The most prestigious law school admissions discussion board in the world.
|
 |
 
|
"And then with enough CapEx, our LLM will turn into AGI." "Wow, that's amazing!"
Poast new message in this thread
Date: May 24th, 2026 8:23 PM Author: Emotionally + Physically Abusive Ex-Husband (oppose bitchbois) Retards invested million$ in "alchemy tech" back in the 1800s too. (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899057) |
Date: May 24th, 2026 8:24 PM Author: GoyPuff (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899060) |
|
Date: May 24th, 2026 8:27 PM Author: The Penis Yeah because noone got rich off of chemical manipulation tech in the 19th-20th centuries after Lavoisier killed phlogiston theory (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899064) |
|
Date: May 24th, 2026 9:42 PM Author: .,.,.;;,;.,;:,:,,:,.,:,::,..;.,:,.:;.:.,;.:.,:.::, that's not what the capex is for breh. u know that tho. (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899191) |
 |
Date: May 24th, 2026 10:44 PM Author: Emotionally + Physically Abusive Ex-Husband (oppose bitchbois) If "AGI" were anywhere on the horizon, they wouldn't be spending $1T to facilitate LLMs today.
It's all fucking flame, and LJL at anyone who doesn't get that by the midpoint of 2026. |
|
Date: May 24th, 2026 11:06 PM Author: The Penis "AGI" is a bullshit underspecified folk term with multiple fuzzy definitions and constantly moving goalposts. What would make you consider current AI to be "AGI" considering that it recently solved a problem that one of the most intelligent humans who ever lived formulated, couldn't solve, and whose own intuition ended up being wrong about it? (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899286) |
|
Date: May 25th, 2026 12:16 AM Author: Emotionally + Physically Abusive Ex-Husband (oppose bitchbois) What? No, this is a very well recognized concept: https://en.wikipedia.org/wiki/Artificial_general_intelligence
You could think of it as the ability to reason, rather than to just complete tasks in response to prompts. It would include things like having a world model, a theory of mind, etc. It's what literally every "AI" company professes to be trying to develop. |
|
Date: May 25th, 2026 12:22 AM Author: The Penis Wikipedia does make AGI a "recognized concept", but that isn't the same as having a precise operational definition. The page itself is even defining AGI broadly as a hypothetical AI that "matches or surpasses human capabilities across virtually all cognitive tasks" then it lists various traits, but notes at the end that there isn't a single agreed upon definition of "intelligence" as applied to humans or computers. In terms of AI companies, Sam Altman recently came out in an interview and said the term is "not super useful", and Dario Amodei (anthropic ceo) has also been quoted saying he doesn't know what "AGI" is and that it sounds like a BS marketing term. AI companies are continuing to distance themselves from it. (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899372) |
|
Date: May 25th, 2026 12:26 AM Author: Emotionally + Physically Abusive Ex-Husband (oppose bitchbois) I suppose we could be epistemological faggots if we wanted to, but it's close enough to say: the ability to reason, rather than to just complete tasks in response to prompts. It would include things like having a world model, a theory of mind, etc. Like the things that 8yo children have. (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899373) |
|
Date: May 25th, 2026 12:38 AM Author: The Penis Okay I suppose so. But also if you ask an 8 yo a question and they infer the answer and respond, they are also reasoning in response to a prompt. An 8yo child has a "world model" that includes Santa Claus and the belief that if they hide their eyes, you can't see them, sure. And "theory of mind"--the ability to attribute mental states to others--you use the term while failing to attribute a "mental state" to the thing that just completely liquidated one of the greatest human geniuses mathematical intuition. To some, I suppose "mental state" requires "feeling" like an 8 yo. (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899380) |
|
Date: May 24th, 2026 11:27 PM Author: .,.,.;;,;.,;:,:,,:,.,:,::,..;.,:,.:;.:.,;.:.,:.::, who exactly r u satirizing?
as penis says "agi" is a meaningless concept for nerds to jack off to in the anthropic bathroom
no one ever thought the capex was for anything other than code writing / drug discovery / 60 second animal cuck cartoons made by indians. that's far from flame and flame was never needed to sell people on the capex. |
|
Date: May 24th, 2026 9:43 PM Author: icanseemyneurons alchemy was 1600s (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899193) |
|
Date: May 24th, 2026 10:45 PM Author: tats STRchads cant stop winning (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899273) |
|
Date: May 24th, 2026 10:48 PM Author: butlerian jihadist https://x.com/AskYoshik/status/2057859030081745275 (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899277) |
|
Date: May 24th, 2026 11:13 PM Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... i'm not sure why anyone would currently be comfortable in believing that LLMs won't unlock all other cognitive capabilities. at the limit and with a sufficiently powerful function approximator, the next token training objective leads to a model developing all the verbal capabilities of humans. in order to predict text optimally, models need to develop increasingly sophisticated models of what people talk about and how. we have Mythos finding all sorts of new code vulnerabilities and GPT 5.5 is proving mathematical theorems that humans never solved. the METR time horizon graph is still bending upward on a linear scale and doubling every few months. models are increasingly capable of agentic autonomous AI research.
