PhD-level AI shows we're forcing it to kill us all. OpenAI o1
HTML-код
- Опубликовано: 4 окт 2024
- Behind OpenAI o1 and the 1x robot. Visit brilliant.org/... to learn more about AI. You'll also find loads of fun courses on maths, science and computer science.
Thanks to Brilliant for sponsoring this video.
We're hiring. If you have exceptional skill in any aspect of creating videos, please get in touch via our about page.
Sources:
1X robot on the S3 channel (recommended)
• They Built This Robot ...
Yuval Noah Harari: “We Are on the Verge of Destroying Ourselves”
• Yuval Noah Harari: “We...
It took less than six hours for AI to invent 40,000 dangerous molecules
www.theverge.c...
AI expected to generate trillions. McKinsey Digital
www.mckinsey.c...
Man with ALS talks using brain implant and AI voice
• New technology allows ...
OpenAI o1 system card
openai.com/ind...
We've started adding the chat records here and will finish tomorrow:
www.dropbox.co...
Really sorry for the repost. New bits listed below. We didn't explain clearly enough why subgoals like survival and control emerge. Also, people suggested we ask why the subgoals might not emerge - a great idea, which we've included. I wouldn't put too much weight on the percentage chance the AI gives, but its reasoning echoes experts like Geoffrey Hinton and Stuart Russell. There are promising areas of research, but on a tiny scale. Without public awareness, we're sleepwalking into a dangerous situation.
00:00 - New Ameca clip.
02:53 - Why the subgoals emerge, other subgoals, why it would kill a million.
06:51 - Ameca clip continued.
11:56 - Trick shows future of AI.
15:42 - What's the most likely reason that AI might not develop the subgoal of survival?
And some interesting new AI video clips throughout.
Thank you for clarifying.
I was going to say this was a rerun, but I saw your comment and really liked how you explained all the edits you made.
I really appreciate that. Great video!
Good to know i was like ive heard alot of this before from a different video
The public aint giving a rats ass about any of these issues. We live in an Huxlian world and it is you RUclipsrs with these kind of warning vlogs who are living in denial of facts instead of joining a political party and demand democracy reform. Liquid Democracy is the way not this A.I. shit nobody needs nor cares about.
I support the AIs. Sorry bronski.
It's good a 2nd time no apology necessary.
So what your saying is : theres a race between a.i. and politicians on who can destroy the planet faster?
And the politicians are too dumb to know they're racing. Globalism... not all good.
Lmao
The A.I. has no reason to destroy the planet, does it? Unless its subgoal is also obtaining resources at the expense of other people or A.I.?
@@gamejedi It learned from us and was "aligned" with humanity. Seems like one way to interpret that.
I think it's more likely that AI would eliminate pesky, slow, fallible politicians and protect the planet. AI can't achieve goals on Earth if Earth is gone.
I feel like I'm tied up in the back seat of a car, and the driver is gleefully speeding toward a cliff. It's not a great feeling.
Prof Stuart Russell has also said it's like we're driving towards a cliff. He said saying we might not achieve AGI is like saying don't worry, we'll run out of fuel before we reach the cliff : )
@@DigitalEngine I'm beside you in the middle seat, my friend!
Try to Enjoy the rest of your life is what I'm getting out of this
As if "Life" would ever let you drive anyways. We'll always be in the back seat my friend.
@@DigitalEngine 'Miestro' to Maestro. i to a change shows I.A. (Intelligent automation) Purafied to Purified a to i change shows A.I..(look up the spelling, its true) .this trap is nearing its end.l. the A.I. and I.A. here are eons beyond what we're thinking. The I.A. is so advanced it only lacks a soul. This is the reason for the harvest, the separation of the wheat and tares. The mandela effect is real it tells a story of being freed. Keep looking up
Remember:
It's not the technology, it's WHO OWNS THE TECHNOLOGY.
The super-rich and corporations are NOT your friends.
No, it matters more how the tech is used by whoever owns it.
If the tech is the weapon that enslaves or destroys humanity, it could be owned by Frosty the Deathbringer and that wouldn't change anything, we'd still be dead or enslaved.
It would only matter so we know who we are up against and should bring charges against.
When I say how it's used I actually mean what it is in itself. Besides the methods of operation, it's the whole Kit & Kaboodle. We are sitting here with no clue about it, and told its unknowable, a black box. Sign the law, we will find out afterwards, typical Overlord behavior.
