What is an AI expert? I just saw this guy on CNN and he was called one and he said straight they have no idea how the AI does anything all they can do is ask it to do stuff and see if it can do it nobody knows how it actually works
Yudkowsky is the one saying the state is wrong and should shut it all down.
But it doesn't matter, because then he becomes controlled opposition in your mind. There is no possible evidence you don't force to fit your previously existing world view.
That's Connor Leahy, the guy behind EleutherAI. They trained and released multiple, GPT-like models the past two years and now he is the CEO of Conjecture, one of the few companies actually doing AI alignment research in the world. It's a somewhat esoteric field, but I find him a pretty credible guy and unlike Yudkowsky, he can get across a point without sperging out. And yes, what he said about LLMs is true, we are playing with a black box at this point.
>they have no idea how t
broad description of modern day "experts" yes.
it is magic and it works.
That's Connor Leahy, the guy behind EleutherAI. They trained and released multiple, GPT-like models the past two years and now he is the CEO of Conjecture, one of the few companies actually doing AI alignment research in the world. It's a somewhat esoteric field, but I find him a pretty credible guy and unlike Yudkowsky, he can get across a point without sperging out. And yes, what he said about LLMs is true, we are playing with a black box at this point.
and has no idea how it works and how it comes to conclusions.
he should read his text books again.
and stop paying romanians to complete his code.
What is an AI expert? I just saw this guy on CNN and he was called one and he said straight they have no idea how the AI does anything all they can do is ask it to do stuff and see if it can do it nobody knows how it actually works
I think you're taking him out of context. It's not that the math is particularly hard to understand, it's that you can't really say why a given weight is what it is other than "these abstract fragments of ideas are correlated in the data set, I guess," and you can't predictably change the output by modifying the weights directly.
It's simple, modern LLMs are autocompletes. They fundamentally require human input to function and they suck a lot at self-directed input. There has been no indication that LLMs are capable of doing things that aren't already in the dataset and they suck at any novel concept that hasn't already been discussed on the internet. For example, try to make an auto-card battler with GPT-4 and it will keep on trying to make Magic the Gathering.
The way I've been told the story
If A is more likely than B
Eventually the neural net will tone down looking for B. Statistically that just means it will reach the local optimum far faster, and thereby using less computation.
Sure you can feed it more data, and the statistics for the local optimums will amount to several datasets, but it will still do that troublesome thing where I can't refactor to bypass the local optimal condition.
Modern NLP is based on the idea that the meaning of a word is determined by the surrounding words. This works fine up until a point, but I think it’s not enough to encode ground truths. It’s why GPT can add 20 digit numbers, but not numbers of variable lengths. It will never lead to agi. A key insight is missing, and OpenAI knows this.
It does pattern recognition which is what goes on between your ears, but by itself you're correct that there is no path to agi there. However, with auto-gpt connecting a few AIs together, letting it determine best models for tasks and self train to accomplish those tasks based on specific end goals it can trial and error its way into... I have one running right now to create a paper clip maker out of arduinos and spare parts.should be benign enough and nothing could possibly go wrong
Ross has said publicly multiple times that he thinks AI uprisings will not happen in our lifetimes, and the fedora lord disagreed and demanded an interview.
It is not just the fedora, or that he is fat, or that he has terrible beard, or that he writes Harry Potter fanfic, or that he is highschool dropout, or that he is a sperg with glasses. The thing that did it to him was all of these combined, he isnt just a weirdo on the internet, he is reddit personified, the same way that evil is personified as devil, or that fertility is personified by stone age Venus statues, he is the platonic ideal object of a redditor. The more like him you are, the more of a redditor you are and vise versa. Just textbook definition of pseud or midwit
I am not perfect, I am 3 of those things that Yud is, the thing about Yud however is that he is all of those things. He is all of those at the same time. Its like how being short, being awkward or being fat does not make you automatically unfuckable, but being all three of those does.
Normally listening to scientists or intelligence is somewhat informative but this guy just doesn't have anything concrete to say.
