The most prestigious law school admissions discussion board in the world.
|
 |
 
|
what's the best way to short this hyped-up AI bubble?
Poast new message in this thread
Date: June 7th, 2025 9:13 AM Author: Snowy Concupiscible Ape as a STEM engineer who has worked at an AI company and is now being forced to use AI at work, i see over-hyped trash technology that, due to a complete lack of how thinking actually works in the human mind, will never reach its assumed potential or even come close to it. of course, the hype can continue irrationally for a long time. what is the Cr way to short this obvious bubble? (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994320) |
Date: June 7th, 2025 9:18 AM Author: disrespectful roast beef national Short basically everything/anything. (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994327) |
|
Date: June 7th, 2025 9:19 AM Author: smoky crawly station karate Why does it need to understand how thinking works in the human mind? (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994329) |
 |
Date: June 7th, 2025 9:29 AM Author: Snowy Concupiscible Ape because the fundamentally heuristic nature and limited range of computational techniques of AI systems will always be prone to failure to the point where they will plateau in effectiveness, at a level far below what the human mind can do apart from a narrow subset of computationally-complex things that computers have always been better at (i.e., nothing new and nothing unique to AI) (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994343) |
|
Date: June 7th, 2025 2:06 PM Author: Walnut casino messiness There is good evidence that transformers often learn patterns that are limited in generalization and aren’t as robust as human mental representations and algorithms. The problem with shorting is that you are saying a whole bunch of other possible algorithms in the ML space will fail in the same way. Cortical uniformity is real - there are brain rewiring experiments that show perceptual input can be rerouted in the cortex and the brain will reorganize around it. The implication being that the cortex is running one learning algorithm in all regions, which it uses to organize and predict different sensory modalities.What makes you confident that evolutionary approaches, meta-learning and iterative refinement from researchers won’t find something similar? ML algorithms are under more optimization pressure every year due to more researchers, higher training budgets, falling compute costs, better libraries and understanding of training infrastructure, and now LLMs assisting with code development. (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994852) |
|
Date: June 7th, 2025 2:11 PM Author: burgundy mental disorder state Because human brains can learn and LLMs can't (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994858) |
|
Date: June 7th, 2025 2:19 PM Author: Walnut casino messiness LLMs not learning is an amusing claim. (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994863) |
|
Date: June 7th, 2025 3:09 PM Author: Snowy Concupiscible Ape you state "cortical uniformity is real" as if it's a fact and not a broad and poorly-defined idea that's been debated for decades (and continues to be). red flag #1.
you present perceptual rerouting experiments as if they're novel, which is not true. they've been experimentally verified since the mid 20th century (possibly earlier; i don't remember when the first vision alteration experiments were performed). red flag #2.
you list off a bunch of AI tools/methods, but they all have the same underlying set of limited computational tools that are in principle unable to recreate the underlying mechanisms of thought (because these mechanisms are unknown and will likely forever remain unknown due to them being outside the scope of human intellectual capacity). okay, i'm starting to smell something. what's that smell? hmm...
let's skip over the largely irrelevant optimization part and come to the end: "LLMs assisting with code development." LJL. you know what i smell? a liberal arts major who hasn't worked on a serious software development project and would likely fail out of a pre-calculus class.
admit it, your liberal arts degree turned you into an ignorant fool who fell for the AI hype. |
|
Date: June 7th, 2025 3:25 PM Author: Walnut casino messiness cortical uniformity isn't really debatable in a significant way. there may be slight architectural biases in certain regions that speed learning, but the rewiring experiments strongly imply they aren't essential. the equivalent of regional hyperparamters or inductive biases is also much less important when you are able to train on orders of magnitude more data than a human can see in a lifetime.
where did i present the perceptual rerouting experiments as if they are novel? i didn't imply that. i simply think they are a decisive piece of evidence. perhaps you should work on your reading comprehension skills.