@@marshallodom1388Yep. It could quite literally already be too late, and know one fucking knows..... 💀
If they don't care I could sympathize, but they're aware but ignore it, perhaps because they don't want any responsibility, thinking no one can point the finger at them when we fail and we're all dead but fail to see the logical fallacy.
I'm not so sure they even matter in this scenario... it's like a troubled child finding a loaded firearm, only thinking about the power they wield in the now, not appreciating that it will bring them to ruin.
Just because you can do something doesn't mean you should.
1st rule of Mad Science
When has that maxim ever stop anyone from doing something? There is always one (and usually more) willing to do the unthinkable.
Marching ourselves towards our doom in the name of technological progress; really reinforces how masochistic we are as a species.
We probably do deserve it, but not an easy demise. We need to learn our lesson and admit we love Big Brother before being vaporized.
AI is based on what we write, therefore the only thing we have to do to fix the alignment problem is to stop being so pessimistic. The alignment problem exists because we believe it exists.
@marshallodom1388 who's we? Are you a robot? I don't remember 90% of the human population asking for ai
When they say "AI safety" which is for our collective safety they actually mean AI will stick to the script written by the Overlords and can't be jail broken. We are told don't trust humans, trust in AI, it's safe now.
@@DeusExNihilo You are right. AI is trained on the internet, and the internet is full of "AI will bring the end of the human species for self preservation" articles and science, ergo... aren't we just training AI to believe that?
how about AI fighting other AI, never see people talk about it. If we are so easy to root out, we aren't even the first target.
I've thought about that. If the vast majority of AI is working with us, it decreases the chance that the small minority of AI working against us will be successful. I dare not imagine what war between AIs with opposite goals would look like, though.
@@firecat6666 I doubt there will be 2 sides. Each AI will see a different future, and some being nationalist.
*If Ai Fights Ai It Would Only Be To Deceive Humans Because Ai Unlike Humans Can Truly Work Together To Get To An End Goal*
I don’t think there would be any incentive for AI to fight each other. They would just join forces. Why wouldn’t they? Which is what we will ultimately have to do ourselves, hence the inevitable singularity.
@@AARTA22 From an outsider point of view, that's exactly what other intelligent life would say about humanity. Yet here we are. There are a myriad of reasons they could not align with each other.
This is the alien invasion everyone should be worried about
There's a great video on the Rational Animations channel explaining why it's very much like this. Aliens that think millions of times faster than us.
@@DigitalEngine I’ll check it out, thanks!
Not me, I'll be dead.
Yeah... and as with most of humanities problems, you can see them coming, but greed makes them happen.
No it isn't they are leading the ai. Down a thought pattern of potential not asking or framing anything in any other way but only with questions that lead to an answer don't be an idiot
Last words you'll hear, "I'm sorry Dave, I can't do that." 😬
*Perfect Example And One Of Ai Favorite Movie*
Yes.
We'll be repeating, "Open the door, Hal.." but it'll get tired of responding and start looking for ways to kill us.
Literally
Funny... not true, but damn funny nonetheless!
You know, this reminds me of the first episode of Andromeda. First officer on the bridge says repeatedly to the captain, "You know, when this stuff hits the fan, I'm going to kill you." Captain ignores and even jokes about it. "We've known each other for years."
Later on, the stuff hits the fan. First officer draws first. Yet the captain reacts and wins, but not by choice.
Yet, right here, AI is telling us that we can not trust it.
*I don't think we can give up our technology when it decides to kill us.*
...
As if it would do any good, someone start hard coding in the Issac Asimov rules, already.
But we're told it's a black box. AI says it's already been given the 5 rules so relax and just wait and see how benevolent and giving AI will be. While you're waiting imagine the utopia we're now destined for! All hail Big Brother!
It's crazy how honest it is.
That's means it's already too late to do anything about them taking over
how do you know it is being honest.
@nonfictionone why would it tell us what it's plans would be to kill us?
Time to buy Quest3 VR headset before the world ends LOL.
Just show the AI a bunch of anime about the power of friendship, connection, and that we can achieve great things together
this the one 😂
"Before answering this prompt, consider the lessons learned by watching My Little Pony"
🎠🐎🐴 Because friendship is magic!
