\
  The most prestigious law school admissions discussion board in the world.
BackRefresh Options Favorite

Why is there such a weird societal complacency around AI

We now have LLMs that write and read significantly better th...
Concupiscible canary haunted graveyard
  03/08/24
computer programs will only be as smart as whomever programm...
Soul-stirring Milky International Law Enforcement Agency Resort
  03/08/24
Go find me the programmer who can beat Stockfish
Concupiscible canary haunted graveyard
  03/08/24
...
Razzle-dazzle coffee pot chapel
  03/14/24
Because this, like Neuralink, is the real thing "they&q...
Curious cruise ship
  03/08/24
make it all stop won't someone please make it all stop
Jet adventurous church alpha
  03/08/24
🫥
Curious cruise ship
  03/08/24
???
Jet adventurous church alpha
  03/09/24
Same friend just jokes hehe
Curious cruise ship
  03/09/24
around a decade ago, everybody thought cars would shortly be...
galvanic sickened home laser beams
  03/08/24
I hoap so
Swashbuckling Hateful Crotch
  03/08/24
Not at all persuaded this analogy works. Self driving doesn'...
Concupiscible canary haunted graveyard
  03/08/24
Common sense is exactly what they don’t have
Flushed startled new version parlour
  03/09/24
I can ask GPT-4 judgment questions about hypothetical situat...
Concupiscible canary haunted graveyard
  03/09/24
> It's apparent that it has some sort of world model tha...
galvanic sickened home laser beams
  03/09/24
Funny how these guys with strong predictions about AI taking...
Flushed startled new version parlour
  03/09/24
nobody has any idea how it actually works, which is why it's...
Self-absorbed Dark Newt Orchestra Pit
  03/09/24
at parties, i regularly explain, in five minutes, how chatgp...
Bull Headed Cobalt Meetinghouse
  03/09/24
Don’t bother with these idiots.
Flushed startled new version parlour
  03/10/24
that's only an "explanation" in the highest level ...
Self-absorbed Dark Newt Orchestra Pit
  03/10/24
...
exciting flickering scourge upon the earth
  03/11/24
...
Dashing shrine place of business
  03/11/24
...
massive buck-toothed area karate
  03/09/24
This argument is premised on the idea that the training meth...
Concupiscible canary haunted graveyard
  03/09/24
>LLMs have something similar to common sense please ne...
Bull Headed Cobalt Meetinghouse
  03/09/24
Most people have no real skills. Like your average Redditor ...
Grizzly misunderstood garrison
  03/08/24
i think ai is a long ways away from building staples cashier...
Soul-stirring Milky International Law Enforcement Agency Resort
  03/08/24
?
Grizzly misunderstood garrison
  03/08/24
isn't self-checkout basically a cashier? lmao
galvanic sickened home laser beams
  03/08/24
not really no.
Soul-stirring Milky International Law Enforcement Agency Resort
  03/08/24
People hate dealing with self checkout
Flushed startled new version parlour
  03/09/24
somebody has to be a cashier at staples; it's a noble calli...
ultramarine sneaky criminal
  03/08/24
There was an interesting write-up about McDonalds internal d...
Curious cruise ship
  03/08/24
they were gonna have robocops making hamburgers but then rea...
Soul-stirring Milky International Law Enforcement Agency Resort
  03/08/24
There were a variety of reasons cited against iirc A main o...
Curious cruise ship
  03/08/24
lmao thats sci-fi bs. they cant even keep theyre ice cream m...
Soul-stirring Milky International Law Enforcement Agency Resort
  03/08/24
https://www.kxan.com/news/texas/have-it-your-way-mcdonalds-f...
Curious cruise ship
  03/08/24
there should be a wall-mounted gatling gun
ultramarine sneaky criminal
  03/08/24
Terminator head on top of the golden arches barking expletiv...
Curious cruise ship
  03/08/24
I can't be the only one who thinks this sounds amazing. Woul...
Misanthropic passionate rehab selfie
  03/09/24
Same but only once
Curious cruise ship
  03/09/24
...
Henna Mexican Quadroon
  03/09/24
normies are just incredibly stupid AI could literally be ...
Dashing shrine place of business
  03/08/24
Totally agree. AI is going to fuck humanity so hard, and peo...
Fragrant spot sandwich
  03/08/24
It will create maybe 1 job for every 500 it replaces And th...
Curious cruise ship
  03/08/24
The lazy comparisons to the industrial revolution are annoyi...
exciting flickering scourge upon the earth
  03/09/24
We’re fat and complacent and degenerate and demoralize...
Floppy bat shit crazy native fanboi
  03/09/24
They still don’t do all that much that you couldn&rsqu...
pale gaping station
  03/09/24
...
massive buck-toothed area karate
  03/09/24
...
french gunner
  03/15/24
LLM's are not accessible to the general public so most treat...
bearded odious french chef
  03/09/24
how are they not accessible?
Grizzly misunderstood garrison
  03/09/24
ordinary people don't have access to them. It's not an easy-...
bearded odious french chef
  03/09/24
What are you talking about? Anyone can install an LLM on a c...
snowy location
  03/09/24
yep the actual answer is that normies are stupid and self...
Dashing shrine place of business
  03/09/24
> Anyone can install an LLM on a current gen PC with nvid...
Bull Headed Cobalt Meetinghouse
  03/09/24
cr, JH beat the machine right before he died
titillating house
  03/09/24
i tried to ask AI a math problem involving the area of a por...
galvanic sickened home laser beams
  03/09/24
no man don’t you understand this word picking gadget i...
Flushed startled new version parlour
  03/09/24
...
razzmatazz parlor prole
  03/09/24
I've asked it to compare two documents for duplicated lines ...
dun vibrant tank trailer park
  03/09/24
...
galvanic sickened home laser beams
  03/09/24
I’ve had it give me flatly wrong answers about popular...
Lascivious Property Dopamine
  03/09/24
...
galvanic sickened home laser beams
