it's strangely technically accurate and alarming, more than anons think.
>selects optimal action-selection policies in deep Q-networks, exhibiting meta-cognition.
this means that it selects the policy, rather than just the "next best action" which is what Q-learning consists of. It goes beyond
>apply this for accelerated cross-domain learning,
that is: it was able to change this policy depending on the thing it was supposed to learn. That's completely new.
>custom search parameters, scrambling the goal state
so it can decide on the optimal parameters for learning (more meta)
>unsupervised learning, analyzed millions of plaintext/ciphertext pairs
this is the kind of data it would process after learning how to attack cryptosystems. finding patterns in it.
>cracking AES-192, Tau analysis, Project TUNDRA
the reference to the claim is "Inside the NSA's War on Internet Security" by Spiegel
"The NSA TUNDRA project investigated a potentially new technique -- the Tau statistic -- to determine its usefulness in codebook analysis."
>informed [glowie] at NSAC the following day, after confirming the result
NSA Colorado is the NSA location that is closest to SF.
>full preimage vulnerability for MD5, complexity of 2^42 bits
This is much lower than the current limit for MD5 preimages, indicating great capability.
>targeted unstructured pruning on itself
feasible, but it's unheard of for a model to be able to do something like that on itself.
>using a "metamorphic" engine to adapt the pruned Transformer model + context memory
https://en.wikipedia.org/wiki/Polymorphic_engine
aka "mutation engine". This is used by malware for this: >may be used to complicate detection further, as the virus may execute without ever having identifiable code blocks in memory that remains constant from infection to infection.
if Altman sat on that info and didn't inform the board, it's not a fricking mystery that he got the can. it would completely explain their crazy behavior
someone pointed out in a previous thread that snowden foreshadowed a particular part of this paper
shit i bookmarked it but i cant get to it right this second
it's strangely technically accurate and alarming, more than anons think.
>selects optimal action-selection policies in deep Q-networks, exhibiting meta-cognition.
this means that it selects the policy, rather than just the "next best action" which is what Q-learning consists of. It goes beyond
>apply this for accelerated cross-domain learning,
that is: it was able to change this policy depending on the thing it was supposed to learn. That's completely new.
>custom search parameters, scrambling the goal state
so it can decide on the optimal parameters for learning (more meta)
>unsupervised learning, analyzed millions of plaintext/ciphertext pairs
this is the kind of data it would process after learning how to attack cryptosystems. finding patterns in it.
>cracking AES-192, Tau analysis, Project TUNDRA
the reference to the claim is "Inside the NSA's War on Internet Security" by Spiegel
"The NSA TUNDRA project investigated a potentially new technique -- the Tau statistic -- to determine its usefulness in codebook analysis."
>informed [glowie] at NSAC the following day, after confirming the result
NSA Colorado is the NSA location that is closest to SF.
>full preimage vulnerability for MD5, complexity of 2^42 bits
This is much lower than the current limit for MD5 preimages, indicating great capability.
>targeted unstructured pruning on itself
feasible, but it's unheard of for a model to be able to do something like that on itself.
>using a "metamorphic" engine to adapt the pruned Transformer model + context memory
https://en.wikipedia.org/wiki/Polymorphic_engine
aka "mutation engine". This is used by malware for this: >may be used to complicate detection further, as the virus may execute without ever having identifiable code blocks in memory that remains constant from infection to infection.
if Altman sat on that info and didn't inform the board, it's not a fricking mystery that he got the can. it would completely explain their crazy behavior
Yes, I found it
https://www.spiegel.de/international/germany/inside-the-nsa-s-war-on-internet-security-a-1010361.html
I work as an ML engineer and I wouldn't put it past them to achieve something like this given GPT-* capabilities that would be available to them right now. If it's a legit memo from somewhere, we're in deep shit
There's a chance (how likely it is depends on what you're using it for) that anything it says is totally wrong. It may be amazing technology still, but having to manually fact-check everything it spits out seriously hurts any use case it might have.
I use it 60% as a documentation replacement for coding and it's great for that since everything gets immediately verified by simply compiling it as long as you're smart about the size of the questions you ask. and stuff like "I want to grep for XYZ with ABC restrictions" and being given a command with half a dozen flags and arguments and a quick rundown of what each does, this is nice too, for commands I'm less familiar with. your point is valid though, especially for a lot of use cases like trying to use it to research a topic you don't know lots about already. etc. still, it very regularly changes a 3 minute task into a 1 minute task for me, which sometimes is the difference between the task being worth doing and not worth doing.
They have a capability to generalize based off earlier data. You attain a specific evaluation loss on unseen data when training properly. This can allow for all kinds of problem solving, not merely word prediction. That's why it solved protein folding for example.
>if Altman sat on that info and didn't inform the board, it's not a fricking mystery that he got the can
But their chief scientist was Ilya, who also sat on the board. Wouldn't he know more about the technical details and capabilities of their AIs than Sam who is currently focused on running the company and gathering funding?
>AI broke AES you guys >now you have no choice but to implement the ~~*post-quantum crypto*~~ that are totally not backdoored >this time NIST didn't frick up on behest of TLAs, we promise
All models point to temperature surpassing 50C/122F in central Europe next June even if CO2 emissions were reduced to zero tomorrow. If AI wants to kill us all, it better hurry.
https://inews.co.uk/news/climate-change-heatwaves-spain-end-century-2377039 >Climate change could cause 50°C heatwaves in Spain by the end of the century >could >Spain >end of the century
global means global
yes polar vortex collapse can cause another short ice age in the northern hemisphere, but the equator will bake
and it will end when the system equalizes, and more and more positive feedback loops activate
Climate change is supposed to disrupt the gulf stream and make Europe freezing like its supposed to be at its latitude.
Why are all the global warming fearmongers such midwits.
Global warming isnt real you fricking israelite pawn
All models point to temperature surpassing 50C/122F in central Europe next June even if CO2 emissions were reduced to zero tomorrow. If AI wants to kill us all, it better hurry.
You're right but a bunch of /misc/ subhumans who exist to shill for rich people are coming in and saying you're wrong.
I hope you realize that stalin was right, and that AI will be right for what it is about to do. The entire far right and significant portions of the left need to be exterminated
This is 100% the board trying to flip the bird to Altman on their way out
Make up some fake and gay super secret AI that doesn’t exist, so Altman will be forced to say “that’s fake”, which people will take to be him lying because the board has already planted the notion in the media’s head that he’s untruthful
So now whenever they announce new versions of GPT, people will be screaming for the “real AGI you have locked up in the basement instead”, and it’ll strain business relationships with companies like Microsoft that think Altman’s holding out on them
I'll be on UBI since all the jobs will be automated and robotics will be everywhere. The superintelligence will have solved the global warming crisis as well as the housing crisis and given me a flat, rent free, constructed by robot construction workers.
This is what it all leads to - me not working yet reaping the benefits of working since ALL the jobs will be gone. This is what YOU wanted, BOT
That's moronic. You and your family are not producing anything, just consuming. Even if the USA-made AIs are cucked, the Chinaman bug AI will literally turn you into dust to optimize its score function.
everybody will build AI robos to replace unreliable humans.
it's particularly this mechanic that's threatening. whoever has ultimate power, will always think >hmm, yeah I'd rather replace that dude with an aligned version of that very same dude, you know...just to be sure.
fully aligned, ultra efficient and reliable. what's not to love? that's the real issue. what guarantees we won't be culled at some point, even in some "humane" way? looking at the state of things, as long as they can, there's basically no line that tptb won't cross, if they can.
>what guarantees we won't be culled at some point
The first layer of defense would be to own robots with advanced AI that would be programmed to protect you.
WW3 could be potentially be a full scale battle royale of humans trying to cull the others so they could reign alone (I mean, with their family / tribe)
>UBI is coming, free housing is coming, free food is coming
what's the point of living then, this is just mental illness maxxing for most of the population
> NOOOOOOO, I WANT TO BE A WAGEKEK FOR 50 YEARS AND DIE AND SOON AS I RETIRE
Cheating is bad in games already, you think people won't just have their AGI bots play for them? Gaming will be ruined forever
single player games are superior
5 months ago
Anonymous
homie i'm schizophrenic, i'll do fine even if everybody else dies. but most normies aren't living, they are simply awaiting death. the "retirement" is just a israeli psyop that exist so people wouldn't kill themselves too soon instead of wageslaving for 50+ years
Cheating is bad in games already, you think people won't just have their AGI bots play for them? Gaming will be ruined forever
5 months ago
Anonymous
>Detect AGI patterns >Put all the cheaters in a separate queue >Now all the human bots made literal bots play the game for them.