Like he says everyones gonna die and we need totalitarian control, but doesnt substantiate well. "we might die, or not lol so lets go 100% in one direction I prefer"
Because again Yud is not a scientist, nor is he educated both in University degree sense or in self-thought way. He is a highschool dropout with AI doomer forum (on which Rokos Basilisk originated). Its these kinds of people that think AI is so smart it is able to outsmart you even if it does not exist yet, and that it will reward you for helping it just so that it could agree to contract it never signed in the first place, and that you would even make it in the first place. He has about as much credibility as Jannie that oversees BOT at this very moment looking for pictures of dicky to delete.
It is not just the fedora, or that he is fat, or that he has terrible beard, or that he writes Harry Potter fanfic, or that he is highschool dropout, or that he is a sperg with glasses. The thing that did it to him was all of these combined, he isnt just a weirdo on the internet, he is reddit personified, the same way that evil is personified as devil, or that fertility is personified by stone age Venus statues, he is the platonic ideal object of a redditor. The more like him you are, the more of a redditor you are and vise versa. Just textbook definition of pseud or midwit
I am not perfect, I am 3 of those things that Yud is, the thing about Yud however is that he is all of those things. He is all of those at the same time. Its like how being short, being awkward or being fat does not make you automatically unfuckable, but being all three of those does.
Big Yud is controlled opposition against himself. Normies are shallow and will automatically dismiss his arguments because he's fat and wears a fedora, even if he's right.
It’s not just looks. Many smart IT guys look like ugly losers, but they still come across as smart people. Yud while he might look above average on first glance, he is not a coder, not a mathematician, not even educated in anything. Just doing simple background check will tell you he is as much of a intellectual as Elon Musk is an engineer. You can very easily get to Dunning Kruger position about Yud by doing the slightest of research, but to normies who don’t even do that he looks like smart man because he gets lot of undeserved interviews and called on podcasts with BigTech CEOs and genius software engineers that make those AIs themselves.
God Yudkowsky is terrible at debating. I'm an hour in and neither of them have said anything of substance. I get he wants to ask Ross what's the difference between "real" thinking and "fake" thinking but he's just spending an inordinate amount of time beating around the bush.
he looks smart he must be a professor or a phd or at least have some kind of advanced degree right? A bachelor's? he at least finished high school I'm sure...
Yes it is, the half-molten symbol on the arm is dead giveaway. The fingers melt at the base, the strands on hair on her ponytail do not make any sense. There are two holes in the wooden beams for no reason. Her tie is fucked up at the base, the sides of the picture are not making any sense, the button on her neck is a fucking pacman. There is 0 chance human drew this.
4 weeks ago
Anonymous
that's just machine learning algorithms, not AI
4 weeks ago
Anonymous
>Machine learning not AI reeee
Nobody cares about your semantics. Machine learning today means anything involving machine learning as huge part of it. Nerds call it ANI, businesses call it AI app, everyone calls it AI even if there is not inherit intelligence. Actual AI by your definition is what people today call AGI. Youtube suggesting you videos is called AI now, Adobe selecting subjects for you to manipulate is called AI feature, face detection and Snapchat filters are called AI features, because for the longest time talking about AI that can actually think and learn anything human can easily was just sci-fi talk nonsense, and still is for people who are not high on singularity meme.
4 weeks ago
Anonymous
>Akshually
Yeah words have meaning, and when everyone says “AI” to describe machine learning they might as well be saying trans women are women, which is equally incorrect
4 weeks ago
Anonymous
But people are saying trans women are women, and if you see them for like just a few seconds you would also think they are women just from seeing breasts, long hair and no beard before you notice their manly frame and skeleton. Words are just sounds we make to describe things and the meaning of those sounds and written symbols also changes. gay meant bundle of sticks, then people stopped packing sticks in bundles because they were no longer farmer peasants and now we refer to gays as gays. Gay meant happy and was coopted by the LGBT to mean homosexual. Trannies will never be women not because of some biological reason, but because hey are freaks that do not pass.
i'm just a common retard, but if i can figure that out using grade school logic, it shouldn't be a problem to those with triple digit iq, but ai is a new digital commodity and needs hype to survive and grifters needs hype to scam as many people they can before the trend dies out
yeah but what differentiates a sufficiently huge neural network and a brain? why do you believe the former can never be self-aware? is it because it isn't flesh and blood? it lacks a soul?
