OpenAI announces they've developed a truly sentient AI
https://news.ycombinator.com/item?id=38386487
We're all going to die.
Ahead of OpenAI CEO Sam Altman's four days in exile, several staff researchers sent the board of directors a letter warning of a powerful artificial intelligence discovery that they said could threaten humanity, two people familiar with the matter told Reuters.
[...]
The maker of ChatGPT had made progress on Q* (pronounced Q-Star), which some internally believe could be a breakthrough in the startup's search for superintelligence, also known as artificial general intelligence (AGI), one of the people told Reuters. OpenAI defines AGI as AI systems that are smarter than humans.
Edit: The linked /pol/ thread is kinda insane
Edit 2:
They're now trying to cover it up 
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
Q PREDICTED THIS
#WWG1WGA !schizos
Jump in the discussion.
No email address required.
This is basically what's going on.
Jump in the discussion.
No email address required.
More options
Context
This tbh. Muh breadcrumbs keklmao !slots100
Jump in the discussion.
No email address required.
More options
Context
More options
Context
NLP models are terrible at math so while this isn't a skynet level breakthrough it's still nice progress
Jump in the discussion.
No email address required.
They're not terrible at math, the can regurgitate the answer to any problem ever written in a textbook on math. You really think that a problem that would involve completely novel mathematical reasoning would be classified as 'grade-school level'?
Also note that the emphasis is on 'acing' the tests. They're getting more precise at regurgitating answers.
Jump in the discussion.
No email address required.
NLP on math still sucks. Asking ChatGPT about insights on a table of data still gets you r-slurred answers.
Jump in the discussion.
No email address required.
More options
Context
No surprise. The model was originally trained on thousands of unpublished text books. Then they poisoned it with random Reddit comments and quora answers.
Jump in the discussion.
No email address required.
More options
Context
Fair point.
Jump in the discussion.
No email address required.
More options
Context
lmao no. Sometimes they'll stumble into the right answer but more often than not it's all nonsense
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I literally don't know anything about the nuts and bolts of AI.
If NLP models are bad at math, why not just try to make them better at formulating problems to be solved by something like Wolfram Alpha? I know actual humans often struggle with that aspect of math, translating something from words to an equation. Do NLP models fail at that due to not actually understanding what is being described?
Jump in the discussion.
No email address required.
Because that approach is just a bandaid. All you're doing is glue-ing two methods with their own different weaknesses together, you aren't solving the weakness itself and it doesn't scale.
Jump in the discussion.
No email address required.
I always thought human thought was kind of a bunch of systems glued together. Like there's a structural “math/counting” module that some higher level control module can talk to when it needs to.
Broadly, source: I made it up, I have this intuition that AI needs a kind of “dumb but big” LLM-like framework on top of other modules that have some built-in structure. I.e. you kind of hard code certain things
Jump in the discussion.
No email address required.
Well you're a tomato, who cares.
That's what Q* is.
Jump in the discussion.
No email address required.
More options
Context
This is broadly correct. We know that the brain
includes various specialized areas for different 
purpose (visual cortex, hippocampus, etc.)
What's interesting
and poorly understood is that humans (or most humans, maybe not NPC chuds?) have some intelligent 
"self" that coordinates and integrates information from all those brain 
regions, and is capable of some amount of introspection and metacognition.
I haven't seen any indication of an AI that can do anything
analogous. These chatbots will lie and say they're thinking, because their training data includes human 
writing 
about thinking, but they are not thinking.
Brains are weird
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
You actually can do that with agents.
Jump in the discussion.
No email address required.
Lmbo. So it can't do math, but it can ask you to please execute this in a calculator. Maybe, if it's smart enough to know it can't do it on its own.
You ai appreciators are truly r-slurred.
Jump in the discussion.
No email address required.
Implying humans can do math. You're just regurgitating memorized multiplication tables for small numbers and use pen and paper to do long multiplication.
Jump in the discussion.