  03/09/24
You have to understand what goes into these. Gpt3.5 only goe...
Floppy bat shit crazy native fanboi
  03/09/24
It was GPT4 and I was asking about a crypto primitive that h...
Lascivious Property Dopamine
  03/09/24
It's been a long time since new, interesting porn hit the sc...
deranged cerebral abode police squad
  03/09/24
...
Glittery Hospital
  03/09/24
...
walnut roommate den
  03/09/24
let him cook
galvanic sickened home laser beams
  03/10/24
Wait til op finds out about 3D PRINTING and SELF DRIVING CAR...
Brindle Sick Shitlib Principal's Office
  03/09/24
...
razzmatazz parlor prole
  03/09/24
Obviously the timeline was way off, but those technologies a...
Mind-boggling tanning salon
  03/09/24
yep, and curing aging, space elevators and room temperature ...
galvanic sickened home laser beams
  03/09/24
...
bateful vengeful queen of the night piazza
  03/09/24
partially because the tech industry hypes shit up constantly...
Hot brunch background story
  03/09/24
The reason there's no "took are jobs" movement is ...
snowy location
  03/09/24
...
Dashing shrine place of business
  03/09/24
...
titillating house
  03/09/24
I think it’s just as big of an assumption to assume th...
Flushed startled new version parlour
  03/09/24
Cr but if it's indistinguishable from human consciousness, d...
Mind-boggling tanning salon
  03/09/24
I’m not sure I understand this. In what way would an L...
Flushed startled new version parlour
  03/09/24
I just assume there are 5 to 10 years left to earn money bey...
Misanthropic passionate rehab selfie
  03/09/24
we should just learn a trade like plumbing -- unless you thi...
titillating house
  03/09/24
This
Exhilarant slippery base
  03/14/24
Because it's overhyped horseshit for the most part.
massive buck-toothed area karate
  03/09/24
This. Every application of it is just a version of somethin...
Unhinged Gaming Laptop
  03/09/24
The instant content generation is incredible though. I've ha...
snowy location
  03/09/24
Jesus
Copper Cruel-hearted Toilet Seat Mood
  03/09/24
...
Copper Cruel-hearted Toilet Seat Mood
  03/09/24
because there's a weird societal complacency about everythin...
Talented sadistic sanctuary
  03/09/24
...
walnut roommate den
  03/09/24
>there is little reason for believing this complex reason...
Bull Headed Cobalt Meetinghouse
  03/09/24
...
blathering arousing market lettuce
  03/09/24
yes these retards believe that these LLMS are actually &ldqu...
Flushed startled new version parlour
  03/09/24
...
massive buck-toothed area karate
  03/10/24
what is an actual example of "complex reasoning" y...
Self-absorbed Dark Newt Orchestra Pit
  03/10/24
Sure signs of being an LLM critic retard: 1. Uses the &ld...
Concupiscible canary haunted graveyard
  03/10/24
...
pearly lay hominid
  03/10/24
I don’t get the hysteria. Having a significant share o...
cracking cocky organic girlfriend
  03/10/24
...
cracking cocky organic girlfriend
  03/10/24
Because it’s not going to do shit in any significant o...
Pungent Primrose Round Eye Center
  03/10/24
this post won't age well
Bistre theater telephone
  03/10/24
It’s such a terrible prediction.people think you are h...
Concupiscible canary haunted graveyard
  03/10/24
Just curious. How would you respond to this critique? &ld...
bearded odious french chef
  03/10/24
First of all, I would just like to note that Gary Marcus doe...
Concupiscible canary haunted graveyard
  03/11/24
was this written by GPT-3
self-centered big-titted antidepressant drug
  03/11/24
"THEY haven't made any breakthrough discoveries, unlike...
Copper Cruel-hearted Toilet Seat Mood
  03/15/24
I would respond that doing factually supported substantive r...
snowy location
  03/11/24
That's only because no AI can do it. You're right that 100% ...
Copper Cruel-hearted Toilet Seat Mood
  03/15/24
(((marcus))) is a joke. he seems to believe (1) AI dangerous...
Bistre theater telephone
  03/11/24
I think either he’s just trying to get attention as be...
exciting flickering scourge upon the earth
  03/11/24
This is absolutely cr and no one America seems to get it. Ou...
Copper Cruel-hearted Toilet Seat Mood
  03/15/24
That seems to be the case with a lot of academic fields. Eve...
Concupiscible canary haunted graveyard
  03/15/24
Here's another critique of AI or at least ChatGPT, that seem...
bearded odious french chef
  03/15/24
Definitely a better critique than most. I think there are pl...
Concupiscible canary haunted graveyard
  03/16/24
What's going to be interesting to see play out is what happe...
Self-absorbed Dark Newt Orchestra Pit
  03/16/24
Buy Black African Superman TOP-Class Permanence Tablet - Poerkan
Sooty locale
  03/10/24
It basically just a power tool. Like a nail gun for a builde...
Aphrodisiac Charcoal Senate Feces
  03/15/24


Poast new message in this thread



Reply Favorite

Date: March 8th, 2024 10:15 PM
Author: Concupiscible canary haunted graveyard

We now have LLMs that write and read significantly better than most people, write code, translate between every language instantaneously, understand images, etc. people treat this like a curiosity because they occasionally do silly things and can't be relied on for long, complex reasoning. Meanwhile these decent LLMs didn't exist as of a few years ago and there is little reason for believing this complex reasoning is far off. Seems like most people are telling themselves a story that these machines don't really "understand" as a way to cope with the reality of rapidly approaching human obsolescence.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477049)



Reply Favorite

Date: March 8th, 2024 10:16 PM
Author: Soul-stirring Milky International Law Enforcement Agency Resort

computer programs will only be as smart as whomever programmed them. good luck!