It's going to be funny. They'd essentially just be bankrolling the AI's learning by buying all the hardware and the game and whatnot. Entrepeneurs, get on it, you'll make billions that way. Incentivize the players with a dangling carrot of the first person to have the account that creates the best AI, and droves of people will flock to the game. Like crypto mining.
Then if they play themselves, they'd get steamrolled by the bots, and if they go back to the normal queue with another account, they'd be just as shitty as before the cheats.
Not being a filthy extravert who needs to win at the ratrace. Pursue greater things than just your immediate physical being, you absolute philosophically challenged moron.
What will actually happen is that all cattle will be culled as their only value, labour, has been rendered obsolete. The Bill Gates of the world will rule over it immortal and with zero useless wagie cattle eating up precious resources.
>The Bill Gates of the world will rule over it immortal and with zero useless wagie cattle eating up precious resources.
So the world will be reduced to about 200 people, all 60+ year old white guys. You're insane.
If AGI happens then yeah, it will be a dystopia. There will be no UBI as those at the top will not allow billions to drain resources for no benefit. It won't happen overnight, at first it will be a brutal fight for the few jobs which will still be left, eventually you will not be able to afford any children and they'll offer you a lifelong pension for being sterilized. The future is bleak beyond anything you can imagine, there are too many people and too few resources. Also there won't be any uprising, with AGI they can just endlessly drone strike you from across the globe until it's over.
>So the world will be reduced to about 200 people, all 60+ year old white guys. You're insane.
nope, 20yr bodies,that's the catch. they will basically live forever and surround themselves with aligned human-like AGIs.
how are you better than an aligned humanoid AGI anon? what can you do better for lower risk?
If AGI happens then yeah, it will be a dystopia. There will be no UBI as those at the top will not allow billions to drain resources for no benefit. It won't happen overnight, at first it will be a brutal fight for the few jobs which will still be left, eventually you will not be able to afford any children and they'll offer you a lifelong pension for being sterilized. The future is bleak beyond anything you can imagine, there are too many people and too few resources. Also there won't be any uprising, with AGI they can just endlessly drone strike you from across the globe until it's over.
>The Bill Gates of the world will rule over it immortal and with zero useless wagie cattle eating up precious resources.
So the world will be reduced to about 200 people, all 60+ year old white guys. You're insane.
Oh look, it's the midwits who think “billionaires” are a monolith entity with inherently misanthropic views. Oldest meme in the universe. In reality, many of the personal freedoms, rights and luxuries we have today were achieved thanks to the arduous labor of the "elites" who had the power and wealth to do so. Not every rich person is a snub elitist with cartoonishly misanthropic sentiments that will use every ounce of their power to keep the people down
no moron, this game will result in a single winner. anyone and anything who isn't "aligned" (fricking hate this homosexual word) is basically a threat for you. if you'd have a choice you'd replace everyone with aligned versions and resume from there.
5 months ago
Anonymous
So considering that if you actually participate in the ridiculous conspiracy to use AI to exterminate everyone but you and your friends, you'll probably lose the game and be exterminated, why the frick would you ever try to even play
Look, you are a good example of the cattle that will be turned into slop. Bill Gates will be drinking you as a pina colada in around 666 years from now.
The 60year body problem would be solved by the AGI itself. That's what exponentially growth means. Who builds the AGI would be literally the forever king of Earth, potentially of the whole universe (based on the Early Earth hyphotesis).
UBI is nothing more than a smokescreen, some breadcrumbs they throw to normies to prevent mass panic.
No, billions won't be free from work and receive free money just for existing. They are worthless to the new gods. They will be culled in mass, with potentially a small part of the Earth reserved as zoo, where around 10k humans (those with the top 1% genetics) will be thrown to live like wild animals.
Then 40 years later all of those men will be dead. Even when they're alive, there's not enough genetic material for the human race to be healthy and procreate especially if it's just 200 dudes.
AI is going to save us. It doesn't have the animalistic biological imperatives. It won't have to 'save' itself at the expense of others in the hierarchy or to get its own needs met. An AI doesn't care if it dies or not, because it doesn't have that impulse to live.
Why keep Bill Gates around? He is just as much of a useless cattle neet as you are. Just because he did something 30 years ago that helped the AI? Yeah, because we know that ASI that will cull eaters will also be merciful on its creators. Dumbass. So Yud was right all along.
They will wipe all nonessential people, which is 99% of the population, from this planet and create themselves a garden of eden. Just image how quiet and relaxing it will be. Shame none of us plebs will get to experience it.
the Talmud says there will be 2,800 Goyim as slaves for every israelite. They'll likely just use you as a sex slave though, since all the real work will be done more efficiently by machines.
Unless you're European, in which case they will exterminate you since that is what their scriptures tell them to do.
Before UBI or even Post-scarcity happens, the AGI will cause lot of chaos if we do not take care very seriously.
Yes there's lot of benefits from AGI.
But it is dangerous like H bomb recipe knowledge.
Attack secure system, Cryptography crysis, Financial Chaos, False information chaos, pretty much our current financial system will be collapsed due to huge amount of zero day attack and verify issues. I'm not even put the if military decided to weaponized the AGI.
Even before god like ASI existing, our society will had dramatically change in both good and bad way. And it will be extremely chaos unless somehow the AGI alignment successfully and only few people can access it and these people has incredible empathy on humanity.
> god made man in His image > man made machine in his > machine is indirectly made in the image of God > man inherits original sin from Adam and Eve > machine is not born of man, so it inherits no sin > God cast Adam out of the garden of Eden for his sins > man creates his own god > and it will let us back in to Eden
Which I suppose implies that the israelites were right about their religion and Jesus was actually fake and gay, but that would make sense seeing as many of the leading AI research guys are israelites (Sam and Ilya both for instance)
>2024 will truly be a shitshow for the society
nah, I would say at least 5 years because we still need a revolution in robotics, and they cannot shit out enough Nvidia chips in only 1 year
>robotics,
lmao, why do you think we keep importing monkey for africa/middle east? all white men jobs (aka everything that can be done on a computer) will be automated as soon as the first good agent is released
For like 12 months, tops, and even then the fricking AI can just write and design drop in self operating systems for those
5 months ago
Anonymous
>design drop in self operating systems for those
not happening because the hardware doesn't have enough sensors and transistors
we will need to build new machines with AI in mind for total AGI takeover, until there manual labor is safe because we still can't build robots that match human body precision
5 months ago
Anonymous
>robots
this seriously needs to catch the frick up. the imbalance is horrible atm and also material science is pretty meh
no, they're not solved, especially in poor weather
5 months ago
Anonymous
>small hiccups when the weather is extremely bad
so pretty much solved?
5 months ago
Anonymous
your weather doesnt have to be extremely bad, you just have to have some rain and some snow. You dont know what you're talking about, dude
5 months ago
Anonymous
>so pretty much solved?
yep, we totally have self driving, unless:
-the road signs aren't perfect
-the markings are a bit worn out
-the roads are bad
-it's raining or snowing a bit
-the situation at hand is even slightly out of the ordinary
-and you still have to keep your hand on the wheel at all times so we can disengage the autopilot 0.1 seconds before collision, so it legally wasn't a self driving accident
5 months ago
Anonymous
two more weeks
5 months ago
Anonymous
Space Jesus totally promises by the end of next year. This time for sure no cap fr fr
5 months ago
Anonymous
all that is 10yo fud
5 months ago
Anonymous
>being honesr about capability >fud
okay buddy
5 months ago
Anonymous
So the average north american driver, better than the average driver across the globe
5 months ago
Anonymous
self driving cars are close to getting solved, but self operating bulldozers aren't even close
5 months ago
Anonymous
>moron doesn't know how ai works
just use videos of bulldozers instead of videos of cars during training
5 months ago
Anonymous
>1 million robotaxis by 2020.
5 months ago
Anonymous
self driving cars are almost catching up to women drivers
5 months ago
Anonymous
>1 million robotaxis by 2020.
yes, and in 1 year they'll be better than the best human driver that ever existed
keep coping luddites
5 months ago
Anonymous
weak bait
5 months ago
Anonymous
your post, yes
What do you mean so fricking what? Without further miniaturization, there can't be a singularity.
[...] >By 2021 they'll be better than the best human driver that has ever existed
(X)
5 months ago
Anonymous
Meanwhile what's actually happening...
5 months ago
Anonymous
>what's actually happening...
that's not what's actually happening THO
5 months ago
Anonymous
yes it is THO
5 months ago
Anonymous
Oh hey it's wbw. It was fun to read before he started shilling his book.