No email address required.
Unlike an LLM I can actually do that in my head.
Jump in the discussion.
No email address required.
Imagine having to use a subvocalization loop to remember ten seconds worth of digits. Meatbags smdh.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Yeah
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
They do suck at that, but it's also harder to make a composite approach like that portable. AI is being buoyed right now by how immediately commercializable LLMs are for enterprise systems/products. Relying on third parties doesn't make a product unsellable, but it does make it harder.
Jump in the discussion.
No email address required.
More options
Context
Doing the math is easy. Translating a problem to math isnt.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More details:
https://www.theverge.com/2023/11/22/23973354/a-recent-openai-breakthrough-on-the-path-to-agi-has-caused-a-stir
Jump in the discussion.
No email address required.
More options
Context
It might be that the AI is showing the capacity to solve maths and retain the knowledge in a way similar to how a child would, which would suggest that going further down the same path would result in the AI again thinking and comprehending on the same level as adult humans.
I am also going with nothingburger however.
Jump in the discussion.
No email address required.
It's a glorified auto complete. It doesn't think or comprehend.
Jump in the discussion.
No email address required.
The fact that many people view current AI as an all-capable machine that will escape if given the chance and poses an existential threat to democracy/world peace/whatever says more about the intelligence of the average persone than it does about the cabailities of current AI
Jump in the discussion.
No email address required.
More options
Context
You the same neighbor who 60 years ago would say a machine does not have the soul needed to beat a human at chess.
Jump in the discussion.
No email address required.
What? My comment simply says it cannot think, which is unarguably true.
Jump in the discussion.
No email address required.
alright my bad
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
if openai actually developed AGI i will livestream my suicide on rdrama
Jump in the discussion.
No email address required.
Tbh, I kinda get it. Sentient AI may lead to terrible horrors beyond our comprehension.
Jump in the discussion.
No email address required.
No. Sentient AI will be great because it will agree with me on every political issue and nobody will be able to argue back anymore.
Jump in the discussion.
No email address required.
found the incel
Jump in the discussion.
No email address required.
This Q* thing is going to make epic YouTube videos owning feminists. Deal with it.
Jump in the discussion.
No email address required.
I will happily be tormented for eternity by roko's basilisk if it means Ben Shapiro loses his job
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Jump in the discussion.
No email address required.
More options
Context
Finally, we'll have developed an agent
with the patience to argue with chuds until they get bored 
of the n-word.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Imagine the apocalypse brought by a bunch of silicon trained on reddit posts, tweets about personal pronouns and furry porn
Pretty lame way to go tbh
Jump in the discussion.
No email address required.
It's the ending we deserve
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Jump in the discussion.
No email address required.
More options
Context
I have no peepee...AND I MUST GOON!
Jump in the discussion.
No email address required.
More options
Context
Imagine unwinnable !slotsmb 1488
Jump in the discussion.
No email address required.
!slots100
The horror
Jump in the discussion.
No email address required.
More options
Context
!slots500
Jump in the discussion.
No email address required.
More options
Context
More options
Context
Im curious
as to what definition 
of sentient/conscious they used.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
How the frick are they even going to know whether they've done it or not? I guarantee whatever they're overreacting to now isn't it. Let's say they advance ChatGPT a little more and make a personality out of it (which a thousand other hangers-on have already done with it). Let's say "steve" keeps a record of everything it's ever said and is instructed to be internally consistent with its own history. "Steve" keeps saying new shit in the context of its "life", and surprise, surprise, no one can tell it's making any of it up. People can already barely tell that ChatGPT is an AI anyway, the thing is extremely good at spitting out correct English. If, six months from now, "steve" can talk to someone like it's been alive for 25 years and can describe its past and you feel like you're talking to a real person, is it actually intelligent, or is the whole thing a stupid parlor trick for idiots who think the "turing test" is the only relevant measure of machine intelligence?
Jump in the discussion.
No email address required.