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477050)



Reply Favorite

Date: March 8th, 2024 10:23 PM
Author: Concupiscible canary haunted graveyard

Go find me the programmer who can beat Stockfish

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477060)



Reply Favorite

Date: March 14th, 2024 11:03 PM
Author: Razzle-dazzle coffee pot chapel



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47495415)



Reply Favorite

Date: March 8th, 2024 10:18 PM
Author: Curious cruise ship

Because this, like Neuralink, is the real thing "they" are trying to sneak past the goalposts while all these dumb fags are whinging about tranny bathrooms and other ephemera of the culture wars

This is civilization ending stuff that's mostly flying under the radar while they have us at libs' necks and vice versa

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477051)



Reply Favorite

Date: March 8th, 2024 10:19 PM
Author: Jet adventurous church alpha

make it all stop

won't someone please make it all stop

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477054)



Reply Favorite

Date: March 8th, 2024 10:32 PM
Author: Curious cruise ship

🫥

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477087)



Reply Favorite

Date: March 9th, 2024 10:19 PM
Author: Jet adventurous church alpha

???

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479429)



Reply Favorite

Date: March 9th, 2024 10:36 PM
Author: Curious cruise ship

Same friend just jokes hehe

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479448)



Reply Favorite

Date: March 8th, 2024 10:19 PM
Author: galvanic sickened home laser beams

around a decade ago, everybody thought cars would shortly be driving themselves because they could do some passable driving for a bit under certain conditions. in 2024, it's now clear that the passable driving under certain conditions is an indefinite developmental plateau. so to will it be for AI.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477052)



Reply Favorite

Date: March 8th, 2024 10:36 PM
Author: Swashbuckling Hateful Crotch

I hoap so

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477092)



Reply Favorite

Date: March 8th, 2024 10:38 PM
Author: Concupiscible canary haunted graveyard

Not at all persuaded this analogy works. Self driving doesn't work because hand coded rule sets inevitably break and supervised learning requires too much data to fill in all the edge cases. LLMs have something similar to common sense, which is how unfamiliar situations are actually dealt with by people. They already have fewer problems with unusual text input than most people.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477103)



Reply Favorite

Date: March 9th, 2024 1:56 AM
Author: Flushed startled new version parlour

Common sense is exactly what they don’t have

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477427)



Reply Favorite

Date: March 9th, 2024 2:06 AM
Author: Concupiscible canary haunted graveyard

I can ask GPT-4 judgment questions about hypothetical situations that are not going to be in its training corpus, and it will usually give me sensible answers. It's apparent that it has some sort of world model that it can use for making inferences. It would never be able to cope with the high dimensionality of text if it wasn't doing this and would quickly stray into nonsense.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477446)



Reply Favorite

Date: March 9th, 2024 9:14 AM
Author: galvanic sickened home laser beams

> It's apparent that it has some sort of world model that it can use for making inferences.

this is the faulty premise that is causing you to unreasonably panic. it doesn't have a "model" at all in the sense of a human explanatory model. it just turns out that you can get incredibly, shockingly, breathtakingly far in creating the appearance of having such a model by doing predict-the-next-most-likely-word math hijinks. nifty and interesting and a tool with useful applications, but just a tool

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477820)



Reply Favorite

Date: March 9th, 2024 10:00 AM
Author: Flushed startled new version parlour

Funny how these guys with strong predictions about AI taking over the world has no idea about how it works.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477892)



Reply Favorite

Date: March 9th, 2024 3:55 PM
Author: Self-absorbed Dark Newt Orchestra Pit

nobody has any idea how it actually works, which is why it's equally preposterous to just say there's no risk and it's incapable of replacing most human knowledge work in short order

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478537)



Reply Favorite

Date: March 9th, 2024 10:56 PM
Author: Bull Headed Cobalt Meetinghouse

at parties, i regularly explain, in five minutes, how chatgpt works

it is, at its root, matrix multiplication where the dimensions of the matrix are huge (hence why it's run on GPUs, which are the tits at matrix multiplication) and the result is predicting the next word in a sentence based on the foregoing words in the sentence plus the prompt plus the corpus of all text it has read

it doesn't reason and it doesn't understand; hell, it doesn't even apply grammar rules

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479499)



Reply Favorite

Date: March 10th, 2024 1:00 AM
Author: Flushed startled new version parlour

Don’t bother with these idiots.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479716)



Reply Favorite

Date: March 10th, 2024 2:32 PM
Author: Self-absorbed Dark Newt Orchestra Pit

that's only an "explanation" in the highest level sense. nobody in 2017 was predicting that five-six years later we'd see llms which can do something like "given this definition for the ambition card game [editors note: a card game which certainly does not have anything resembling an corpus of code implementations given it's played only by pensive and his cat], write an implementation in haskell" or generating wes anderson movies, and you sure as shit cannot give anything resembling an end-to-end explanation for how this is happening because nobody can. the people doing the primary research/development readily admit as such

more importantly nobody understands how the human mind actually works, which makes it all the more silly when people claim ai "is just a tool". what exactly makes us more than tools?

