th, 2025 12:27 PM Author: Trip point (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49412445) |
|
Date: April 20th, 2026 1:13 AM Author: kemp (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828595) |
|
Date: April 20th, 2026 1:31 AM Author: the walter white of this generation (walt jr.) It won't be taken seriously until it can do things in the physical world. You just can't respect an "intelligence" that can't move a coffee mug across a room. (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828610) |
|
Date: April 20th, 2026 1:40 AM Author: kike panopticon it's a bit like the breathless overselling that went on during the mid-20th century regarding technology, space travel, et cetera.
at the time, it felt like things were moving so rapidly that SURELY our computers and rockets would soon morph into robotic butlers and flying cars a la The Jetsons!
people deadass believed that the Moon 'landing' was an obvious auguring of interplanetary space travel and 'colonization' in the near-term. even in the 80s, they were still making sci-fi movies set in 'THE YEAR 1999' where people had flying cars and lived on Mars, like this was 100% plausible futurecasting.
the AI thing will fizzle, just like all tech has. we'll get all the low-hanging fruit benefits of AI, while 'AGI' talk quietly fades away. |
|
Date: April 20th, 2026 2:03 AM Author: the walter white of this generation (walt jr.) (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828641) |
|
Date: April 20th, 2026 2:42 AM Author: .,.,...,..,.,.,:,,:,...,:::,...,:,.,.:..:. This is generally a reasonable criticism of technological hype but the arguments for near term AGI are more complicated than “we made a lot of progress quickly, therefore we are near the end.” The success in modeling multiple different modalities using a highly generic architecture with minimal implementation differences tells you something meaningful about intelligence. Gradient descent with transformers works in multiple different domains because it’s a tractable approximation of Solomonoff induction. Small generalizing circuits make up more of the weight space and are easier for gradient descent to stumble on. It’s basically just a dumb circuit search process that tends to find programs from data that are likely to generalize. This becomes more true the more data you train on, with more epochs and with heavier regularization and more parameters. As compute budgets increase and training techniques become more efficient, this process converges closer and closer to optimal prediction of whatever modality you are training on. Even if we run out of ideas for how to improve the training techniques, the parametric circuit search can be increasingly applied to the learning algorithms themselves. There’s really no plausible obstacle that could stop this from happening relatively soon. Manifesting intelligence is no longer contingent on brilliant insights or ideas but FLOPS and engineering (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828660) |
|
Date: April 20th, 2026 2:47 AM Author: The Penis What definition are you using for "AGI" that current systems don't meet and you think might not even meet within the next year or so? I think autonomous operational competence might take another 1-3 years. But current systems already meet broad cognitive generality at the minimum, and arguably exceed most humans across a large fraction of intellectual tasks. (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828663) |
|
Date: April 20th, 2026 2:00 AM Author: The Penis Isn't "AGI" just supposed to be AI created thats as smart as humans? Current AI has already far surpassed the avg human, and on many tasks it has surpassed all humans. Most of the reasons for not calling it "AGI" seem to be sociological. A lot of it is definitional hedging too. If you want AGI to mean autonomous general agents that replace all human jobs or something like that then it might be a long time before we reach it. (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828639) |
|
Date: April 20th, 2026 2:23 AM Author: kemp I thought it was like when AI can autonomously improve and otherwise “run” itself and no longer needs any human inputs to exist and grow (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828648) |
|
Date: April 20th, 2026 2:29 AM Author: The Penis No that's way too strong and specific. It is supposed to just be general competence across a wide range of domains that is comparable to or superior to humans. To me it's kind of insane to say current systems haven't met that. (http://www.autoadmit.com/thread.php?thread_id=5787428&forum_id=2#49828650) |