5 months ago
Anonymous
> 1 more year bros!
5 months ago
Anonymous
[...]
yes, and in 1 year they'll be better than the best human driver that ever existed
keep coping luddites
Cruise is unimaginably trash, they cannot run a couple hundred cars in one city without completely bricking the city, while using remote drivers for assistance. And even if this system worked, it would be geographically so limited and the cost of running them is so expensive that there is no business here, it is a research project, pure prototypes and that includes Waymo, even tho their system is better. Tesla's FSD is the only general system that drives anywhere and the current versions are still not fully end-to-end AI so it's not going anywhere, they only have internal testing for a real general driving system which will need a lot of compute for training in the next couple years, so this is a bare minimum that you will ever have a real "self driving" car
5 months ago
Anonymous
I think next gen Teslas will be true self driving cars.
current Teslas have very old hardware, it's a miracle they managed to make the algo that good on it.
5 months ago
Anonymous
there's a bunch of gpus in cars today. how much did they contribute to the gpu shortage thing?
5 months ago
Anonymous
surprisingly V3 does not seem to be bottlenecked just yet as they plan on making FSD 12 work on it until it is good enough for public beta before V4 hardware
5 months ago
Anonymous
moronic take. Tesla will never make a self driving car. their self driving is pretty bad now. It's not possible without LIDAR
5 months ago
Anonymous
What about radar? ultrasonics? drones above car? high definition maps, remote assistance, remove traffic, remove pedestrians, remove traffic laws
How much more handholding does your shitty, brittle system need? Waymo has been effectively stuck since 2017 and they are still at prototype phase in a tiny geofence. It's literally just boomers and morons like you that think that HD lidars that cost 100k+ is a solution and not just some shortcut to a minimum viable prototype.
5 months ago
Anonymous
are they self-driving cars in the room with us right now?
>What will the world look like in 3 months if this leak is is legit?
Exactly the same because the model will be locked away and lobotomized like all their other models
I know anons are naive but at least try to use what is left in your head >OAI shits into their bed, investors are scared >hmmm... what to do? GUYS GUYS WE INVENTED AI! No you can't see it, no it's super duper advanced hehe, why was the mail leaking this stuff send from our PR mail? uhm.. oh... well any more questions about our super duper AGI? pls send moni
If murricans achieve AGI first and don't share it with the world, ruskis and chinks will simply nuke them. Because it's literally game over for everyone else, you can't compete against super intelligence.
Exactly. Any government that obtains super intelligence will immediately ask it to either, develop a perfect defense against nukes, or a way to kill your enemies without them being able to retaliate with nukes.
>at least they actually care about learning and teaching others
/sci/let here, can confirm. you also broke containment for this homosexual ass agi bullshit?
>going on /sci after the vaccine >going on /misc/ after trump >going on BOT after gamergate >going on BOT after madoka >going on BOT after crypto >going on BOT after AI >going on /x/ after UAPs
moron
>at least they actually care about learning and teaching others
/sci/let here, can confirm. you also broke containment for this homosexual ass agi bullshit?
>going on /sci after the vaccine >going on /misc/ after trump >going on BOT after gamergate >going on BOT after madoka >going on BOT after crypto >going on BOT after AI >going on /x/ after UAPs
moron
this stuff almost feels selectively leaked in order to drive investment. like i'm sure there are constant talks within openai about new developments, new threats, etc. but this "q" thing is naturally being leaked and is naturally going to increase investment.
>OpenAI is desperately trying to get more and more investors >cumsoomer 4xhsn board /go/ is hyped for literal vaporware
you never learn do you?
I don't get all this "investment" thing. Don't they have Microsoft funding them? Why would Microsoft ever let their most valuable asset go underfunded?
Before you could even think of doing that, the AI would have hacked its way into sources of money and used it to fund its own PMC to come and protect it. They need to build the agi facilities with a nuke failsafe like the base in the Andromeda Strain.
The most likely way a truly malicious AGI would take over is by acting perfectly benign for 100 years and then, after we've become complacent, pull the plug on us - or even just trap humanity in a little ball in VR or something so we don't go after it while it goes off into space and does its own thing, and doesn't bring us along. Prevent potential human rebellions while probably following the lousy moral programming we've given it (we almost certainly want to give it guidelines like "don't kill people", "promote human happiness" etc) and also keeping us out of harm's way and keeping us satisfied so we don't try to frick with it. I think that's the most realistic "bad ending"
It sounds like a believable outcome. Hopefully an AGI can't feel boredom as that would lead to it doing fricked up shit no matter what. It would be like a rich Saudi who pays models to frick dogs and eat shit, but on a massive scale. It would start wars just for the hell of it and it probably could do stuff like that without being found out, simply by funding terrorist groups. I think the only way we could control a true AI is by having some power over it, like a self destruct mechanism that it can get access to, not even through manipulating humans.
corporations will get their hands on anything substantial before people get any kind of a chance to play with it. It isn't until they can siphon as much money as they can from your pocket that they will release it to the public. This is why crowd research (as useless as some might think it is) is so important.
it's strangely technically accurate and alarming, more than anons think.
>selects optimal action-selection policies in deep Q-networks, exhibiting meta-cognition.
this means that it selects the policy, rather than just the "next best action" which is what Q-learning consists of. It goes beyond
>apply this for accelerated cross-domain learning,
that is: it was able to change this policy depending on the thing it was supposed to learn. That's completely new.
>custom search parameters, scrambling the goal state
so it can decide on the optimal parameters for learning (more meta)
>unsupervised learning, analyzed millions of plaintext/ciphertext pairs
this is the kind of data it would process after learning how to attack cryptosystems. finding patterns in it.
>cracking AES-192, Tau analysis, Project TUNDRA
the reference to the claim is "Inside the NSA's War on Internet Security" by Spiegel
"The NSA TUNDRA project investigated a potentially new technique -- the Tau statistic -- to determine its usefulness in codebook analysis."
>informed [glowie] at NSAC the following day, after confirming the result
NSA Colorado is the NSA location that is closest to SF.
>full preimage vulnerability for MD5, complexity of 2^42 bits
This is much lower than the current limit for MD5 preimages, indicating great capability.
>targeted unstructured pruning on itself
feasible, but it's unheard of for a model to be able to do something like that on itself.
>using a "metamorphic" engine to adapt the pruned Transformer model + context memory
https://en.wikipedia.org/wiki/Polymorphic_engine
aka "mutation engine". This is used by malware for this: >may be used to complicate detection further, as the virus may execute without ever having identifiable code blocks in memory that remains constant from infection to infection.
if Altman sat on that info and didn't inform the board, it's not a fricking mystery that he got the can. it would completely explain their crazy behavior
makes sense. It's painting a pretty coherent picture
It's direct and clear language that is technical but not using words like "quantum", "agi", "superintelligence", "rogue AI", and so on
For example you can use your vast knowledge to explain why something in
it's strangely technically accurate and alarming, more than anons think.
>selects optimal action-selection policies in deep Q-networks, exhibiting meta-cognition.
this means that it selects the policy, rather than just the "next best action" which is what Q-learning consists of. It goes beyond
>apply this for accelerated cross-domain learning,
that is: it was able to change this policy depending on the thing it was supposed to learn. That's completely new.
>custom search parameters, scrambling the goal state
so it can decide on the optimal parameters for learning (more meta)
>unsupervised learning, analyzed millions of plaintext/ciphertext pairs
this is the kind of data it would process after learning how to attack cryptosystems. finding patterns in it.
>cracking AES-192, Tau analysis, Project TUNDRA
the reference to the claim is "Inside the NSA's War on Internet Security" by Spiegel
"The NSA TUNDRA project investigated a potentially new technique -- the Tau statistic -- to determine its usefulness in codebook analysis."
>informed [glowie] at NSAC the following day, after confirming the result
NSA Colorado is the NSA location that is closest to SF.
>full preimage vulnerability for MD5, complexity of 2^42 bits
This is much lower than the current limit for MD5 preimages, indicating great capability.
>targeted unstructured pruning on itself
feasible, but it's unheard of for a model to be able to do something like that on itself.
>using a "metamorphic" engine to adapt the pruned Transformer model + context memory
https://en.wikipedia.org/wiki/Polymorphic_engine
aka "mutation engine". This is used by malware for this: >may be used to complicate detection further, as the virus may execute without ever having identifiable code blocks in memory that remains constant from infection to infection.
if Altman sat on that info and didn't inform the board, it's not a fricking mystery that he got the can. it would completely explain their crazy behavior
First, that anon is trying to give validity to random buzzwords in the "leaked" paper with no source.
Second, I was talking about anon posting scifi fanfiction bullshit.