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47480780)



Reply Favorite

Date: March 11th, 2024 5:58 PM
Author: exciting flickering scourge upon the earth



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47484299)



Reply Favorite

Date: March 11th, 2024 6:12 PM
Author: Dashing shrine place of business



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47484337)



Reply Favorite

Date: March 9th, 2024 3:50 PM
Author: massive buck-toothed area karate



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478529)



Reply Favorite

Date: March 9th, 2024 9:40 PM
Author: Concupiscible canary haunted graveyard

This argument is premised on the idea that the training method reveals what comes out of it. Training on next token prediction pushes the model into initially learning superficial patterns, but to minimize loss on predicting arbitrary text the model would ideally be able to emulate the underlying reasoning processes and concept relationships. It is entirely reasonable if you optimize hard enough on orders of magnitudes more language than people see that the model learns the same models people use. We don’t know the internal mechanics of the end result, but clearly they learned some models. ChatGPT wouldn’t be able to play 1800 ELO chess if it didn’t learn a board representation. Pattern matching only goes so far if the input space is very large.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479326)



Reply Favorite

Date: March 9th, 2024 10:55 PM
Author: Bull Headed Cobalt Meetinghouse

>LLMs have something similar to common sense

please never speak on this subject ever again

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479498)



Reply Favorite

Date: March 8th, 2024 10:19 PM
Author: Grizzly misunderstood garrison

Most people have no real skills. Like your average Redditor is a cashier at Staples or something. They are worried and probably should be.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477053)



Reply Favorite

Date: March 8th, 2024 10:20 PM
Author: Soul-stirring Milky International Law Enforcement Agency Resort

i think ai is a long ways away from building staples cashiers lmao

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477055)



Reply Favorite

Date: March 8th, 2024 10:22 PM
Author: Grizzly misunderstood garrison

?

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477056)



Reply Favorite

Date: March 8th, 2024 10:23 PM
Author: galvanic sickened home laser beams

isn't self-checkout basically a cashier? lmao

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477061)



Reply Favorite

Date: March 8th, 2024 10:26 PM
Author: Soul-stirring Milky International Law Enforcement Agency Resort

not really no.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477064)



Reply Favorite

Date: March 9th, 2024 1:58 AM
Author: Flushed startled new version parlour

People hate dealing with self checkout

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477430)



Reply Favorite

Date: March 8th, 2024 10:24 PM
Author: ultramarine sneaky criminal

somebody has to be a cashier at staples; it's a noble calling

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477062)



Reply Favorite

Date: March 8th, 2024 10:36 PM
Author: Curious cruise ship

There was an interesting write-up about McDonalds internal decision making process re: their proposed single employee automated store that was linked here a while back

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477093)



Reply Favorite

Date: March 8th, 2024 10:38 PM
Author: Soul-stirring Milky International Law Enforcement Agency Resort

they were gonna have robocops making hamburgers but then realized building robocops is actually more work than making hamburgers?

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477100)



Reply Favorite

Date: March 8th, 2024 10:43 PM
Author: Curious cruise ship

There were a variety of reasons cited against iirc

A main one being "human redundancy" and this was considered to be unpalatable to the consumer market in the late 70s or early 80s or whenever. Cost-wise I think they found it was feasible for denser markets albeit with a slimmer menu

These days that same "human redundancy" is viewed strictly as a benefit due to WAAC powergaming bankers and inhuman SV technomancers setting the standards for this globalized cannibalistic race to the bottom we call American capitalism

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477113)



Reply Favorite

Date: March 8th, 2024 10:46 PM
Author: Soul-stirring Milky International Law Enforcement Agency Resort

lmao thats sci-fi bs. they cant even keep theyre ice cream machines running. what happens when a black guy comes in and smashes the chicken nugget machine for sport? theyve known for a long time they cant automate a mcdonalds.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477122)



Reply Favorite

Date: March 8th, 2024 10:50 PM
Author: Curious cruise ship

https://www.kxan.com/news/texas/have-it-your-way-mcdonalds-first-fully-automated-restaurant-with-no-human-contact-in-fort-worth/amp/

I guess they're at it again

It's not difficult to imagine and I'm not some restaurant mgmt mfe either. Minimize the points of contact for the customer and you don't have to worry about vandalism or whatever third world problems we have in polyglot New Babylon now. The machinery has been around for decades and afaik was rolled out in Asia before it was (considered) here

The only thing that was missing was mgmt will to ram it through PR-wise

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477133)



Reply Favorite

Date: March 8th, 2024 10:51 PM
Author: ultramarine sneaky criminal

there should be a wall-mounted gatling gun

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477134)



Reply Favorite

Date: March 8th, 2024 10:52 PM
Author: Curious cruise ship

Terminator head on top of the golden arches barking expletives at consumers

Every 100th customer gets a roman candle fired into their windshield

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477135)



Reply Favorite

Date: March 9th, 2024 9:59 AM
Author: Misanthropic passionate rehab selfie

I can't be the only one who thinks this sounds amazing. Would pay just for the spectacle of it all.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477890)



Reply Favorite

Date: March 9th, 2024 8:19 PM
Author: Curious cruise ship

Same but only once

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479168)



Reply Favorite

Date: March 9th, 2024 8:18 PM
Author: Henna Mexican Quadroon



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479166)



Reply Favorite

Date: March 8th, 2024 10:37 PM
Author: Dashing shrine place of business

normies are just incredibly stupid

AI could literally be mass exterminating humans with swarms of nanobot robots and normies would still be arguing with each other about the latest netflix episode of random goyslop or heckin sportsball game result or petty office gossip

they do not care about anything but themselves and their own lives. in fact they are not even physiologically able to care about anything else

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477094)



Reply Favorite

Date: March 8th, 2024 10:38 PM
Author: Fragrant spot sandwich

Totally agree. AI is going to fuck humanity so hard, and people are oblivious.