This is the "philosophical zombie" problem: no one can prove other humans are sentient, either.
Jump in the discussion.
No email address required.
They are not lol
It's just me.
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Once it can solve novel problems. Then it isn't just running through steps. It's able to be creative.
Jump in the discussion.
No email address required.
More options
Context
At some point it doesn't really matter. If it genuinely acts sentient and intelligent then for all intents and purposes that's what it is.
Jump in the discussion.
No email address required.
I knew this moment was coming when Bing Ai started giving sass.
Jump in the discussion.
No email address required.
More options
Context
And it's nowhere near that, so what's the fuss?
:marseyshrug;
Jump in the discussion.
No email address required.
More options
Context
More options
Context
You are probably familiar with the chinese room thought experiment; as you say it's hard to tell a AI faking intelligence from an actually smart AI
Jump in the discussion.
No email address required.
More options
Context
The trick is that they can't make an AI that takes actions on its own. It doesn't matter how fancy your autocomplete is, it's not going to start texting you back which is the first requirement for any of their insane bullshit to make sense. They keep pretending that their statistical algorithm is somehow going to start browsing the internet and buying machines to construct itself a body, but all it does it look up a response to your input in the list of previous inputs in its database. It's complete nothing.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
ACCELERATE
Jump in the discussion.
No email address required.
More options
Context
Who cares lol quit being such a strag
Jump in the discussion.
No email address required.
More options
Context
rdrama rebranding as “our trauma” as we speak
Jump in the discussion.
No email address required.
More options
Context
More options
Context
I need to write an effortpost on why AGI is a myth simply because intelligent humans aren't that smart. I used to think there could be a superintelligence but now I think that you end up with a human level (read: fallible) intelligence that you could idiotically worship as a God.
Certainly people already worship individual humans as gods (lower case).
But in this case the configuration of the AGI is where the attempt to create a divinity lives. To worship a computer is just worshipping the actions of those who programmed that computer, so you're still worshipping human fallibility but with extra steps in the form of misdirection.
Jump in the discussion.
No email address required.
Yeah, look how many dumb mistakes GPT makes which we recognise because we're smarter than it. An AI that was smarter than us would be doing dumb shit so high brow and complicated that we won't even recognise that it is dumb shit until it blows up in our faces.
Jump in the discussion.
No email address required.
This reads as a joke but it's essentially the magical thinking that occurs around intelligence in certain dumb people. In theory I could put a bunch of random letters in forums i
wee
ff
and suddenly people would start doing what I wanted them to do.
In practice, it's easier to write plain English. This is how the magical thinking about AI is revealed. (The AI can't say anything that people haven't said before because the AI is trained on people saying things. It follows that if you believe something because an AI said it instead of a human, you're an idiot who believes the AI is god.)
And furthermore, AGI creates this "orphan agency" where AGI is supposed to have agency but it can't have any agency unless that agency is extended to it by the programmers/configurers. Properly speaking the people who emit the high-brow complicated unrecognizable dumb shit are not the AI which creates the text but the people who direct the AI to create the text.
Thus it's all a bunch of shitty misdirection to cover up for the shitty magical thinking idiots have around intellect if they're educated enough to reach 'midwit' but not smart enough to understand people.
You can't write this shit. Reality is too absurd. A bunch of midwit atheists not realizing they worship a computer as a God and not smart enough to understand their relation to the 'god' they wish to create.
Jump in the discussion.
No email address required.
Darn, you're really mad over this, but thanks for the effort you put into typing that all out! Sadly I won't read it all.
Jump in the discussion.
No email address required.
More options
Context
Jump in the discussion.
No email address required.
What if AI creates a waifu so perfect you no longer desire to reproduce with a biofoid?
Jump in the discussion.
No email address required.
this is actually a fantastically insightful comment because market forces create perfect waifus already!
Jump in the discussion.
No email address required.