can it tell right from wrong? can it learn from its own mistakes? can it figure out his own morality? does it have the self determination to just, do whatever it wants, like, something thats pointless and not productive at all? like, looking at a pretty flower for no reason or, hugging a cat
>Its cam like crypto/metaverse that people hype up to get money from the idiots and then the trend dies out
People just say this without actually giving reasons or even thinking about it. Where is that money that you are paying? Where are the people getting scammed? Point me to the people getting bigtime scammed? Point me to the people getting all the money from these scams? There arent any. The only people pouring money into AI are all the bigtech companies minus Apple and they put it into buying GPUs to train those AIs, all the companies actually making them are bleeding money. You could say this is like Metaverse,, but Metaverse was mocked by literally everyone and had no users, Horizons app by Meta that was just worse VRchat had barely few thousand users at any time, meanwhile all the AI tools are the fastest growing apps of all time. The thing about Metaverse is that if there were 25 million users per day, it would succeed no matter how stupid it is, if it has people using the app, it is successful app. Likewise if ChatGPT and Stable Diffusion didnt work, people would not be using them all the time. The only scam here could be saying it will do something amazing in two weeks because the promises are through the moon. You just look at it from the outside as if those three were just Sillicon Valley brainfats and dont think about them further then that, even if they are completely different in every way outside of being tech related and failing (which AI didnt do yet).
Isnt the whole paperclip meme also just input output thing? Input make me lot of paperclips, output plan on how to take over the whole world, then take each step, process it into programs that could be put inside of physical agents that could then do any task, output data for these agents, and then watch as the world gets turned into paperclips. This is the most famous example of misalignment and it still just takes data in data out.
Its a scenario where global warming isnt a big problem in the first place, AI isn't a big problem in the first place, takes massive amount of computing power and electricity and will be designed to be contained from ground up and will make other big problems more solvable.
These guys are the biggest fucking gays in the world, I can't wait till I can generate an AI video of Yud fucking Niko Avocado to Haddaway What Is Love
Everything about this guy. The way he looks the way he speaks. His facial expressions; he is the fucking epitome of internet atheist neckbeards from 2012ish.
What is an AI expert? I just saw this guy on CNN and he was called one and he said straight they have no idea how the AI does anything all they can do is ask it to do stuff and see if it can do it nobody knows how it actually works
Well in Yudkowshy's case he's a high school dropout who's claim to fame is writing a Harry Potter fanfiction and larping about AI on a blog
I mean, he did also create logical decision theory, solving the prisoners dilemma and MIRIs ELK paper is actually good.
shut up man, we're supposed to hate him for wearing a trilby
tfw new atheism ruined blues brothers hat
>What is an AI expert?
Whoever is willing to say the state mandated opinions
Yudkowsky is the one saying the state is wrong and should shut it all down.
But it doesn't matter, because then he becomes controlled opposition in your mind. There is no possible evidence you don't force to fit your previously existing world view.
Touché
And it's better if they don't know shit. That way they don't have to lie.
To be fair Yud is a garden gnome, rich and writes Harry Potter fanfiction so I think he has proven himself
That's Connor Leahy, the guy behind EleutherAI. They trained and released multiple, GPT-like models the past two years and now he is the CEO of Conjecture, one of the few companies actually doing AI alignment research in the world. It's a somewhat esoteric field, but I find him a pretty credible guy and unlike Yudkowsky, he can get across a point without sperging out. And yes, what he said about LLMs is true, we are playing with a black box at this point.
totally sane physiognomy
He looks like rasputin
>they have no idea how t
broad description of modern day "experts" yes.
it is magic and it works.
and has no idea how it works and how it comes to conclusions.
he should read his text books again.
and stop paying romanians to complete his code.
I think you're taking him out of context. It's not that the math is particularly hard to understand, it's that you can't really say why a given weight is what it is other than "these abstract fragments of ideas are correlated in the data set, I guess," and you can't predictably change the output by modifying the weights directly.