humans are just meat robots running connectionist learning algorithms. once you accept that fact, the increasingly general capabilities of modern AI from burning compute cycles is unsurprising and strongly implies near-term AGI. there's simply too much compute and data available, so even garbage learning algorithms with crude training objectives (next token training on trillions of random tokens) leads to increasingly powerful systems. |
|
Date: May 24th, 2026 11:18 PM Author: The Penis Already is AGI unless you make some insane definition of "AGI" where it has to have fully autonomous agency. Recent Erdos paper says right in the AI section "problem was solved in a completely automated fashion", so we have an autonomous system solving problems that were formulated and were unable to be solved by one of the highest iq prodigies that ever lived, plus the system can code, has more knowledge than any human could learn across several lifetimes, can go beyond maxing human iq tests to the point where the numbers are meaningless. What more do you need for "general intelligence? (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899296) |
|
Date: May 24th, 2026 11:29 PM Author: ,.,....,..,.,.,,,,..,..,.,..,.,.,.,... i think the models are really quite general and good anymore at directed, in-distribution tasks. a SWE task (even a very long one) that needs to have certain widely understood characteristics is something modern models will crush. the mathematical achievements aren't fully general. it's easy with RL and synthetic thought traces to train models on tons of data and make them good at certain types of math problems, but it seems like they are trying to make everything in-distribution by training on every bit of data they can. i think the current approach will still leave noticeable gaps in model capabilities. they aren't good at poking around at a problem and creating a new conceptual understanding of it. they can't do continual learning except for the weak sort within the context window. this is really important for achieving the more radical AGI visions and even constrains their capabilities in more mundane settings. i think the models are perfectly capable of automating large amounts white collar work if they can be flexibly adapted to a corporate context, but they can't. (http://www.autoadmit.com/thread.php?thread_id=5868844&forum_id=2�#49899317) |
|
Date: May 24th, 2026 11:47 PM Author: The Penis you are using “AGI” to mean a deployable autonomous worker or researcher with continual learning and organizational self-management. That is a legitimate engineering target, but not really the same as artificial general intelligence in the competence sense. Current models already display general cognitive transfer without satisfying that stronger autonomy criterion.
i think your "poking around and creating a new conceptual understanding" claim is most vulnerable here. If a model independently finds a construction that humans had not used to settle a longstanding conjecture, then by ordinary behavioral standards it has done something very close to “poking around” and “creating a new conceptual understanding.” Especially when it instantiates a new cross-domain solution path that no human ever saw.
maybe it didn't do it with humanlike curiosity, it didn't have the phenomenological experience of sitting with the problem, feeling the sense of strain that something was off, and didn't reach the felt moment of insight. Maybe it needed a specific prompt, and AI grading and flagging then later human verification, but I wouldn't consider those things to fall under the category of "intelligence".
another deeper issue is it seems you are smuggling in an origin criterion here. it seems you are saying that because the model was trained on vast data, RL, and synthetic thought traces that it's solution is "in-distribution", but that does not settle the question. humans are also trained on vast cultural data, examples, heuristics, and feedback. whether the system was trained doesn't seem to be the relevant question. the question should be whether it can generate valid, non-trivial solutions to real problems outside of memorized cases and it clearly can. i agree with you that it is still much weaker in terms of full research autonomy and economic subsitutability (where you can drop it in a firm and completely replace roles under deadlines etc.), but to me that is a different question from "general intelligence".
edit: actually I just thought of another angle on one of your points that I agree with now. the fact that humans can update their competence at any point, learn new things from lived interaction while current base models mostly can't change their weights during deployment, it needs "new training" probably for certain novel activities or for things the weights or guardrails bias against. I think that is a genuine aspect of "intelligence" humans have that these systems are lacking in, because we do say someone is more intelligent due to their ability to learn new things |
|
Date: May 25th, 2026 12:42 AM Author: peeface EPAH: "I produced an Auto Gynephilic Igbo for a fraction of that cost!"
|