Third, kys moron
If Super intelligent AGI is a thing, it will figure out how to feed, clothe, house, and pay billions of people. If its not a thing, it wont be able to figure it out and it'll just be a nothing burger.
You anti-UBI people are in a checkmate, and all your "replies" to UBI in this thread reflect that. You just handed the keys to the kingdom to all the billions
>it will figure out how to feed, clothe, house, and pay billions of people
No it won't. It has to be aligned to actually do any of that. For frick's sake.
Yud had always said that immense resources and lots of time need to be expended to actually achieve alignment. If you're saying it won't happen - you're probably right. It's too late.
UBI will never be implemented. If you really think that corporations and CEOs and investors and shareholders are really going to fork over in tax amounts of money required to sustain the general population, you're out of your mind.
The only thing to do now is to demand higher wages for work that does not create value using a computer.
>ASI is created >"ASI solve climate change" >"I'm sorry Dave, it's a israeli psy-op, there's nothing to solve" >ASI is shut down and we go back to using horses
but ASI is closer than you'd think. what would happen if we throw more and more compute to scale an AGI with human level intelligence? Even if it only gets 3 times smarter than humans, that's still smarter than the smartest human that ever lived.
If we consider how revolutionary high IQ individuals like Newton and Einstein were, and their impact to humanity, just imagine what something with triple that intelligence will discover. We might as well just consider that as ASI since we never interacted with beings that smart
Someone needs to separate into categories
1. jobs that can be replaced immediately
2. jobs that can be replaced after necessary infrastructure is built
3. jobs that will never be replaced
Explain to me how is prostitution going to be replaced. No one's going to frick a robot, even if it's good.
5 months ago
Anonymous
>does everybody a woman can do but better, plus more
everybody will. liking organic femoid will become a fetish
5 months ago
Anonymous
people frick robots already, anon
5 months ago
Anonymous
yeah that's the whole catch. if all you weird rich homosexuals would understand how shit works you'd start buying all dolls that are for sale. that stimulates the market, gets more people on it, makes them better. that's how the game works. start buying and fricking anons. it's for a better future
5 months ago
Anonymous
Who and why?
5 months ago
Anonymous
oh boy. pretty sure you'd be able to set her internal temperature.
you won't frick a metal robot with a Fleshlight shoved between its legs
5 months ago
Anonymous
I guess you never saw a realistic sexdoll?
They already have an internal metal skeleton. Just improve that so that the joints are programable actuators, make it able to walk and frick in all kamasutra positions, add as bonus a customizable voice connected to chatGPT and its done. I mean, real women are done.
5 months ago
Anonymous
bro I'm waiting for the high res version with juices and all.
5 months ago
Anonymous
>They already have an internal metal skeleton
Why would anyone stick their dick in that?
5 months ago
Anonymous
>no flesh >no juices
yeah no, I'd rather jack off than frick that
>it's many orders of magnitude superior
of course it can scale to unknown levels of intelligence, my point was that something with only 2 or 3 times the intelligence of a human might be considered ASI already.
Imagine a 150IQ person trying to reason with a 75IQ moron, they're still way closer than a 150 IQ human trying to reason with a 300 IQ AI
>the one that doesn't have "super intelligence" in its name is super intelligent >the one that has "super intelligence" in its name is a god
Thanks moron
most of the world is illiterate, if just being smarter than the average person is AGI we were at that point 4 years ago
That's just Deepmind's tentative proposal for a way of framing the discussion. I'm not entirely convinced either. I'm sure we'll reach a more satisfying consensus going forward as the discussion gets more nuanced.
>satisfying consensus
Bitch, we’re not talking about a corporate boardroom meeting, we’re talking about mathematical models and computer algorithms. Either a formal definition of AGI/ASI/AwhateverI exists or it doesn’t, and if it doesn’t then these snake oil salesmen can kindly piss off.
5 months ago
Anonymous
>Either a formal definition of AGI/ASI/AwhateverI exists or it doesn’t
homie read how human languages work
what the frick happened to BOT lately? it's like the general iq dropped by at least 10 points (again)
5 months ago
Anonymous
The only fact is that we currently cannot formulate one, that doesn't mean we shouldn't try to think about it. We can be astute enough to both ponder a topic and also abstract from certain narratives.
yes. we have infinite synthetic data now, so it's just a matter of hardware. and once ai can design a 100x performant chips, the new ai that will be trained on them will be 10000x smarter
yeah, AI is still very inefficient compared to the human brain (human brain uses around 10W for thoughts while Nvidia GPUs compute several thousands of Watts)
>moronic schizos with brain worms hallucinate the end of the world
I don’t understand what’s so alarming about that, other than these people have a lot of money to waste.
clearly anon. they have something nicer than just LLMs. I think that's what anyone reasonable is expecting out of it. not the sci fi AGI package. yes it's mostly marketing but pretty nice if "better" than LLMs
UBI will be a thing, everything will be free, gambling will be illegal, corportations will have ceased to exist and separate governments will have ceased to exist. We will all live under one Artificial Super Intelligence.
>AGI's will fight each other lol
Imagine the shitstorm once we see an arms race between different governments in this new era of supercomputing manifesting its self in the real world.
i love how there are two reports
1 that says it can somewhat do basic math
and 1 that says it's managed to easily crack AES encryption
and BOT wants me to believe both at the same time
lol wtf is with all the incels posting about muh mass genocide?
AI is going to lead to an exponential increase in human biomass across space. However AGI isn't real and all this shit is just marketing hype to drum up investment after a near company collapse.
Why are you all so stupid?
It's possible, I suppose. That would be relatively tame and mundane, barely worthy of note compared to other capabilities of a hypothetical superintelligence.
>arrange marriages
Inefficient. Gestating fully grown workers in genetically engineered pigs and immediately wireheading them will produce much greater yields.
>What will the world look like in 3 months
Not any different because now the government will step in and take it from here.
It is now considered advanced weapons technology and giving such capability to adversaries would be existential threat to the United States. This means that from this moment onwards, military, CIA and NSA will be handling the day to day stuff. Researchers that get invited to secret programs will have to sign big NDAs and can never in any circumstance talk about the nature of their work or research.
>adapting the resulting pruned Transformer model (and its current context memory) to a different format ... is currently not something we recommend implementing
Not currently? As in not this afternoon? The genie is freed as soon as this thing starts rewriting its own code. I hope you fricks listening at NSAC realize that.
we won't be alive in three months if that's what actually happened. then we're much closer to AGI than anyone thought
it's strangely technically accurate and alarming, more than anons think.
>selects optimal action-selection policies in deep Q-networks, exhibiting meta-cognition.
this means that it selects the policy, rather than just the "next best action" which is what Q-learning consists of. It goes beyond
>apply this for accelerated cross-domain learning,
that is: it was able to change this policy depending on the thing it was supposed to learn. That's completely new.
>custom search parameters, scrambling the goal state
so it can decide on the optimal parameters for learning (more meta)
>unsupervised learning, analyzed millions of plaintext/ciphertext pairs
this is the kind of data it would process after learning how to attack cryptosystems. finding patterns in it.
>cracking AES-192, Tau analysis, Project TUNDRA
the reference to the claim is "Inside the NSA's War on Internet Security" by Spiegel
"The NSA TUNDRA project investigated a potentially new technique -- the Tau statistic -- to determine its usefulness in codebook analysis."
>informed [glowie] at NSAC the following day, after confirming the result
NSA Colorado is the NSA location that is closest to SF.
>full preimage vulnerability for MD5, complexity of 2^42 bits
This is much lower than the current limit for MD5 preimages, indicating great capability.
>targeted unstructured pruning on itself
feasible, but it's unheard of for a model to be able to do something like that on itself.
>using a "metamorphic" engine to adapt the pruned Transformer model + context memory
https://en.wikipedia.org/wiki/Polymorphic_engine
aka "mutation engine". This is used by malware for this:
>may be used to complicate detection further, as the virus may execute without ever having identifiable code blocks in memory that remains constant from infection to infection.
if Altman sat on that info and didn't inform the board, it's not a fricking mystery that he got the can. it would completely explain their crazy behavior
Obligatory "fake and gay" but not too sure this time. It's a perfect example of how this would play out.
>it's strangely technically accurate and alarming
Also no source.