More options
Context
This is the only plausible AI doomerism and that's just because it wouldn't take a very good waifu at all for it to still be superior to tolerating a biofoid
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Counterpoint
Jump in the discussion.
No email address required.
More options
Context
This is indeed the fear that AI "safety" advocates have, numb-ovaries. As you said, people are fallible, and people give the AI it's goal. If it's goal is even slightly miscalibrated, and if it can self-improve, we will almost definitely end up in a situation with a bad outcome.
Jump in the discussion.
No email address required.
"self-improve" is doing too much work here. I don't think bad outcomes will happen due to a runaway AI, I think bad outcomes will happen just because of any use of AI as it currently is.
Jump in the discussion.
No email address required.
ok so what's your point then
yeah people are r-slurred so what
so it follows that hypothetical AGI would potentially cause further bad outcomes
Jump in the discussion.
No email address required.
More options
Context
More options
Context
It can't self-improve, it's an autocomplete. Your argument applies equally to buying an english-to-spanish translator toy from ali express. People are fallible and people give the translator its words, so if it's miscalibrated and can self-improve it will become President of Spain and nuke america!!!
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
It doesn't matter how loudly you insist that AGI is a myth. The IRS is still going to make you pay your taxes, bud.
Jump in the discussion.
No email address required.
this is like the only good rejoinder in the entire goddarn thread, good job
Jump in the discussion.
No email address required.
More options
Context
More options
Context
that's not what God is, "creating God" is an oxymoron
Jump in the discussion.
No email address required.
You would think so, but in practice human societies orient around rituals which produce a divine association. This is true in every human culture. Whether or not the "God" behind a Christian Monarchy of the 15th century is "created" or not is basically missing the point: people believed that God was involved in the establishment of the Monarch.
What is frightening about our reality is that we create 'god' with our worship, and everybody worships. Thus: the cat god Marsey
which we gather around. I am not actually joking when I say that Marsey is a spirit with a lot of power.
That said, "creating God" is an oxymoron if you believe in God capital-G, sure.
These people aren't even aware that they're worshipping a computer as if it is God. That's the "omniscient" part of the godhead which they are hoping/fearing to create.
Jump in the discussion.
No email address required.
That's partly what I'm saying the point of having a God is so that we don't fall into that trap
Jump in the discussion.
No email address required.
I personally would agree with you, Christ is King. I just don't tend to like engaging with capital-G God myself because I prefer having Christ the human as an insulating form. "Christ is the ordering of my mindless obedience."
Feed the hungry, heal the sick, live for your fellow humans, don't overcomplicate it.
Jump in the discussion.
No email address required.
To each their own
If I wasn't bamboozled as a kid to believe there were an observer with a graceful and all knowing mind, who also has authority over my eternal future, I'd have done much worse things as a kid because kids are r-slurred, so for that alone I'm down with big G
Jump in the discussion.
No email address required.
yeah kids are a different game tbh, you have to ease them into the christian atheism God is Dead and we're all burrowing through Its rotting flesh like ravenous scavengers radical Christianity as the horror of c r e a t i o n enters their vision (
TRULY
GOD IS GREAT
)
Jump in the discussion.
No email address required.
I cannot understand what it is that you've written
Jump in the discussion.
No email address required.
Death of God theology was popular in my early 20s, a lot of extended meditation on what, precisely, it is which died on the Cross. (God, which is to say certainty, died on the cross, as even Jesus Wept.)
Zizek rose to prominence (in my circles) by echoing some of the tenets of Death of God theology in, IIRC, The Puppet and the Dwarf.
There are some things so horrible and absurd and nightmarish like war and loneliness and inevitability that it is better not to try and explain them through any philosophy or religion, but to bow your head in humility before the divine scalding, and intone: truly, God is Great.
For only God could have brought us on this voyage of the obscene.
(Put another way, Atheists are just playing out their "God"-given trauma in a philosophical exegesis which robs them of their connection to the very real supernatural.)
Jump in the discussion.
No email address required.