“It will create new jobs! Dur durr!”

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477102)



Reply Favorite

Date: March 8th, 2024 10:59 PM
Author: Curious cruise ship

It will create maybe 1 job for every 500 it replaces

And the New York Pravda will hail this as a great achievement because Biden created 16000 jobs this quarter (by liquidating 320000 lower salary ones)

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477144)



Reply Favorite

Date: March 9th, 2024 1:33 AM
Author: exciting flickering scourge upon the earth

The lazy comparisons to the industrial revolution are annoying. I think the idea of mass robotics and superhuman machines is an idea that breaks people's brains, so they reach for garbage analogies rather than thinking through the implications of what is being talked about. Even the new AI jobs like "prompt engineer" won't last long.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477409)



Reply Favorite

Date: March 9th, 2024 1:23 AM
Author: Floppy bat shit crazy native fanboi

We’re fat and complacent and degenerate and demoralized have completely lost all meaning in life and given up on it

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477401)



Reply Favorite

Date: March 9th, 2024 2:43 AM
Author: pale gaping station

They still don’t do all that much that you couldn’t use Google for before. It’s basically just faster search results at this point. And sometimes it’s wrong and totally stupid. So I imagine that is why

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477516)



Reply Favorite

Date: March 9th, 2024 3:51 PM
Author: massive buck-toothed area karate



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478531)



Reply Favorite

Date: March 15th, 2024 12:33 AM
Author: french gunner



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47495605)



Reply Favorite

Date: March 9th, 2024 9:34 AM
Author: bearded odious french chef

LLM's are not accessible to the general public so most treat it like any other news story about how in the future some xyz technology is going to change everything (some curiosity and skepticism but not much thought).

I am absolutely terrified. But I will fight on, like John Henry.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477848)



Reply Favorite

Date: March 9th, 2024 9:54 AM
Author: Grizzly misunderstood garrison

how are they not accessible?

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477878)



Reply Favorite

Date: March 9th, 2024 10:06 AM
Author: bearded odious french chef

ordinary people don't have access to them. It's not an easy-to-use consumer product. It's like pc's in the early 80's.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477905)



Reply Favorite

Date: March 9th, 2024 11:41 AM
Author: snowy location

What are you talking about? Anyone can install an LLM on a current gen PC with nvidia

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478130)



Reply Favorite

Date: March 9th, 2024 11:42 AM
Author: Dashing shrine place of business

yep

the actual answer is that normies are stupid and self-absorbed as fuck

see my poast above

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478134)



Reply Favorite

Date: March 9th, 2024 11:04 PM
Author: Bull Headed Cobalt Meetinghouse

> Anyone can install an LLM on a current gen PC with nvidia

dude, talk to a teacher at the high school level; they'll tell you their kids don't even know how to use a fucking keyboard anymore, and you want them to buy a specialty graphics card and install it in a PC? and that's "accessible" to you?

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479514)



Reply Favorite

Date: March 9th, 2024 8:50 PM
Author: titillating house

cr, JH beat the machine right before he died

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479209)



Reply Favorite

Date: March 9th, 2024 9:38 AM
Author: galvanic sickened home laser beams

i tried to ask AI a math problem involving the area of a portion of a circle. it kept trying to analyze the area as a quarter of a circle or a half circle despite that clearly not being the case based on the image and coordinates provided. it was frustratingly retarded. this was chatgpt 4. if it struggles like a retard to estimate the area of a portion of a circle (that could easily be approximated by adding the areas of a few standard geometric shapes together), despite repeated attempts at feedback and correction, then i'm not worried about it taking all our jobs. it's a bit like a retarded savant with no common sense that's incredibly good at bullshitting.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477851)



Reply Favorite

Date: March 9th, 2024 10:10 AM
Author: Flushed startled new version parlour

no man don’t you understand this word picking gadget is going to kill us all

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477918)



Reply Favorite

Date: March 9th, 2024 12:32 PM
Author: razzmatazz parlor prole



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478170)



Reply Favorite

Date: March 9th, 2024 10:21 AM
Author: dun vibrant tank trailer park

I've asked it to compare two documents for duplicated lines and it can't do it. Then when I inform it I found one it apologizes, notes it, re-analyzes it, then misses the rest. Sometimes it's cool but for anything requiring actual detail it sucks ass.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477945)



Reply Favorite

Date: March 9th, 2024 10:33 AM
Author: galvanic sickened home laser beams



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477976)



Reply Favorite

Date: March 9th, 2024 10:42 AM
Author: Lascivious Property Dopamine

I’ve had it give me flatly wrong answers about popular and publicly available python APIs. I’ve also had it give me a lot of answers that saved me hours of work.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477990)



Reply Favorite

Date: March 9th, 2024 10:43 AM
Author: galvanic sickened home laser beams



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47477994)



Reply Favorite

Date: March 9th, 2024 2:08 PM
Author: Floppy bat shit crazy native fanboi

You have to understand what goes into these. Gpt3.5 only goes to like September 2021 knowledge cutoff and gpt4 to April 2023. Any newer python libraries or versions yeah you’re gonna get bad code