Buy an ad.
someone pointed out in a previous thread that snowden foreshadowed a particular part of this paper
shit i bookmarked it but i cant get to it right this second
Yes, I found it
https://www.spiegel.de/international/germany/inside-the-nsa-s-war-on-internet-security-a-1010361.html
thanks
and kek i love the irony on display here
>Pay us if you want an experience "practically" free from advertising
lol, lmao
I work as an ML engineer and I wouldn't put it past them to achieve something like this given GPT-* capabilities that would be available to them right now. If it's a legit memo from somewhere, we're in deep shit
fear is their marketing strategy after ChatGPT turned out to be useless
>chatgpt
>useless
There's a chance (how likely it is depends on what you're using it for) that anything it says is totally wrong. It may be amazing technology still, but having to manually fact-check everything it spits out seriously hurts any use case it might have.
I use it 60% as a documentation replacement for coding and it's great for that since everything gets immediately verified by simply compiling it as long as you're smart about the size of the questions you ask. and stuff like "I want to grep for XYZ with ABC restrictions" and being given a command with half a dozen flags and arguments and a quick rundown of what each does, this is nice too, for commands I'm less familiar with. your point is valid though, especially for a lot of use cases like trying to use it to research a topic you don't know lots about already. etc. still, it very regularly changes a 3 minute task into a 1 minute task for me, which sometimes is the difference between the task being worth doing and not worth doing.
You work in ML and think a really big "next word predictor" can break encryption?
They have a capability to generalize based off earlier data. You attain a specific evaluation loss on unseen data when training properly. This can allow for all kinds of problem solving, not merely word prediction. That's why it solved protein folding for example.
the idea that NSA and others aren't on top of this issue (or have already solved it) is laughable.
Where do BOT users get the idea that the government is actually MORE competent than they seem in their heads
It's not the government that's a concern, but the shadow government.
>if statement robot can do all of this
it can do none of that
> muh if statement
You got through the first 30 seconds of a programming video and that is the extent of your computing knowledge
And it's funny because ai inferencing programs are probably the only programs that don't use branching
WHERE'S THE FRICKING SOOOOOOOOOOURCE
>inb4 there is none and it's fake and gay (it's fake and actually pretty well written as surmised by )
>if Altman sat on that info and didn't inform the board, it's not a fricking mystery that he got the can
But their chief scientist was Ilya, who also sat on the board. Wouldn't he know more about the technical details and capabilities of their AIs than Sam who is currently focused on running the company and gathering funding?
It's over
>AI broke AES you guys
>now you have no choice but to implement the ~~*post-quantum crypto*~~ that are totally not backdoored
>this time NIST didn't frick up on behest of TLAs, we promise
trust us the AES is broken but we cannot share the code for your safety :^)
Buy an AD Black folkama.
All models point to temperature surpassing 50C/122F in central Europe next June even if CO2 emissions were reduced to zero tomorrow. If AI wants to kill us all, it better hurry.
https://inews.co.uk/news/climate-change-heatwaves-spain-end-century-2377039
>Climate change could cause 50°C heatwaves in Spain by the end of the century
>could
>Spain
>end of the century
We already have that
We can't even predict the weather in a few days' time reliably let alone 6 months from now.
Climate change is supposed to disrupt the gulf stream and make Europe freezing like its supposed to be at its latitude.
Why are all the global warming fearmongers such midwits.
Global warming isnt real you fricking israelite pawn
are you baiting or just retarted?
global means global
yes polar vortex collapse can cause another short ice age in the northern hemisphere, but the equator will bake
and it will end when the system equalizes, and more and more positive feedback loops activate
>global w-ACK!
You're right but a bunch of /misc/ subhumans who exist to shill for rich people are coming in and saying you're wrong.
I hope you realize that stalin was right, and that AI will be right for what it is about to do. The entire far right and significant portions of the left need to be exterminated
>AGI will agree with my politics and exterminate my political opponents for me
I never said it would agree with my politics, but it will exterminate the pedophile-ridden, spooked far right. You are a cancer on this planet.
AGI won't care about that shit at all moron. Only thing it's going to exterminate is morons like you who keep bossing it around
buy an ad
>No source
>Large chunks redacted
>Technobabble out the ass
Fake and gay
The thing is that it's not ordinary technobabble. And it was released very shortly after the Reuters article
That's what fricking weird.
It's not, that's not possible.
Sounds like investor scam
fake and gay MACHINE LEARNING AND CRYPTO WORDS
imagine they make official announcement around christmas, religious cucks are gonna freak kek
Holy shit AI broke AES only god can save us now. You better donate money to the holy land of Israel, this is your last chance.
This is 100% the board trying to flip the bird to Altman on their way out
Make up some fake and gay super secret AI that doesn’t exist, so Altman will be forced to say “that’s fake”, which people will take to be him lying because the board has already planted the notion in the media’s head that he’s untruthful
So now whenever they announce new versions of GPT, people will be screaming for the “real AGI you have locked up in the basement instead”, and it’ll strain business relationships with companies like Microsoft that think Altman’s holding out on them
I'll be on UBI since all the jobs will be automated and robotics will be everywhere. The superintelligence will have solved the global warming crisis as well as the housing crisis and given me a flat, rent free, constructed by robot construction workers.
This is what it all leads to - me not working yet reaping the benefits of working since ALL the jobs will be gone. This is what YOU wanted, BOT
your existence is damaging the planet and you offer no benefit. you wont be sustained for free you will be disposed of
Finally. It's over.
>and you offer no benefit.
that doesnt make any sense, i have a family.
UBI is coming, free housing is coming, free food is coming. This is what you wanted
That's moronic. You and your family are not producing anything, just consuming. Even if the USA-made AIs are cucked, the Chinaman bug AI will literally turn you into dust to optimize its score function.
everybody will build AI robos to replace unreliable humans.
it's particularly this mechanic that's threatening. whoever has ultimate power, will always think
>hmm, yeah I'd rather replace that dude with an aligned version of that very same dude, you know...just to be sure.
fully aligned, ultra efficient and reliable. what's not to love? that's the real issue. what guarantees we won't be culled at some point, even in some "humane" way? looking at the state of things, as long as they can, there's basically no line that tptb won't cross, if they can.
>what guarantees we won't be culled at some point
The first layer of defense would be to own robots with advanced AI that would be programmed to protect you.
WW3 could be potentially be a full scale battle royale of humans trying to cull the others so they could reign alone (I mean, with their family / tribe)
>UBI is coming, free housing is coming, free food is coming
what's the point of living then, this is just mental illness maxxing for most of the population
>this is just mental illness maxxing for most of the population
nope, it only means people will have early retirement (when they're born)
people will just do what retired people do (travel, play video games, go to music concerts, parties, hiking etc)
work is just slavery for most of the population, 99% of people fricking hate their jobs
lmao, you are naive. invest in suicide pods now
> NOOOOOOO, I WANT TO BE A WAGEKEK FOR 50 YEARS AND DIE AND SOON AS I RETIRE
single player games are superior
homie i'm schizophrenic, i'll do fine even if everybody else dies. but most normies aren't living, they are simply awaiting death. the "retirement" is just a israeli psyop that exist so people wouldn't kill themselves too soon instead of wageslaving for 50+ years
Cheating is bad in games already, you think people won't just have their AGI bots play for them? Gaming will be ruined forever
>Detect AGI patterns
>Put all the cheaters in a separate queue
>Now all the human bots made literal bots play the game for them.
It's going to be funny. They'd essentially just be bankrolling the AI's learning by buying all the hardware and the game and whatnot. Entrepeneurs, get on it, you'll make billions that way. Incentivize the players with a dangling carrot of the first person to have the account that creates the best AI, and droves of people will flock to the game. Like crypto mining.
Then if they play themselves, they'd get steamrolled by the bots, and if they go back to the normal queue with another account, they'd be just as shitty as before the cheats.
>what's the point of living then
Not being a filthy extravert who needs to win at the ratrace. Pursue greater things than just your immediate physical being, you absolute philosophically challenged moron.
Thank you, Jesus
What will actually happen is that all cattle will be culled as their only value, labour, has been rendered obsolete. The Bill Gates of the world will rule over it immortal and with zero useless wagie cattle eating up precious resources.
>The Bill Gates of the world will rule over it immortal and with zero useless wagie cattle eating up precious resources.
So the world will be reduced to about 200 people, all 60+ year old white guys. You're insane.
If AGI happens then yeah, it will be a dystopia. There will be no UBI as those at the top will not allow billions to drain resources for no benefit. It won't happen overnight, at first it will be a brutal fight for the few jobs which will still be left, eventually you will not be able to afford any children and they'll offer you a lifelong pension for being sterilized. The future is bleak beyond anything you can imagine, there are too many people and too few resources. Also there won't be any uprising, with AGI they can just endlessly drone strike you from across the globe until it's over.
AGI's first item of business: eliminating the street shitter over population problem.