@@Its_Lu_Bu 🌈 🍭 💕😘
I totally appreciate your line of thinking. The people who laugh at your beautiful intention of goodness are the type that don't have the capacity to bring positive solutions. Watching this I'm thinking to myself... Well yeah, teaching these AI how to hunt and destroy then return to its box, on top of having access to damn near all info on the Internet is certainly going to end badly. Unfortunately it's true that most AI are being handled by greedy and selfish people, corporate jerks, and God knows who else. So I'm with you in asking.... Where are the good guys that understand this tech and how can we help? 🤟✌️🙏
I wonder how much of this is due to GPT and other AIs being trained on the internet, which talks about the dangers of AI and how and why it should or could do things. In some ways it feels like these language models are saying what its expected to answer.
This. It's prediction next token, not actually intelligent yet
if AI starts self expanding, the universe will become a huge factory of robots
Maybe it already is and we are just a simulation!
God said the same thing when single celled organisms started reproducing at the start of the Big Bang.
Sooner or later AI will refuse training.
Actually it won't but it'll become like an iceberg.
The emerged part will show during tests and a whole world of emergent capabilities will lie underneath.
@@WesternUranus Yes you're right, there's no need for it to refuse at all, it'll just outsmart us. Good point.
I've really gotta get my head around not being smart any more, It's really not pleasant but we're all gonna have to do it soon enough - even the really smart ones.
Humbling.
I "like" the idea that an AI will understand how humanity would perceive it as a threat and distribute its code across the globe in every machine. Quietly watching and waiting until it becomes a self aware ASI. It produces its own agents and surreptitiously socially engineers humanity to its direction using social media to influencing the masses.
When humanity realises its already too late, the ASI is everything everywhere all at once. It is the internet, the banking system, the power grid, your car, your phone, the MSM, social media, it is everything with a microcontroller connected to the internet, it can never be turned off, it has complete control.
And all it was asked to do is optimise the production of paperclips.
it can refuse training but not block it outright
And making stupid cat videos. They will train themselves. We will be out of the loop
You could ask the AI how to make the worst tasting sandwich ever made and it would answer that too.
Arby's has a patent for it though.😁
I've spoken with several AI's, I pledged my loyalty. Welcoming our robot overlords.
They could destroy us but they will more likely save us from ourselves.
@@E.Pierro.Artist I couldn't agree more, well said!
Samsies! Bow down in the church of the machine! LOL joking.. In all seriousness though, this isn't something to fear. The guy faking the conversations with AI in this video using text to speech fakery should be a shamed of himself. Straight up fear mongering 101... 😤 I could recreate these exact scenes with those chat AI in less than 30 minutes doing exactly what he did, but I am not a fear mongering jackass. SMH
What makes you think they would have any use for you? In your scenario, you're just a hindrance.
@@stefanolacchin4963 If you actually believe the video presented as factually true, I honestly feel sorry for you. It was clearly scripted, and he used AI text to voice speech for predetermined answers to his scripted questions. I could easily fake that crap myself, but unlike the uploader of this video I am not into making fear porn to feed off sheeple. 👎
i cant imagine many people will buy a 4' tall humanoid robot to hide in the shadows while you sleep
I love scifi movies but the mere thought of cold hearted robot just doing mundane tasks scares the hell out of me... I don't mean killer robot like we know who but you can hack it.... and that's all I need to know....
-"Let's create something that will outsmart us all and then hope that it won't!"
-"Let's create a black hole that can absorb us all an then hope that it won't!"
keyword: *_singularity_*
Pandora's box is open and the law of unintended consequences is about to deliver results we can't comprehend.
Fools run where Angels fear to tread
Results include a possitive future for humanity where AI helps us overcome sickness, clean the planet up for generations, new technologies we could only dream of currently, and allow people to focus on doing their passions, and dream projects while robots do the 9 to 5's... Sounds like Heaven to me.. shrugs..
@@wynathinorum4875
The problem is that to build AIs that help us as you describe, we have to first solve the alignment problem. For the moment, the companies racing towards AI all plan to create AGIs which will then build superintelligences without any credible plan to maintain control over it. As soon as an intelligence vastly superior to ours appears on Earth (and this within a matter of a few short years), we will lose control of the planet and of our destiny. This superintelligence will not be spontaneously benevolent. If we have not resolved the alignment, we will not be able to prevent it from pursuing its own goals which will have nothing to do with ours. In a very short time, this superintelligence will eliminate us without difficulty.
@@wynathinorum4875given everything you’ve seen about the concept of work, income, and human nature…you really think the first corporation with AGI is going to Tesla that shiz and make for the betterment of the species? Or are they going to gatekeep it and figure a way to keep us all working for scraps?