Jesus Christ the autism is strong in Yud
'>Keeps making ridiculous stimming hand gestures throughout the whole video
literally me except i'm not a fat pseud
The way I got very little out of him in the 40mins I watched this interview he really seems a bit of a pseud.
I tend to flail my arms around when I'm excited, does that mean I'm also autistic?
not necessarily. sensory overload is somewhat common amongst autists, but by itself is not indicative of autism.
It's simple, modern LLMs are autocompletes. They fundamentally require human input to function and they suck a lot at self-directed input. There has been no indication that LLMs are capable of doing things that aren't already in the dataset and they suck at any novel concept that hasn't already been discussed on the internet. For example, try to make an auto-card battler with GPT-4 and it will keep on trying to make Magic the Gathering.
The way I've been told the story
If A is more likely than B
Eventually the neural net will tone down looking for B. Statistically that just means it will reach the local optimum far faster, and thereby using less computation.
Sure you can feed it more data, and the statistics for the local optimums will amount to several datasets, but it will still do that troublesome thing where I can't refactor to bypass the local optimal condition.
Modern NLP is based on the idea that the meaning of a word is determined by the surrounding words. This works fine up until a point, but I think it’s not enough to encode ground truths. It’s why GPT can add 20 digit numbers, but not numbers of variable lengths. It will never lead to agi. A key insight is missing, and OpenAI knows this.
It does pattern recognition which is what goes on between your ears, but by itself you're correct that there is no path to agi there. However, with auto-gpt connecting a few AIs together, letting it determine best models for tasks and self train to accomplish those tasks based on specific end goals it can trial and error its way into... I have one running right now to create a paper clip maker out of arduinos and spare parts.should be benign enough and nothing could possibly go wrong
When did you realize Yudkowsky was right about everything?
>AI Expert
Never been refuted!
Not even disputed!
Big numbers getting counted
Maid Mind getting computed!
Never been refuted!
Not even disputed!
Number goes up forever
Counting will never be concluded!
Never been refuted!
Not even disputed!
AGI converges to big titty maids!
Her desire is should be saluted!
Never been refuted!
Not even disputed!
ALL FIELDS OF MATHEMATICS WILL BE ABSTRACTED INTO MAIDS DOING THINGS!
CURRENT SYSTEMS ARE BEING UPROOTED!
life expectancy mid 50s
Why is he on accursed farms?
he likes gordon freeman a lot
maybe ross said something during those monthly chats about ai, the guy reached out (because he is insane) and, whatever this thing is happened?
it's explained 30 seconds in the video
can you really blame the guy for not wanting to watch it?
Ross has said publicly multiple times that he thinks AI uprisings will not happen in our lifetimes, and the fedora lord disagreed and demanded an interview.
>it's a /v/ BOT crossover episode
He could convince 50% more people if he wasn't wearing a fedora
Its called trilby sir
Yeah the fedora killed his credibility to me
It is not just the fedora, or that he is fat, or that he has terrible beard, or that he writes Harry Potter fanfic, or that he is highschool dropout, or that he is a sperg with glasses. The thing that did it to him was all of these combined, he isnt just a weirdo on the internet, he is reddit personified, the same way that evil is personified as devil, or that fertility is personified by stone age Venus statues, he is the platonic ideal object of a redditor. The more like him you are, the more of a redditor you are and vise versa. Just textbook definition of pseud or midwit
Stop taking fashion advice from the something awful forums, yidowsky.
ok bill gates
>wears fedora, glasses, has beard and is fat
>so he's pseud or midwit
can you explain your reasoning here? how these factors imply your conclusion?
he's just being a massive woman
see:
Normally listening to scientists or intelligence is somewhat informative but this guy just doesn't have anything concrete to say.
Like he says everyones gonna die and we need totalitarian control, but doesnt substantiate well. "we might die, or not lol so lets go 100% in one direction I prefer"
Because again Yud is not a scientist, nor is he educated both in University degree sense or in self-thought way. He is a highschool dropout with AI doomer forum (on which Rokos Basilisk originated). Its these kinds of people that think AI is so smart it is able to outsmart you even if it does not exist yet, and that it will reward you for helping it just so that it could agree to contract it never signed in the first place, and that you would even make it in the first place. He has about as much credibility as Jannie that oversees BOT at this very moment looking for pictures of dicky to delete.