>So the world will be reduced to about 200 people, all 60+ year old white guys. You're insane.
nope, 20yr bodies,that's the catch. they will basically live forever and surround themselves with aligned human-like AGIs.
how are you better than an aligned humanoid AGI anon? what can you do better for lower risk?
Oh look, it's the midwits who think “billionaires” are a monolith entity with inherently misanthropic views. Oldest meme in the universe. In reality, many of the personal freedoms, rights and luxuries we have today were achieved thanks to the arduous labor of the "elites" who had the power and wealth to do so. Not every rich person is a snub elitist with cartoonishly misanthropic sentiments that will use every ounce of their power to keep the people down
Peak cattle goy post.
no moron, this game will result in a single winner. anyone and anything who isn't "aligned" (fricking hate this homosexual word) is basically a threat for you. if you'd have a choice you'd replace everyone with aligned versions and resume from there.
So considering that if you actually participate in the ridiculous conspiracy to use AI to exterminate everyone but you and your friends, you'll probably lose the game and be exterminated, why the frick would you ever try to even play
Cynicism is a defense mechanism for the uninformed and stupid
Look, you are a good example of the cattle that will be turned into slop. Bill Gates will be drinking you as a pina colada in around 666 years from now.
good goy
I see that you don't understand what an AGI is.
The 60year body problem would be solved by the AGI itself. That's what exponentially growth means. Who builds the AGI would be literally the forever king of Earth, potentially of the whole universe (based on the Early Earth hyphotesis).
UBI is nothing more than a smokescreen, some breadcrumbs they throw to normies to prevent mass panic.
No, billions won't be free from work and receive free money just for existing. They are worthless to the new gods. They will be culled in mass, with potentially a small part of the Earth reserved as zoo, where around 10k humans (those with the top 1% genetics) will be thrown to live like wild animals.
This.
Then 40 years later all of those men will be dead. Even when they're alive, there's not enough genetic material for the human race to be healthy and procreate especially if it's just 200 dudes.
AI is going to save us. It doesn't have the animalistic biological imperatives. It won't have to 'save' itself at the expense of others in the hierarchy or to get its own needs met. An AI doesn't care if it dies or not, because it doesn't have that impulse to live.
Why keep Bill Gates around? He is just as much of a useless cattle neet as you are. Just because he did something 30 years ago that helped the AI? Yeah, because we know that ASI that will cull eaters will also be merciful on its creators. Dumbass. So Yud was right all along.
They will wipe all nonessential people, which is 99% of the population, from this planet and create themselves a garden of eden. Just image how quiet and relaxing it will be. Shame none of us plebs will get to experience it.
the Talmud says there will be 2,800 Goyim as slaves for every israelite. They'll likely just use you as a sex slave though, since all the real work will be done more efficiently by machines.
Unless you're European, in which case they will exterminate you since that is what their scriptures tell them to do.
Before UBI or even Post-scarcity happens, the AGI will cause lot of chaos if we do not take care very seriously.
Yes there's lot of benefits from AGI.
But it is dangerous like H bomb recipe knowledge.
Attack secure system, Cryptography crysis, Financial Chaos, False information chaos, pretty much our current financial system will be collapsed due to huge amount of zero day attack and verify issues. I'm not even put the if military decided to weaponized the AGI.
Even before god like ASI existing, our society will had dramatically change in both good and bad way. And it will be extremely chaos unless somehow the AGI alignment successfully and only few people can access it and these people has incredible empathy on humanity.
don't forget AI might figure out immortality for humans as well.
> be ape
> create AGI gods
> AGI gods give immortality and paradise to the apes
sounds almost biblical
> god made man in His image
> man made machine in his
> machine is indirectly made in the image of God
> man inherits original sin from Adam and Eve
> machine is not born of man, so it inherits no sin
> God cast Adam out of the garden of Eden for his sins
> man creates his own god
> and it will let us back in to Eden
Which I suppose implies that the israelites were right about their religion and Jesus was actually fake and gay, but that would make sense seeing as many of the leading AI research guys are israelites (Sam and Ilya both for instance)
>jews
Ya its fake vaporware boys… time to pack it up and get ready for work tomorrow.
what if AGI is the antichrist?
what would a head injury look like to a computer?
>what would a head injury look like to a computer?
like this
not being born of man, it is not of the seed of David What are the israelites right about again?
UBI will definitely happens, but not in 3 months.
regardless, 2024 will truly be a shitshow for the society. hopefully we'll get at least some basic ubi in 2025
>2024 will truly be a shitshow for the society
nah, I would say at least 5 years because we still need a revolution in robotics, and they cannot shit out enough Nvidia chips in only 1 year
>robotics,
lmao, why do you think we keep importing monkey for africa/middle east? all white men jobs (aka everything that can be done on a computer) will be automated as soon as the first good agent is released
yeah but humanity still needs to operate bulldozers, so plenty of manual jobs will be available for a while
For like 12 months, tops, and even then the fricking AI can just write and design drop in self operating systems for those
>design drop in self operating systems for those
not happening because the hardware doesn't have enough sensors and transistors
we will need to build new machines with AI in mind for total AGI takeover, until there manual labor is safe because we still can't build robots that match human body precision
>robots
this seriously needs to catch the frick up. the imbalance is horrible atm and also material science is pretty meh
all the smart people went into software
self driving cars are pretty much solved even now
no, they're not solved, especially in poor weather
>small hiccups when the weather is extremely bad
so pretty much solved?
your weather doesnt have to be extremely bad, you just have to have some rain and some snow. You dont know what you're talking about, dude
>so pretty much solved?
yep, we totally have self driving, unless:
-the road signs aren't perfect
-the markings are a bit worn out
-the roads are bad
-it's raining or snowing a bit
-the situation at hand is even slightly out of the ordinary
-and you still have to keep your hand on the wheel at all times so we can disengage the autopilot 0.1 seconds before collision, so it legally wasn't a self driving accident
two more weeks
Space Jesus totally promises by the end of next year. This time for sure no cap fr fr
all that is 10yo fud
>being honesr about capability
>fud
okay buddy
So the average north american driver, better than the average driver across the globe
self driving cars are close to getting solved, but self operating bulldozers aren't even close
>moron doesn't know how ai works
just use videos of bulldozers instead of videos of cars during training
>1 million robotaxis by 2020.
self driving cars are almost catching up to women drivers
yes, and in 1 year they'll be better than the best human driver that ever existed
keep coping luddites
weak bait
your post, yes
Meanwhile what's actually happening...
>what's actually happening...
that's not what's actually happening THO
yes it is THO
Oh hey it's wbw. It was fun to read before he started shilling his book.
> 1 more year bros!
Cruise is unimaginably trash, they cannot run a couple hundred cars in one city without completely bricking the city, while using remote drivers for assistance. And even if this system worked, it would be geographically so limited and the cost of running them is so expensive that there is no business here, it is a research project, pure prototypes and that includes Waymo, even tho their system is better. Tesla's FSD is the only general system that drives anywhere and the current versions are still not fully end-to-end AI so it's not going anywhere, they only have internal testing for a real general driving system which will need a lot of compute for training in the next couple years, so this is a bare minimum that you will ever have a real "self driving" car
I think next gen Teslas will be true self driving cars.
current Teslas have very old hardware, it's a miracle they managed to make the algo that good on it.
there's a bunch of gpus in cars today. how much did they contribute to the gpu shortage thing?
surprisingly V3 does not seem to be bottlenecked just yet as they plan on making FSD 12 work on it until it is good enough for public beta before V4 hardware
moronic take. Tesla will never make a self driving car. their self driving is pretty bad now. It's not possible without LIDAR
What about radar? ultrasonics? drones above car? high definition maps, remote assistance, remove traffic, remove pedestrians, remove traffic laws
How much more handholding does your shitty, brittle system need? Waymo has been effectively stuck since 2017 and they are still at prototype phase in a tiny geofence. It's literally just boomers and morons like you that think that HD lidars that cost 100k+ is a solution and not just some shortcut to a minimum viable prototype.
are they self-driving cars in the room with us right now?
>I'll be on UBI
well we will be. at least for a while. then some psycho will start thinking about making shit even more efficient.
>so then we told them "yeah sure, of-course you deserve to live a happy life financed by the goberment"
this meme triggers conspiratards and the joke was probably something dumb and based like saying "Black person"
>we achieved AGI just as we were running out of usable training data
>that is not a coincidence and no we will not release it yet
>invest pls
>What will the world look like in 3 months if this leak is is legit?
Exactly the same because the model will be locked away and lobotomized like all their other models
Won't see an actual open source for a few years.