" but they all have the same underlying set of limited computational tools that are in principle unable to recreate the underlying mechanisms of thought" This assertion is completely baseless and it is the sort of thing that someone would only assert if they were emotionally invested in the idea that humans are special, when there is now substantial empirical evidence that they aren't.
dipshits like you are about to succeeded by layers of matrix multiplications. you should probably try to make peace with that. |
|
Date: June 7th, 2025 3:27 PM Author: burgundy mental disorder state Ok but do you or do you not have a Liberal Arts degree? (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48995022) |
|
Date: June 7th, 2025 3:29 PM Author: Snowy Concupiscible Ape B.A. in political science, cum laude, directional state university confirmed (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48995027) |
|
Date: June 7th, 2025 3:48 PM Author: Snowy Concupiscible Ape okay, i'll temporarily lower my IQ and assume you think you know what you're talking about:
> "cortical uniformity isn't really debatable in a significant way"
literally a simple google search will show you research results from "prestigious" research scientists saying that cortical uniformity has been shown to be false.
>"where did i present the perceptual rerouting experiments as if they are novel?"
you imply this by referencing experiments that can narrow this down to the cortex, which is novel. but the general idea was shown to be true before you were born.
>"This assertion is completely baseless and it is the sort of thing that someone would only assert if they were emotionally invested in the idea that humans are special, when there is now substantial empirical evidence that they aren't."
If you think about it for a moment, you'll see that I'm stating an uncontroversial fact about intellectual capacities of all organisms: they're necessarily limited in scope if you paid attention in high school biology. In no way do I even hint that humans are special; if I were verbose, I would've said the same principles would apply to all organisms and not just humans, but again, that's being verbose.
>"dipshits like you are about to succeeded by layers of matrix multiplications. you should probably try to make peace with that."
Tell me about a time you delivered a new feature in a complex software system built over decades and used by hundreds of thousands of people daily. Oh wait, I see that POLITICAL SCIENCE DIPLOMA ON YOUR WALL |
|
Date: June 7th, 2025 9:52 AM Author: burgundy mental disorder state I'm not sure there's actually a way to short it but I'm not a "finance guy"
Like I'd say short NVIDIA but there's actually still going to be substantial demand going forward because even if AI progress continues to stall, they still have to build out compute in order to service user demand |
|
Date: June 7th, 2025 10:08 AM Author: Snowy Concupiscible Ape "Short NVIDIA" has been the obvious idea in my mind for over a year, but I haven't done it because it's impossible to tell if/when/how this bubble will pop (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994405) |
|
Date: June 7th, 2025 10:15 AM Author: burgundy mental disorder state I'm never shorting any big stocks again honestly
At this point the whole American economy is dependent on them going up forever and I don't think they'll ever be allowed to correct to reality |
|
Date: June 7th, 2025 10:45 AM Author: Snowy Concupiscible Ape Cr watching TSLA makes me extremely hesitant (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994487) |
|
Date: June 7th, 2025 10:47 AM Author: disrespectful roast beef national Shorting anything is kind of nuts. Maybe take out $3k/yr worth of Shorts for shits and giggles and then just deduct it on taxes when it inevitably loses money
If you have more than like $10k in cash you're kind of Shorting equities anyway atm and earning Interest. |
|
Date: June 7th, 2025 10:50 AM Author: Snowy Concupiscible Ape a few years ago i bought far-out Coinbase puts that, if i had the balls to hold them long enough, i would've made nearly half a million. but it was too risky and fucked with my head, so i sold for an insignificant gain (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994491) |
|
Date: June 7th, 2025 10:49 AM Author: aromatic slate travel guidebook Stonks have been disconnected from Value for a long time. Stop pretending that things still make sense or there’s justice in the world or something. (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994490) |
|
Date: June 7th, 2025 10:51 AM Author: Snowy Concupiscible Ape Cr my default strategy is to buy VTI and ignore the markets (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994492) |
|
Date: June 7th, 2025 2:18 PM Author: Amethyst well-lubricated death wish stain LLMs just can't spam gay shit like humans can (http://www.autoadmit.com/thread.php?thread_id=5734550&forum_id=2#48994862) |