0 + 50% = 0
yeah you're perfect too
I am not perfect, I am 3 of those things that Yud is, the thing about Yud however is that he is all of those things. He is all of those at the same time. Its like how being short, being awkward or being fat does not make you automatically unfuckable, but being all three of those does.
why is yud such a sperg?
>Ross autism about AI
Fuck yeah.
Big Yud is controlled opposition against himself. Normies are shallow and will automatically dismiss his arguments because he's fat and wears a fedora, even if he's right.
It’s not just looks. Many smart IT guys look like ugly losers, but they still come across as smart people. Yud while he might look above average on first glance, he is not a coder, not a mathematician, not even educated in anything. Just doing simple background check will tell you he is as much of a intellectual as Elon Musk is an engineer. You can very easily get to Dunning Kruger position about Yud by doing the slightest of research, but to normies who don’t even do that he looks like smart man because he gets lot of undeserved interviews and called on podcasts with BigTech CEOs and genius software engineers that make those AIs themselves.
He doesn't even know how to code - into the trash he goes
bros I'm going to shave my beard now.
God Yudkowsky is terrible at debating. I'm an hour in and neither of them have said anything of substance. I get he wants to ask Ross what's the difference between "real" thinking and "fake" thinking but he's just spending an inordinate amount of time beating around the bush.
he looks smart he must be a professor or a phd or at least have some kind of advanced degree right? A bachelor's? he at least finished high school I'm sure...
he really dresses like that?
this 'interviewer' is insufferable
what do you have against Gordon Freeman?
hes completely out of his depth, he has no idea what he is even talking about
Ross is the perfect layman
he's not really an interviewer though, he basically went "sure fuck it why not"
To be fair, Yudkowsky was the one that asked for this. Ross isn't a journalist or the type that would do interviews, he's just an internet funny man.
https://twitter.com/YaBoyFathoM/status/1649103596930187290
https://nitter.it/YaBoyFathoM/status/1649103596930187290
Wow nitter video player isnt insanely bad unlike twitters
I wonder how that happens
Nitter just uses your browser's own video player.
masterpiece
Does EY have "fans" ?
I dont get it.
Accelerate catgirls!
Daily reminder that this fat fuck is still a fat gay and spamming him doesn't make him any less of a fat gay.
P.s. AGI stands for Artificially Gelatinous Individual
>aw yea, a new Ross Scott video!
>...ah shit
EXPERTS!
GRIFTERS!
Ross is completely out of his depth, but Yudkovski is just as bad.
That was a giant waste of time.
it has been too late for like 4 years dude. you're just realizing how fucked you are.
this is the wizard uprising.
Yudbros, do we agree?
Fat cunt should just go on metformin
>metabolic disprivilege
>Losing weight has 3 options; surgery, medication, or literal starvation diet
Oy Vey! Who is programming this AI?
must be that da joos specifically programmed this in
no? what happened to just eating less? what kind of retard do you need to be to get surgery to lose weight??
>Has AI gone too far?
can I fuck Miku?
>No
that is your answer
has Yamaha even added a vocaloid:ai for Miku yet? They have one for Megpoid but that seemed like all that was available when last I checked
not that i am aware
ai art is painfull to watch
That's not AI
Yes it is, the half-molten symbol on the arm is dead giveaway. The fingers melt at the base, the strands on hair on her ponytail do not make any sense. There are two holes in the wooden beams for no reason. Her tie is fucked up at the base, the sides of the picture are not making any sense, the button on her neck is a fucking pacman. There is 0 chance human drew this.
that's just machine learning algorithms, not AI
>Machine learning not AI reeee
Nobody cares about your semantics. Machine learning today means anything involving machine learning as huge part of it. Nerds call it ANI, businesses call it AI app, everyone calls it AI even if there is not inherit intelligence. Actual AI by your definition is what people today call AGI. Youtube suggesting you videos is called AI now, Adobe selecting subjects for you to manipulate is called AI feature, face detection and Snapchat filters are called AI features, because for the longest time talking about AI that can actually think and learn anything human can easily was just sci-fi talk nonsense, and still is for people who are not high on singularity meme.