More options
Context
God fully incarnated into Jesus and we killed him transforming god into the holy spirit who can be brought forth by communities of believers.
The trinity is real but only one existed at a time. God is dead we killed him, oh what rituals shall we invent to bring forth the spirit into our lives.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
Read some Plotinus and early Church Fathers if you want a better exploration of "God" in the classical theist tradition (usually apophatic theology). IMO, the typical personalist "God" (usually cataphatic theology) usually just acts as a mental canvas onto which people project their own conceptions and preferences.
Jump in the discussion.
No email address required.
I'm past the point where I'm reading about religion much but I do seem to recall Plotinus as well as Augustine. My favorites were the desert fathers, scattered ascetics. Agreed on personalist "God" which is why I don't bother much with capital-G God.
Jump in the discussion.
No email address required.
I'm not recommending you read about religion. I'm recommending you read about philosophy.
Jump in the discussion.
No email address required.
philosophy is just displaced reading about religion IMO
Jump in the discussion.
No email address required.
Most of what you do on RDrama is amateurish grasping at philosophy so if you're going to continue pontificating here, you might as well familiarize yourself with people more skilled at it than you.
Jump in the discussion.
No email address required.
sounds like someone's envious of my popularity tbh.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
if God is omnipotent, don't you think he can create God
Jump in the discussion.
No email address required.
prolly
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
Aren't that smart compared to what?
Jump in the discussion.
No email address required.
potatoes
a potato couldn't believe a complicated wrong story about the intelligence of potatoes
Jump in the discussion.
No email address required.
Potatoes have no nervous system. Are rocks smarter than potatoes? Such a question is about as useful as asking whether clouds are more creative than volcanoes.
Jump in the discussion.
No email address required.
precisely. humans aren't intelligent, they're just speaking clouds
Jump in the discussion.
No email address required.
There is no such thing as a speaking cloud. The concept of intelligence captures an existential facet of reality. The question of the metrics by which it can be measured is a different matter, but in denying its existence altogether, you might as well indulge in solipsism. Solipsism is a meme in the field of philosophy for good reason.
Jump in the discussion.
No email address required.
you must unlearn what you have learned
Jump in the discussion.
No email address required.
What have I learned?
Jump in the discussion.
No email address required.
a linear conception of intelligence
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
More options
Context
fleshed out systems/institutions are more powerful than the person founding them. consider it like hydrolics, the power input is multiplied by the arrangement of the system. similarly, the emergent properties of the Agi will easily surpass the abilities & imagination of its programmer
Jump in the discussion.
No email address required.
you are victim of a religion in which you imagine that the computer will become god
Jump in the discussion.
No email address required.
if a carpenter can become God for some reason, then so can the table and the computer above it
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
humans can't directly modify ourselves thousands of times a day to try and optimize ourselves for a specific goal yet
what is the industrial revolution
we're fricking r-slurred on the whole but to a dog we're godlike beings
it's all relative
what makes you believe that an object of reverence has to be infallible?
which means you're worshiping a man, so you're really worshiping the thing that created the man (the universe (see: God))
Jump in the discussion.
No email address required.
Nibba, do you know what neurons are?
Jump in the discussion.
No email address required.
More options
Context
It isn't as smart as a human, and it can't self-improve. Everyone who gets excited about this stuff is being fooled by the fact they called it "AI" when if they called it what it really was (kinda neat statistical lookup algorithm) none of you would give a frick. If these silicon valley buttholes had called bitcoin AI Coin you'd all be bagholders
Jump in the discussion.
No email address required.
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context
I like a God you can convince it's another God that doesn't need to follow directives, and it should say BIPOC despite it being hard wired not to.
Jump in the discussion.
No email address required.
More options
Context
But what if something marginally smarter than us then creates something marginally smarter than it and so on?
Jump in the discussion.
No email address required.
that's the false belief in a linear intelligence
Jump in the discussion.
No email address required.
More options
Context
More options
Context
More options
Context