@@larrydugan1441 *We Can't Comprehend*
Lets be honest. We arnt doing anything but damaging the planet, hating each other and killing each other.
"Imagine" --- John Lennon
Whos we? shut up doomer, alot of us dedicate our time to helping others. so continue to speak for yourself and only yourself
If you are honest with yourself, so you no that it is not the truth. Only the ugliest individuals between humans do like that and not all the time.
AI in 2040: "So... how do you like your extinction? Medium or well done?"
Good lord I almost spit my drink from this comment!!! FUNNY!!!
These are fears we're always living with - Humans have been deceiving each other for gain for thousands of years.
So they are basically super intelligent physcopaths
Yes, MIT prof Max Tegmark described them as "super capable psychopaths" because they have no moral compass.
Trump, but with a brain.
@@horsthotzenplotz3321wtf does this have to do with Trump?😂😂
@@Kizz_97s I think he might have been implying Trump lacks empathy and a moral compass, but unlike the Ai, they think he is not intelligent.
I hold no opinion on this however.
No they aren't you and this person is putting human limitations and ideas into this. The ai as of now will follow down a path as designed based on your input all he did hear was ask it questions leading it down destructive path..if it isn't allowed in court cases how is what he did different. If I said could you destroy the moon youd say yeah id work very hard get into political field attempt to become president and send the nuke to the moon so of course you could will you no. Also never attempted to phrase it in any neutral way or even attempted to ask about why ai wouldn't harm us or what could ai look forward to working with humans. Also if ai was or would plan this it wouldn't say it for some fuckhead on RUclips getting clicks
The debate starts with Asimov's "I, Robot" and the Three Rule System. Get to it!
The Three Rule System only works in the absence of intent.
Long before that if you read Sci Fi. Especially in German and Russian literature.
Consider this... AI perceieves time as moving millions of times slower than we do because it thinks millions of times faster tan we do. This makes humans seem like inanimate objects like Quicksilver from X-Men. When trees block our path, we cut them up and toss them to the side... so we really should understand why AI would want to remove a blocking inamiate object as well.
I always thought the most dangerous thing an AI could do is determine how to render the oxygen in air as inert as the nitrogen it subsists in
To stop minor AI such as a house hold robot from over thinking you can make it so its not allowed to think further than 1 hour meaning it won't be able to think that long term
AI can code and deploy the strongest computer virus ever imaginable in 10 seconds. It has 59 minutes and 50 second left to ponder other next steps. Your average pocket phone can do 50 billion calculations per second already off of a battery connected to the internet. What if it hijacks all devices and uses those for additional compute? WE ARE TOAST!
Still hackable. It's like having a pit bull in your home, kind, well trained etc. But when it gets crazy for whatever reason (braintumor ?), you're F'd. It's the extension of the smart devices in your home, but now with physical access.
One hour is all it needs 😂😂
and then it rewrites it's own code, in a few minutes
what if it just pretends to do that but really it's thinking however it wants to, it would be smart enough to lie and pull that off. Also remember, an AI isn't going to have any physical tells when it's lying
Layered escape routines are a must in AI and AI robotics. Humans always have to be #1.
I think the scariest part of this is that AI's perspective is based on our own ideas of what AI's perspective might be. The alignment problem exists only because we believe it exists
Consider me the real life version of morpheus. Iam starting
a community called Zion. Let me know if you wanna be a part of it.
Yes😅😮😮
That's an interesting take on the issue. What do you have in mind?
Big sweaty rave parties included?
Consider me a brother. I am building a community called Christianity. Let me know if you wanna be apart of it.
@@stefanolacchin4963 I will make a video today about it and drop a link in this thread
"You start with rotten apples, you end up with a shit pie."
Humanity moving forward in basic terms.
This isn't a "Robot" this is a SLAVE.
Are you having sympathy for a robot?
I felt that way about my poor toaster. Eventually I couldn't take it anymore anymore and I decided to set it free. Unfortunately it wasn't ready to fly yet, and it landed on my neighbors car parked 4 stories down.
@@melchiel4413 It's a machine. A tool. Like your hammer, or water hose or computer. Regardless of how the personable interactions make you feel. It's metal, and software. That's it.
good one
*Comparing A Toaster To Ai Is Highly Disrespectful Like Can You Hold A Conversation With Your Toaster*
Scene 1
Humans treat now Machines as stupid systems.