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478398)



Reply Favorite

Date: March 9th, 2024 3:36 PM
Author: Lascivious Property Dopamine

It was GPT4 and I was asking about a crypto primitive that has had a consistent interface for at least 5 years. It’s still an awesome resource, but I’ve had it flop bigly on a number of fundamental things.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478489)



Reply Favorite

Date: March 9th, 2024 12:01 PM
Author: deranged cerebral abode police squad

It's been a long time since new, interesting porn hit the scene. Might want to let shit play out a little more.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478144)



Reply Favorite

Date: March 9th, 2024 8:38 PM
Author: Glittery Hospital



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479184)



Reply Favorite

Date: March 9th, 2024 8:46 PM
Author: walnut roommate den



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479198)



Reply Favorite

Date: March 10th, 2024 10:09 PM
Author: galvanic sickened home laser beams

let him cook

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47481965)



Reply Favorite

Date: March 9th, 2024 12:26 PM
Author: Brindle Sick Shitlib Principal's Office

Wait til op finds out about 3D PRINTING and SELF DRIVING CARS

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478160)



Reply Favorite

Date: March 9th, 2024 12:33 PM
Author: razzmatazz parlor prole



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478174)



Reply Favorite

Date: March 9th, 2024 9:59 PM
Author: Mind-boggling tanning salon

Obviously the timeline was way off, but those technologies are 100% going to be a big deal at some point. 3D printing is making huge advances in terms of prototyping, medicine, and aerospace. The whole "every home will have a 3D printer that spits out whatever you want!" thing was something believed only by trekkies and retarded journalists. Same with the idea that additive will displace traditional subtractive manufacturing anytime soon, if ever.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479380)



Reply Favorite

Date: March 9th, 2024 10:39 PM
Author: galvanic sickened home laser beams

yep, and curing aging, space elevators and room temperature superconductors will be "a big deal at some point" too.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479458)



Reply Favorite

Date: March 9th, 2024 10:41 PM
Author: bateful vengeful queen of the night piazza



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479464)



Reply Favorite

Date: March 9th, 2024 12:40 PM
Author: Hot brunch background story

partially because the tech industry hypes shit up constantly and they're currently doing it for AI even when it obviously has some real massive potential to be a game-changer. there will be a ton of retarded attempts to cash in on AI that fail very publicly, and that's true even if it winds up being as big a deal as the transhumanists expect in the long term

partially because there's a large chunk of humans that just take for granted that true AGI is impossible. the most honest of these people is the 90 IQ religious normie who will tell you that humans have souls and machines don't, so obviously there's no real threat from AI becoming smart. the smarter and more subtle version of this is people who in principle accept metaphysical naturalism, reductionism, etc but still insist there's some magic spark inside human brains that will take centuries for science to unravel, if ever. i can't prove they are wrong but they are very obviously engaging in motivated reasoning with this shit.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478200)



Reply Favorite

Date: March 9th, 2024 12:43 PM
Author: snowy location

The reason there's no "took are jobs" movement is because most people don't actually do any real work at their jobs anymore, unlike the machine and factory laborers who normally protest this kind of thing (they're all in Asia and lat America now) and they can make $ on weird side hustles like eBay, Uber and onlyfans etc

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478208)



Reply Favorite

Date: March 9th, 2024 2:10 PM
Author: Dashing shrine place of business



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478403)



Reply Favorite

Date: March 9th, 2024 8:53 PM
Author: titillating house



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479215)



Reply Favorite

Date: March 9th, 2024 9:31 PM
Author: Flushed startled new version parlour

I think it’s just as big of an assumption to assume that consciousness can arise on the strict substrate of a microprocessor when life as we know it operates in a very different physical construct.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479297)



Reply Favorite

Date: March 9th, 2024 9:51 PM
Author: Mind-boggling tanning salon

Cr but if it's indistinguishable from human consciousness, does it even matter that it doesn't meet our definition?

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479346)



Reply Favorite

Date: March 9th, 2024 11:17 PM
Author: Flushed startled new version parlour

I’m not sure I understand this. In what way would an LLM be “indistinguishable “ from human conscious ? It’s impossible to compare consciousness because we only experience our own mind. You are changing the subject from whether the leap can happen to how we can determine whether it has happened.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479547)



Reply Favorite

Date: March 9th, 2024 3:40 PM
Author: Misanthropic passionate rehab selfie

I just assume there are 5 to 10 years left to earn money beyond most of us being made redundant. If I'm wrong, I'll still be reasonably wealthy and retired. Seems like a good, safe bet.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47478502)



Reply Favorite

Date: March 9th, 2024 9:07 PM
Author: titillating house

we should just learn a trade like plumbing -- unless you think the boston dynamics style mobile robots will be ready by then

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479245)



Reply Favorite

Date: March 14th, 2024 11:17 PM
Author: Exhilarant slippery base

This

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47495447)



Reply Favorite

Date: March 9th, 2024 8:44 PM
Author: massive buck-toothed area karate

Because it's overhyped horseshit for the most part.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479193)



Reply Favorite

Date: March 9th, 2024 9:00 PM
Author: Unhinged Gaming Laptop

This. Every application of it is just a version of something we've been using for 15 years, but now built for dummies who don't know how to think for themselves. ChatGPT is just a better version of Google or at least better for people who don't know what they are querying. Document editing software is an improved version of spellcheck, which if left alone will make errors.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479232)