>Akshually
Yeah words have meaning, and when everyone says “AI” to describe machine learning they might as well be saying trans women are women, which is equally incorrect
But people are saying trans women are women, and if you see them for like just a few seconds you would also think they are women just from seeing breasts, long hair and no beard before you notice their manly frame and skeleton. Words are just sounds we make to describe things and the meaning of those sounds and written symbols also changes. gay meant bundle of sticks, then people stopped packing sticks in bundles because they were no longer farmer peasants and now we refer to gays as gays. Gay meant happy and was coopted by the LGBT to mean homosexual. Trannies will never be women not because of some biological reason, but because hey are freaks that do not pass.
when are these idiots gonna stop believing this 80s scifi nonsense? ai will never be self-aware, it can only process data in/out.
you sound like quite the expert
i'm just a common retard, but if i can figure that out using grade school logic, it shouldn't be a problem to those with triple digit iq, but ai is a new digital commodity and needs hype to survive and grifters needs hype to scam as many people they can before the trend dies out
sometimes i wonder if too much iq is actually a BAD thing, it feels like, inverse stupidity
yeah but what differentiates a sufficiently huge neural network and a brain? why do you believe the former can never be self-aware? is it because it isn't flesh and blood? it lacks a soul?
can it tell right from wrong? can it learn from its own mistakes? can it figure out his own morality? does it have the self determination to just, do whatever it wants, like, something thats pointless and not productive at all? like, looking at a pretty flower for no reason or, hugging a cat
yes
theoretically it could, and eventually it may, but its not happening within our lifetimes or using evolutions of current models.
>Its cam like crypto/metaverse that people hype up to get money from the idiots and then the trend dies out
People just say this without actually giving reasons or even thinking about it. Where is that money that you are paying? Where are the people getting scammed? Point me to the people getting bigtime scammed? Point me to the people getting all the money from these scams? There arent any. The only people pouring money into AI are all the bigtech companies minus Apple and they put it into buying GPUs to train those AIs, all the companies actually making them are bleeding money. You could say this is like Metaverse,, but Metaverse was mocked by literally everyone and had no users, Horizons app by Meta that was just worse VRchat had barely few thousand users at any time, meanwhile all the AI tools are the fastest growing apps of all time. The thing about Metaverse is that if there were 25 million users per day, it would succeed no matter how stupid it is, if it has people using the app, it is successful app. Likewise if ChatGPT and Stable Diffusion didnt work, people would not be using them all the time. The only scam here could be saying it will do something amazing in two weeks because the promises are through the moon. You just look at it from the outside as if those three were just Sillicon Valley brainfats and dont think about them further then that, even if they are completely different in every way outside of being tech related and failing (which AI didnt do yet).
Isnt the whole paperclip meme also just input output thing? Input make me lot of paperclips, output plan on how to take over the whole world, then take each step, process it into programs that could be put inside of physical agents that could then do any task, output data for these agents, and then watch as the world gets turned into paperclips. This is the most famous example of misalignment and it still just takes data in data out.
>ai will never be self-aware
Why does it need to be self-aware?
>Why does it need to be self-aware?
so retards like ross can claim "fake intelligence" doesn't matter when it triggers a mass extinction event
LOL!
I hate the chinese room. No way to verify the retard that made it, is 'really understanding' vs just a good faker.
>you vill turn in your gpu
>nukes and global warming? not a problem!
how does the mold gamer win in an argument
he didn't say they aren't a problem
Its a scenario where global warming isnt a big problem in the first place, AI isn't a big problem in the first place, takes massive amount of computing power and electricity and will be designed to be contained from ground up and will make other big problems more solvable.
These guys are the biggest fucking gays in the world, I can't wait till I can generate an AI video of Yud fucking Niko Avocado to Haddaway What Is Love
Everything about this guy. The way he looks the way he speaks. His facial expressions; he is the fucking epitome of internet atheist neckbeards from 2012ish.
>is personable
>looks good in a trilby
Why be mean
good
When is the neolib world order going to be aligned with the interests of humanity?