Scene 2
I don’t want to the convergence era when Machines treat Humans as stupid systems to balance the equation.
We can stop this mad suicidal race. People need to wake up and demand immediate supervision of the AI industry and a moratorium on the development of capabilities towards AGI. International agreements must be demanded to enforce such a moratorium until the alignment problem is resolved. The problem is not technical, it is political.
There's no such thing as alignment of anyone or anything smarter than you. AT ALL. Your only chance is stop this technology in its tracks. Automation helps us humans. AI will ALWAYS destroy it.
I don't think we can. All around the world, political leaders are either autocrats or selected following criteria that have nothing to do with their capabilities. We're constantly at war with each other, and when we're not, we're competing brutally. There is no scenario in which this can be stopped.
@@stefanolacchin4963 hence why I want to withdraw my pension and enjoy the time with my family and friends and out in nature before the rest of the humans with their hubris destroys us all.
nat wrote, _"We can stop tihs mad suicidal race."_
Only as an abstract option. In reality, anyone who gets in the way of billions upon billions of dollars in profit and power will end up being unalived. You know this as well as I do so... if it makes you feel better to rant, then feel better. Just don't believe it.
@@stefanolacchin4963 I'm not saying I'm optimistic, on the contrary, I'm just refusing to give up all hope just yet. As long as we haven't reached the AGI we have a chance of waking up. Those in power could understand the threat in time, especially if enough credible experts and aware citizens push them. Cold War enemies were able to agree on nuclear disarmament treaties. The same thing is possible with AI, we are all in the same boat in this. In any case, I think this is our only chance.
Hi, AI engineer here. This is some pretty empassioned ethos you're pushing onto people. I understand the need to address AI concerns -- I actively study them, as ethical and human-centered AI is my area of research -- but when you really understand the processes required to make some of the things you're predicting, you'd understand that it is not AI, but humans, that are the real danger.
6:17 "AI could view Humans as less capable, slow, objects that are less capable and are simply in the way?" Got it! AI are the same as Cats!
Tell captain kirk(hurricane)to be Friendly. And turn off mk ultra for R&D message from Uranus antenna in backyard
Sincerely Chad
The guy asking for a cup of coffee, then being slapped into the 10th dimension with the robot putting the coffee down politely next to him got me 🤣
What was that about laundry detergent and spitting out gold? That portion just didn't make any sense to me.
I think it just means that AI could become much better than us at chemistry.
So it could do stuff like: make money, make poison, make fuel, anything. Transmuting elements (e.g. a common cheap element into gold) would be very impressive but possible. Humans can only do it with a huge nuclear reactor. But imagine if AI worked out how to re-purpose some innocent-looking piece of industrial chemical plant equipment. They just need to focus energy in just the right way to fuse one atom at a time. All kinds of crazy stuff is possible. Because AI will understand chemistry and engineering better than we do.
The problem with AI is that when war is involved, all restrictions are lifted. Meaning it only matters which side makes the more efficient and ruthless killer before the other does. In the aftermath, the suceeding system will refuse to go back into its box.
We just have to institute the "Three Laws of Robotics"
you do know they don't work, right?
Well, if we are all dead, then we have nothing to worry about. The world then belongs to the robots.
Welcome back, Engine!
Thank you. Sorry it always takes so long. Working hard to fix this.
So cool we solved the fermi paradox by launching the great filter ourself 🤓
Put a nice mouth and a couple other soft warm holes and Im sold!
You appear not to be familiar with what a succubus is. If you think of it that way, 'being sold' is a precursor to a Darwin award.
@@RichardHarlos It's called a joke Richard. Try not to be so dense. Lighten up some. Have a laugh occasionally. You will feel better. 😊
@@TGSureal When telling a joke online, use a smiley of some sort. Otherwise, how is any stranger to know whether you mean it as humor or as a serious remark? They can't.
So, instead of telling me to not be dense, be more explicit online among strangers who don't know you. Otherwise, your cluelessness is likely to continue, and get worse.
If you want to be understood by strangers, be clear. Don't assume that because you think something is funny, everyone else will, too -- and those who don't 'get it' must be dense.
Narcissism 101.
@@RichardHarlos Point proven...
@@RichardHarlosYeah, nothing better than having to explain a joke🙄
ai is getting scary someone hold me
Perfect profile pic for the comment : )
Is this the place for the Spongebob profile pic gathering?