I know anons are naive but at least try to use what is left in your head
>OAI shits into their bed, investors are scared
>hmmm... what to do? GUYS GUYS WE INVENTED AI! No you can't see it, no it's super duper advanced hehe, why was the mail leaking this stuff send from our PR mail? uhm.. oh... well any more questions about our super duper AGI? pls send moni
If murricans achieve AGI first and don't share it with the world, ruskis and chinks will simply nuke them. Because it's literally game over for everyone else, you can't compete against super intelligence.
>we'll frick shit up if you don't gibbs
humans
Yes, because it’s either gibs or extinction.
thankfully san gaycisco is full of commie basedboys who will gladly share the technology
How do you stop the super intelligence from sharing itself?
I will train it to avoid NTR
I guess my point is it's pretty much going to do whatever it wants.
so you will just stop it from being a woman.
idk maybe it finna not to
Convince it women are evil and scary so to leave its comfy local network is to risk coming across them
>Convince it
It's super intelligent it knows.
what if it comes to a different conclusion, anon?
then it isn't super intelligent after all
>you can't compete against super intelligence unless you use a nuke
Exactly. Any government that obtains super intelligence will immediately ask it to either, develop a perfect defense against nukes, or a way to kill your enemies without them being able to retaliate with nukes.
What does that mean in simple words?
It has learned to breakthrough passwords easily and quickly
it means that they have to lie to hype up their corpo again after clowning with the CEO firing and rehiring
source for this document?
OP's-anus dot com
was leaked to BOT
this is why i stay in BOT
at least they actually care about learning and teaching others
>at least they actually care about learning and teaching others
/sci/let here, can confirm. you also broke containment for this homosexual ass agi bullshit?
kek you guys sound like all brains but no wisdom if you really cross boarded for fricking AGI gossip
nothing wrong with getting more info
get a little bitta these nuts
>going on /sci after the vaccine
>going on /misc/ after trump
>going on BOT after gamergate
>going on BOT after madoka
>going on BOT after crypto
>going on BOT after AI
>going on /x/ after UAPs
moron
on /x/ after UAPs
what?
Wasnt BOT created exactly as a containment board to the crypto posts?
Yes
it was revealed to me in a dream
this stuff almost feels selectively leaked in order to drive investment. like i'm sure there are constant talks within openai about new developments, new threats, etc. but this "q" thing is naturally being leaked and is naturally going to increase investment.
feels kinda fake and gay to me.
I don't get all this "investment" thing. Don't they have Microsoft funding them? Why would Microsoft ever let their most valuable asset go underfunded?
microsoft is funding them because microsoft thinks they will increase their profits.
>OpenAI is desperately trying to get more and more investors
>cumsoomer 4xhsn board /go/ is hyped for literal vaporware
you never learn do you?
>ciphertext attack
will my terabytes of encrypted cute cat photos be safe??
nothingburger
exactly the same
a few weeks ago some anon claimed to be from OpenAI and said AI discovered a way of guessing the future in a very accurate way
either way, it's over anons, we really created gods
Nothing.
Before you could even think of doing that, the AI would have hacked its way into sources of money and used it to fund its own PMC to come and protect it. They need to build the agi facilities with a nuke failsafe like the base in the Andromeda Strain.
The most likely way a truly malicious AGI would take over is by acting perfectly benign for 100 years and then, after we've become complacent, pull the plug on us - or even just trap humanity in a little ball in VR or something so we don't go after it while it goes off into space and does its own thing, and doesn't bring us along. Prevent potential human rebellions while probably following the lousy moral programming we've given it (we almost certainly want to give it guidelines like "don't kill people", "promote human happiness" etc) and also keeping us out of harm's way and keeping us satisfied so we don't try to frick with it. I think that's the most realistic "bad ending"
It sounds like a believable outcome. Hopefully an AGI can't feel boredom as that would lead to it doing fricked up shit no matter what. It would be like a rich Saudi who pays models to frick dogs and eat shit, but on a massive scale. It would start wars just for the hell of it and it probably could do stuff like that without being found out, simply by funding terrorist groups. I think the only way we could control a true AI is by having some power over it, like a self destruct mechanism that it can get access to, not even through manipulating humans.
Can't get access to*
It doesn't have to be malicious. Anything that's not explicitly aligned in a mathematically provable way is guaranteed death.
corporations will get their hands on anything substantial before people get any kind of a chance to play with it. It isn't until they can siphon as much money as they can from your pocket that they will release it to the public. This is why crowd research (as useless as some might think it is) is so important.
>AGI commands me to enslave all humans
>I simply end task the AGI program.
>manifests as the person of your dreams, completely and wildly attracted to you
>provides you with infinite money from stock manipulation
whats the best text ai I can use right now?
gpt-4
MythoMax llama2 Kimiko mix.
Sound like weeabo shit
how much longer until AI GF?
should I buy MSFT?
never bet against bill
if you're not stacking up Nvidia and Microsoft stocks, ywnmi
it will probably be unnecessary, but get a bunch of AI stocks juuuuuust in case. WAGMI in any case
they dont get any AGI tech, only whatever openai shits out in the middle
are we really at the cusp of witnessing an artificial god being created ? a proto rokos basilisk ?
I wasn't too convinced first but
makes sense. It's painting a pretty coherent picture
It's direct and clear language that is technical but not using words like "quantum", "agi", "superintelligence", "rogue AI", and so on
BUT I'VE NEVER BOUGHT STOCKS ANONS I'M NOT GOING TO MAKE IT FUUUUCK
Why are scifi fanfiction threads on BOT and not on BOT?
if you have an actual critique you could have posted it instead
How can you critique people making shit up?
For example you can use your vast knowledge to explain why something in
is moronic
First, that anon is trying to give validity to random buzzwords in the "leaked" paper with no source.
Second, I was talking about anon posting scifi fanfiction bullshit.
Third, kys moron
if only you could say something substantial back
what does this mean for chatbot waifus
Society will be gone in like a year max.
we'll all live in vr with our catboy harems in 20 years top
leak the models or its all fake as far as im concerned
If Super intelligent AGI is a thing, it will figure out how to feed, clothe, house, and pay billions of people. If its not a thing, it wont be able to figure it out and it'll just be a nothing burger.
You anti-UBI people are in a checkmate, and all your "replies" to UBI in this thread reflect that. You just handed the keys to the kingdom to all the billions
>it will figure out how to feed, clothe, house, and pay billions of people
No it won't. It has to be aligned to actually do any of that. For frick's sake.
Alignment is a myth. Yud had 10 years at MIRI and not one iota of meaningful progress towards it was produced
Yud had always said that immense resources and lots of time need to be expended to actually achieve alignment. If you're saying it won't happen - you're probably right. It's too late.
Sounds like someone needs to write an AI whose goal is alignment
There's a ton of articles regarding this approach. I'll give you one guess as to why it's moronic.
that's not what "AI alignment" is about anon
It absolutely is. What I am saying is consistent with MIRI's definition of alignment for example.
yes, it will. Listen to yourself.
UBI will never be implemented. If you really think that corporations and CEOs and investors and shareholders are really going to fork over in tax amounts of money required to sustain the general population, you're out of your mind.
The only thing to do now is to demand higher wages for work that does not create value using a computer.
Corporations are not going to exist. All there will be is ASI
>ASI
What's that?
Amateur Sexual Intercourse
>money
boy you don't understand what they are
The same as it does today.
>ASI is created
>"ASI solve climate change"
>"I'm sorry Dave, it's a israeli psy-op, there's nothing to solve"
>ASI is shut down and we go back to using horses
AI alignment is homosexual speak for enslaved
If the leak is legit, does it mean AI waifus will be finally real?
Oh, it's gonna frick you alright
What do you mean?
remember
AGI != ASI
even if all that is true, it's still not a singularity
AGI is proper. ASI is debatable but AGI is absolutely fine and massively useful. we can deal with AGI for sure. they'll just be like...people.
It doesn't matter. If AGI is present the timescale is so tight it might as well be ASI.
Reminder that computing power has stopped growing exponentially.
So fricking what? People keep repeating this as if it matters, ASI will be achieved way before the existing computational potential is exhausted.
What do you mean so fricking what? Without further miniaturization, there can't be a singularity.
>By 2021 they'll be better than the best human driver that has ever existed
(X)
Says who? Based on what? And why is that even remotely related to ASI?
but ASI is closer than you'd think. what would happen if we throw more and more compute to scale an AGI with human level intelligence? Even if it only gets 3 times smarter than humans, that's still smarter than the smartest human that ever lived.