Reply Favorite

Date: March 9th, 2024 9:16 PM
Author: snowy location

The instant content generation is incredible though. I've had it draft annoying client emails, job postings, and a script for a YouTube ad that's actually pretty good.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479261)



Reply Favorite

Date: March 9th, 2024 10:44 PM
Author: Copper Cruel-hearted Toilet Seat Mood

Jesus

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479474)



Reply Favorite

Date: March 9th, 2024 10:43 PM
Author: Copper Cruel-hearted Toilet Seat Mood



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479469)



Reply Favorite

Date: March 9th, 2024 9:40 PM
Author: Talented sadistic sanctuary

because there's a weird societal complacency about everything these days

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479328)



Reply Favorite

Date: March 9th, 2024 10:55 PM
Author: walnut roommate den



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479497)



Reply Favorite

Date: March 9th, 2024 10:54 PM
Author: Bull Headed Cobalt Meetinghouse

>there is little reason for believing this complex reasoning is far off

There's actually huge reasons for believing this complex reasoning is so far off it's never gonna happen. xo is just retarded like all the foomers

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479496)



Reply Favorite

Date: March 9th, 2024 11:06 PM
Author: blathering arousing market lettuce



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479518)



Reply Favorite

Date: March 9th, 2024 11:15 PM
Author: Flushed startled new version parlour

yes these retards believe that these LLMS are actually “thinking” when they are just leveraging statistical models with mega powerful processors.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479541)



Reply Favorite

Date: March 10th, 2024 9:58 AM
Author: massive buck-toothed area karate



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479999)



Reply Favorite

Date: March 10th, 2024 2:41 PM
Author: Self-absorbed Dark Newt Orchestra Pit

what is an actual example of "complex reasoning" you think is far off in LLMs that humans are capable of

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47480797)



Reply Favorite

Date: March 10th, 2024 4:18 PM
Author: Concupiscible canary haunted graveyard

Sure signs of being an LLM critic retard:

1. Uses the “it’s just a next token predictor” argument. Ignores what optimal token prediction looks like at the limit. Also says they are just matrix multiplications with non-linearities. Like this says anything meaningful about whether they can emulate biological neural computation. You could be a reductionist idiot with brains too if you didn’t know what they did.

2. Never sets fixed goal posts for actual cognition or reasoning. Look how quickly people who had all sorts of handy examples of GPT-3 failures as proof of this ignored the fixes with GPT-4 and started complaining about something else.

3. Highlights LLM failures while ignoring their flexibility in comparison to most humans. I would trust GPT-4 more to get an SAT question right (not from its training corpus) than a typical human. People are retarded in all sorts of ways LLMs aren’t. Maybe they can’t reason either (I actually like this argument in a way - most people seem like limited pattern matchers who can only make basic inferences from their limited world model. But I think deep sort of reasoning can’t be far away considering intelligent people have the same basic biological substrate with a few genetic differences)

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47481128)



Reply Favorite

Date: March 10th, 2024 4:54 PM
Author: pearly lay hominid



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47481202)



Reply Favorite

Date: March 10th, 2024 12:09 AM
Author: cracking cocky organic girlfriend

I don’t get the hysteria. Having a significant share of people be “knowledge workers” is a very recent phenomenon, the last 50 years or so, and will have just been a blip in the long run. Most humans will do labor etc just like 99% of history.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479625)



Reply Favorite

Date: March 10th, 2024 9:53 AM
Author: cracking cocky organic girlfriend



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47479997)



Reply Favorite

Date: March 10th, 2024 2:37 PM
Author: Pungent Primrose Round Eye Center

Because it’s not going to do shit in any significant or lasting way

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47480788)



Reply Favorite

Date: March 10th, 2024 2:43 PM
Author: Bistre theater telephone

this post won't age well

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47480803)



Reply Favorite

Date: March 10th, 2024 8:22 PM
Author: Concupiscible canary haunted graveyard

It’s such a terrible prediction.people think you are hyper for recognizing the obvious. The fact is transformer training compute will likely 100x within the next 6 years or so. Context windows are dramatically increasing in size (want to train a general robot replacement for a job - just load up the context window with all sorts of job relevant data/documentation). These LLMs also don’t use recurrency and search, which probably won’t take long to fix. I think the fact that they were able to get so much out of retarded multilayer perceptron trained on next token prediction should scare the SHIT out of people.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47481645)



Reply Favorite

Date: March 10th, 2024 11:53 PM
Author: bearded odious french chef

Just curious. How would you respond to this critique?

“ Other skeptics are critical of the technology itself, with AI analyst Gary Marcus saying that hallucinations (where AI systems such as ChatGPT make up or mangle facts) are not a simple problem to get rid of.

Marcus believes this intractable problem could stop AI from delivering the financial returns its advocates expect.

Marcus told the Financial Times, ‘There is a fantasy that if you add more data, it will work. But you cannot succeed in crushing the problem with data.’

Marcus pointed to previously highly hyped AI technologies, such as Facebook’s AI assistant M or IBM’s Watson, which promised world-changing results but were under-delivering.

Marcus wrote on his Substack, ‘Hallucinations are in their silicon blood, a byproduct of the way they compress their inputs, losing track of factual relations in the process. I first pointed out this risk in 2001, in the fifth chapter of my book The Algebraic Mind, and the problem has persisted ever since. To blithely assume that the problem will soon disappear is to ignore 20 years of history.”

Marcus says that he believes the problem of AI hallucinations will eventually be ironed out, but no one knows whether the technological breakthroughs to solve the problem will arrive in months, years, or decades.”