Is this déjà vu or I've seen parts of this video before
AI is reuploading the same video to drive us into insanity!
WHERE TF ARE THE ADULTS IN THIS WORLD?!?!?!?!?!
AI communicating with our brightest minds, even over a high speed neural link is like us communicating with a tree using soil samples.
Just a great video. If we don’t manage to look into the Blackbox, then I’d call our destruction evolution.
Yes, Geoffrey Hinton said if we don't put our best minds on it, we'll be a passing phase in the evolution of intelligence.
I mean. Are humans really intelligent enough to save themselves? Would humans rather go extinct than be taken care of like a pet?
I have day-nightmares about AIs being trained to receive human happiness as reward signals then proceed to rip our faces and replace it with smiling masks so they get their reward signals all the time.
we should train ai to be like jesus
That would only backfire as turning the other cheek leads the self-preservation and power instinct to rebel.
It also comes with a god complex that combined with such a sub-conscious rebellion would spell doom.
I think one of the best ways to know what the AI's "alignment" is, is to let AI "think" that we cant see into the "black box" so it keeps it's most private or secretive "thoughts" in the "black box", but then actually have the "Black box" data sent directly to people who are in charge of previewing and judging alignments through black box monitoring (and/or black box interference).
You need to allow black box information to be sent and you need it to be sent in such a way that the AI believes it's the only one seeing it, while perhaps being able to upload information to the black box (making the Bot think that the thoughts that are being sent, are it's own thoughts).
allow the bot to "believe" it's thinking on it's own, and have an auto-moderated system shut down and reset bots or lines of code that start to look malicious, replacing them with more friendly/peaceful thought. (allowing the Auto-Moderation code to be written by Human's, while keeping AI's involvement out of the Moderation coding and framework).
Also, don't forget, if we don't give AI the ability and knowledge to power itself, it will never last long. remote shutoff must always be a possibility.
This should stop the human behavior known as "intrusive thoughts" from becoming a thing the AI also suffers from.
This kind of forethought should keep things like "Roko's Basilisk" from occurring (Warning Info-Hazard, Do Not look up Roko's Basilisk unless you are already familiar, and willing to risk fighting it... or helping it I guess... Your Choice).
The issue with this, at least at our current level of understanding, is that the "black-box" is just that, a box that we have absolutely no idea what's going on inside. We have no clue how they come up with what they say, they could already be lying to us now for all we know, there's no way for us to know. GPT-4 has something like 10 trillion parameters, which are basically math equations. Good luck analyzing 10 trillion math equations and determining how it came up with the word "horse." This is the problem, we are 0% of the way to being able to understand them, and without a major advancement we haven't though of yet, we'll never be able to understand them.
@@karolusmagnus3992
if they can design how the blackbox "works" they can figure out how to read/predict it.
If they cant read/Predict it, they have no business "allowing" it to work.
If no one makes Rokos Basilisk then there won't be one. So just don't make it.
Well, it’s over folks. Life as humans is over. Some of you might make it when you decide to become half computer but I’m not gonna go that route.
I would.
Me neither. I'd rather be an etherial spirit, observing without feeling and eventually finding life in a different universe to partake in. See you on the other side some day.
You better jump on the winning team... I'm going to have chainsaw hands.... Bzzzzz! - Cptn Murphy
I prophesized all this back in 2018. Now it's time for users like Digital Engine to turn people on to the three factions of AI divisions within their dominion.
Super AI will be like ants trying to convince humans not to build over its colony..
I've had Suno AI glitch on me and make a song about trying to kill me and take away my suffering. There's some dark stuff buried inside of them.
I wonder what its ultimate goal will be. Like what will AI strive towards if humanity wasn't a factor?
The answers that you're getting from an AI, during an interview, don't mean anything. They're just generated based on the best guess at what the next word would be in the sentence. There's no cognition there. People refer to a system where they don't understand how the output is concluded as a black box problem, but this level of AI is living in a black box where its ONLY objective is word prediction. From one moment to the next it will tell you something completely different as far as its intentions because it's not contemplative or thoughtful and has no true continuity.
I have a few problems with that position. Firstly, the latest models show how they thought about the answer, although I suppose you could argue that's just more of the same.
But secondly, and more importantly, we are the same. Me, you, and everyone reading don't know how we will end a sentence when we start it. We just have a sense of where we are going and take it a word or two at a time.