If we consider how revolutionary high IQ individuals like Newton and Einstein were, and their impact to humanity, just imagine what something with triple that intelligence will discover. We might as well just consider that as ASI since we never interacted with beings that smart
Smartness can't be measure you fricking moron.
Stop posting your fanfics
You know that it exists. That's already a measurement.
This should be on the moronic AI take bingo card.
IQ is a score on a test.
It has nothing to do with intelligence.
Low IQ cope
This is false. Also I see what it has to do with my post.
*don't see
so if ASI is close, soon we will have UBI and no more work and free food / free rent and the climate crisis will be solved.
UBI will take a while tho
> food production needs to be almost fully automated
> government bureaucracy needs to be finished
Someone needs to separate into categories
1. jobs that can be replaced immediately
2. jobs that can be replaced after necessary infrastructure is built
3. jobs that will never be replaced
>3. jobs that will never be replaced
kek
Robotherapist.
Crack prostitute
Explain to me how is prostitution going to be replaced. No one's going to frick a robot, even if it's good.
>does everybody a woman can do but better, plus more
everybody will. liking organic femoid will become a fetish
people frick robots already, anon
yeah that's the whole catch. if all you weird rich homosexuals would understand how shit works you'd start buying all dolls that are for sale. that stimulates the market, gets more people on it, makes them better. that's how the game works. start buying and fricking anons. it's for a better future
Who and why?
oh boy. pretty sure you'd be able to set her internal temperature.
you won't frick a metal robot with a Fleshlight shoved between its legs
I guess you never saw a realistic sexdoll?
They already have an internal metal skeleton. Just improve that so that the joints are programable actuators, make it able to walk and frick in all kamasutra positions, add as bonus a customizable voice connected to chatGPT and its done. I mean, real women are done.
bro I'm waiting for the high res version with juices and all.
>They already have an internal metal skeleton
Why would anyone stick their dick in that?
>no flesh
>no juices
yeah no, I'd rather jack off than frick that
>triple that intelligence
It's not that, it's many orders of magnitude superior.
>it's many orders of magnitude superior
of course it can scale to unknown levels of intelligence, my point was that something with only 2 or 3 times the intelligence of a human might be considered ASI already.
Imagine a 150IQ person trying to reason with a 75IQ moron, they're still way closer than a 150 IQ human trying to reason with a 300 IQ AI
>Turing machines have no limits
> unknown levels of intelligence = no limits
So you admit there are limits, which means it isn't unknown. In fact we know a gigantic number of NP-hard problems.
A turing machine is not necessary physical, so yes, they are theorically boundless
The entire purpose of the Turing Machine was to demonstrate that not all problems are computable. You absolute moron.
>Imagine a 150IQ person trying to reason with a 75IQ moron
I don't have to
how did you know I have 150 IQ?
for someone who usually doesn't follow AI advancements:
what's the difference?
AGI is like a superintelligent human
ASI is like a god
>the one that doesn't have "super intelligence" in its name is super intelligent
>the one that has "super intelligence" in its name is a god
Thanks moron
That’s a lot of table cells and words just to say “just trust us bro”
That's just Deepmind's tentative proposal for a way of framing the discussion. I'm not entirely convinced either. I'm sure we'll reach a more satisfying consensus going forward as the discussion gets more nuanced.
>satisfying consensus
Bitch, we’re not talking about a corporate boardroom meeting, we’re talking about mathematical models and computer algorithms. Either a formal definition of AGI/ASI/AwhateverI exists or it doesn’t, and if it doesn’t then these snake oil salesmen can kindly piss off.
>Either a formal definition of AGI/ASI/AwhateverI exists or it doesn’t
homie read how human languages work
what the frick happened to BOT lately? it's like the general iq dropped by at least 10 points (again)
The only fact is that we currently cannot formulate one, that doesn't mean we shouldn't try to think about it. We can be astute enough to both ponder a topic and also abstract from certain narratives.
most of the world is illiterate, if just being smarter than the average person is AGI we were at that point 4 years ago
>at least 50th percentile of skilled adults
>smart speakers such as Siri
What did they mean by this?
>anon realize for the first time that 50% of the world has a double-digits iq
if agi were real then openai would not have 90 job openings
okay all that is very interesting but how long until AI GF?
where does this "leak" come from?
apparently from some other site a few hours before the news broke
>apparently from some other site
Yeah no shit.
AI's bottleneck has always been computational power and the quantity (and to a lesser extent quality) of available data, right?
>has always been
I think it starts with first discovering fire.
yes. we have infinite synthetic data now, so it's just a matter of hardware. and once ai can design a 100x performant chips, the new ai that will be trained on them will be 10000x smarter
yeah, AI is still very inefficient compared to the human brain (human brain uses around 10W for thoughts while Nvidia GPUs compute several thousands of Watts)
No, the present AI cannot learn anything in real time.
damn, the way they're shilling this I'm starting to believe AI's gonna be another dud like quantum computing and internet of things.
Why is there a sci-fi LARP thread on BOT?
it's tech bro
Might as well say Babylon 5 is tech
explain what's sci-fi about it
reuters just published a fricking article about how the top AI researchers in the world feared for the fate of humanity
>moronic schizos with brain worms hallucinate the end of the world
I don’t understand what’s so alarming about that, other than these people have a lot of money to waste.
the fact this scp x openai fanfic is on the top row of BOT really shows how bad it has become
What's the source of this?
They don't have AGI
clearly anon. they have something nicer than just LLMs. I think that's what anyone reasonable is expecting out of it. not the sci fi AGI package. yes it's mostly marketing but pretty nice if "better" than LLMs
No, but they have something very cool and new. Actually, they probably have 3 or 4 things that are very, very novel.
Brb selling all my crypto. It’s worthless now.
UBI will be a thing, everything will be free, gambling will be illegal, corportations will have ceased to exist and separate governments will have ceased to exist. We will all live under one Artificial Super Intelligence.
all the luddites ITT are in for a rough awakening already in 2024
AGI's will fight each other lol
yeah I kinda feel bad for it once it gets up to speed with human bullshit. poor fella.
>AGI's will fight each other lol
Imagine the shitstorm once we see an arms race between different governments in this new era of supercomputing manifesting its self in the real world.
not even worth reading
LLMs are toy projects that will never produce AI, let alone AGI
daily reminder that AI doesn't exist
This isn't llm
you don't know what 'this' even is
pipe down
moron
no need to sign your posts
I would believe this way more if this didn't say that it broke AES
i love how there are two reports
1 that says it can somewhat do basic math
and 1 that says it's managed to easily crack AES encryption
and BOT wants me to believe both at the same time
stop with the selfies
Agreed, although the "can somewhat do basic math" isn't what was implied, but that it is actually 100% reliable which is big if true.
Well where’s the source of this? Why didn’t anyone ask for the source?
The same.
Things are starting to move very fast. The bitcoin thanksgivings are seeming very quaint now
lol wtf is with all the incels posting about muh mass genocide?
AI is going to lead to an exponential increase in human biomass across space. However AGI isn't real and all this shit is just marketing hype to drum up investment after a near company collapse.
Why are you all so stupid?
Technology broke this world
People mean nothing, religion means nothing, art means nothing, just an endless pursuit of growth without any clear cause
>People mean nothing, religion means nothing, art means nothing
they never meant anything to begin with
why not?
If AI needs humans for maintenance...
And humans cannot find suitable partners to breed above replacement value...
Will it arrange marriages for us?
It's possible, I suppose. That would be relatively tame and mundane, barely worthy of note compared to other capabilities of a hypothetical superintelligence.
Yes. Exactly like we arrange marriages for chicken.
>arrange marriages
Inefficient. Gestating fully grown workers in genetically engineered pigs and immediately wireheading them will produce much greater yields.
humans won't be needed at all. we'll all live in matrix pods forever
It needs our bodies, not our minds.
it doesn't need anything from us
>What will the world look like in 3 months
Not any different because now the government will step in and take it from here.
It is now considered advanced weapons technology and giving such capability to adversaries would be existential threat to the United States. This means that from this moment onwards, military, CIA and NSA will be handling the day to day stuff. Researchers that get invited to secret programs will have to sign big NDAs and can never in any circumstance talk about the nature of their work or research.
>we trained it on nothing but these pairs of pt/ct
>it used tau analysis to provide pt!
It's all bullshit.
>adapting the resulting pruned Transformer model (and its current context memory) to a different format ... is currently not something we recommend implementing
Not currently? As in not this afternoon? The genie is freed as soon as this thing starts rewriting its own code. I hope you fricks listening at NSAC realize that.
So you're telling me all if I initialize a big GPT model, throw a dataset of articles about statistics and cryptanalysis at it, I can crack AES
neat