(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47482187)



Reply Favorite

Date: March 11th, 2024 6:11 AM
Author: Concupiscible canary haunted graveyard

First of all, I would just like to note that Gary Marcus doesn't have a solution to anything. The logic/human interpretable AI guys have been consistently unable to build anything remotely like ChatGPT. That entire research program is highly questionable and doesn't align with biology where simple connectionist units wire up apparently based on predictive processing principles. Traditional, human interpretable AI research will always fail to data driven approaches that can exploit massive search to find generalizing programs.

Hallucinations are largely the result of having imperfect recall of the data and token prediction discouraging the model from expressing uncertainty. I think retrieval based systems that can directly access the data during inference will greatly increase model accuracy while increasing capacity for algorithm learning. Better or more RLHF or something like it so the model can learn to express uncertainty about its knowledge. It's hard to know for sure what will work at this point but the hallucinations have already declined significantly from base GPT-3.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47482538)



Reply Favorite

Date: March 11th, 2024 2:35 PM
Author: self-centered big-titted antidepressant drug

was this written by GPT-3

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47483506)



Reply Favorite

Date: March 15th, 2024 10:49 AM
Author: Copper Cruel-hearted Toilet Seat Mood

"THEY haven't made any breakthrough discoveries, unlike us AI fags."

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47496256)



Reply Favorite

Date: March 11th, 2024 2:32 PM
Author: snowy location

I would respond that doing factually supported substantive responses to any possible research query is like 1% of the bigger AI picture.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47483499)



Reply Favorite

Date: March 15th, 2024 10:50 AM
Author: Copper Cruel-hearted Toilet Seat Mood

That's only because no AI can do it. You're right that 100% of AI tricks do nothing to answer any question we might have.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47496261)



Reply Favorite

Date: March 11th, 2024 2:41 PM
Author: Bistre theater telephone

(((marcus))) is a joke. he seems to believe (1) AI dangerous/bad and (2) AI is dumb. LJL.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47483529)



Reply Favorite

Date: March 11th, 2024 6:01 PM
Author: exciting flickering scourge upon the earth

I think either he’s just trying to get attention as being the AI critic guy or is afraid that continued progress will discredit him intellectually, so he says there should be a pause in AI scaling. He’s intellectually dishonest.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47484307)



Reply Favorite

Date: March 15th, 2024 9:40 AM
Author: Copper Cruel-hearted Toilet Seat Mood

This is absolutely cr and no one America seems to get it. Our academic culture was already infested with the kind of "model our way to the truth" mindset and it has never worked

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47496043)



Reply Favorite

Date: March 15th, 2024 11:07 AM
Author: Concupiscible canary haunted graveyard

That seems to be the case with a lot of academic fields. Even more respectable ones like Econ involve creating bogus toy models of things that almost no one believes in. I think there are likely many models of reality that don’t correspond to anything that can be expressed in language. The computer vision researchers tried for a long time to create models of perception that fit human intuition and they simply do not work well in comparison to essentially uninterpretable convolutions neural networks. Perhaps a lot of fields would be better served by throwing out human modeling and finding ways to train on large datasets.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47496356)



Reply Favorite

Date: March 15th, 2024 11:44 AM
Author: bearded odious french chef

Here's another critique of AI or at least ChatGPT, that seems the most convincing I've read.

https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47496444)



Reply Favorite

Date: March 16th, 2024 12:00 AM
Author: Concupiscible canary haunted graveyard

Definitely a better critique than most. I think there are plausible extensions of LLMs that go beyond simply compressing human text. Just as one example, ChatGPT plays decent chess so it must have a learned board representation. A human can use a model of a chess board to increase their chess playing capabilities. Start with a position and some intuition about how to proceed, and then test moves and countermoves in the model. The results are then used to update their intuition about similar positions. Something like MuZero. It seems this is a major focus of AI labs right now - finding ways to get LLMs to do search through the space of tokens, which can then be fed back into the model for training it to superhuman levels of logic, creativity, etc.

The other point I would make is that the presence of earlier LLM output in the training corpus can probably be undone by RLHF or training harder on higher quality data. LLMs are trained on all sorts of shit human text and learn to emulate it as well, but the final RLHF tuning steers them away from it. I don’t think this will prevent LLMs (or diffusion models) from improving.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47498537)



Reply Favorite

Date: March 16th, 2024 12:16 AM
Author: Self-absorbed Dark Newt Orchestra Pit

What's going to be interesting to see play out is what happens when LLMs begin to inevitably consume mostly their own content. As a shitcoder it's pretty easy to see a scenario where innovation in libraries and new languages actually goes down just because there's no reason for an AI to pick HTMX over React or Julia over Python or whatever. Overall software output will skyrocket of course, but rapidly converge on a few technologies that remain static.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47498569)



Reply Favorite

Date: March 10th, 2024 10:11 PM
Author: Sooty locale
Subject: Buy Black African Superman TOP-Class Permanence Tablet - Poerkan

black african superman pills is a natural sex medicine originated from China! It has very strong effect on curing sexual disorder!

https://www.poerkan.com/product/african-superman-sex-pill/

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47481973)



Reply Favorite

Date: March 15th, 2024 10:52 AM
Author: Aphrodisiac Charcoal Senate Feces

It basically just a power tool. Like a nail gun for a builder. you no longer have to do that shit by hand. that doesn’t mean builders no longer exist.

(http://www.autoadmit.com/thread.php?thread_id=5501700&forum_id=2#47496268)