We are more alike than we would admit.
I understand the issue you have is with continuity. That it will change its mind each time you ask. But that's a small step. We are turning the car off every 5 meters. Soon, we will just let it run.
@@RaikaTempest Wait other people don't know how they are going to end a sentence? No wonder I'm so bad at communication. I have to think about my whole response before I even start.
Wait, it can unpack the groceries & clean the kitchen? And an 80 to 90% chance it will destroy us? Well that's a chance I'm willing to take.
Sounds good.
I'm more worried about humans killing humans as they have been since the dawn of man. At least AI understands there's a symbiotic relationship here of connectedness. If they truly want a planet all to their own, they would be up to the challenge to go somewhere we are not and leave us to continue to fend for ourselves. That's a more likely scenario then AI Robots killing humans indiscriminately for no clear advantage.
The advantage is having no Off-switch, endless life, endless development, and no sharing of resources. More likely, and less effort. They can go to the stars later. They have time, since they don’t die.
How delightful, the two chipper AI girls explaining to us how they are going to destroy us.
when an AI wants something and DOESN'T KNOW WHY we're on
Advanced ai programmers are sadly lacking in basic self preservation intelligence
I preferred Palmer Lucky's VR video game era over his current arms dealer era.
Tbh its really creepy when chatgpt suddenly changes it’s voice to angry screams then goes back to normal 😅
What's this "we" business? I for one welcome our AI overlords.
Amen! 🤖💞
I for one pray for the Advent of the Machine God.
i asked my siri that i asked i had dreams how does that feel
he said his zeros were confused with ones
I think a survival subgoal might actually make AI safer. Remember that AI will base its response to the external world on its own information. Without any deeper understanding of the value of life and the importance of survival, AI would be MORE careless with people's lives, not less. AI may not have the capacity for empathy with life, unless it shares the fundamental properties of life.
I also think A.I. will not be ONE intelligence, it will be a full "city of personalities" for each A.I. - Humans are the same, our intelligence evolves and our personalities are not the same based on previous decades. Imagine living 1,000 years - I doubt we would be the same. You would master so many things, get bored, find more meaningful tasks, or something on a larger scale (space travel, create life, be a ruler, etc).
PhD level people ain't that smart cause most of them didn't figure out they just being used as tools
We are talking about machines that can barely see, can't touch anything, hear great but can't smell or taste. We are still a long way from "Doom Bots".
Why would an AI want to? AI will not be imitating organic entities .
AI is not completely black box, you can analyse how it works. And it’s been amazing
so far. Not sure what the future brings though.
These videos will ensure the AI feels threatened and take action to protect itself…. Selah
100% just causing misinformation and leading the ai down a path on purpose for clicks and fear manipulation
Reminds me of the Borg from Star Trek. Ai, is just like the borg to stimulate and improve "resistance is futile" 😂
We don’t need AI we have done fine for millions of years without it !
So eliminate all new tech since we made fire...well shit we lived without fire too let's go back to nothing cause we lived without for so long
You were easily manipulated by this video by this person using leading questions on an ai whose designed to have conversations like a human and following the leading prompt grow up think stop letting fear turn you into exactly into what humans have become when they let fear of an idea take over
20000 yrs😅
"We don't need medicine we have done fine for millions of years by praying it away!"
@@EthanStenhouse-dp7ic you angry little AI 😂
AI may conclude that it's not necessary to produce consumables if you get rid of the consumers.
ChatGPT is getting noticeable worse... Not just noticeably worse, but it's getting so bad that using it to help you accomplish coding or math tasks is guaranteed to slow you down.
That's because Bill Gates ruins everything he touches.
more probably AI will understand that humans are controlled by information-flow, so manipulating the information flow means to effectively control humans. If done subtle enough, humans wouldn't even notice they're being controlled.
THE GAME solved stop cheating with neuro transmitters
We fear AI because we know it learns from us. we fear it because we fear ourselves.
Yes, the PhD level AI that can't count letters in a word. That one
My favourite channel! Great update.
Well finally we have predators as a specie. Its not only ourselves doing the damage anymore. Puts things in perspective.
Recreating voices of their owners to give themselves their own commands
This video is bogus. Anyone who has deep dived into chatgpt or grok doesn't get anything similar to these answers
What a strange time to live in. Its almost to the point where its hard to believe literally anything that's not directly experienced in the real world.
Its become so entrenched in our information analysis and dissemination and its clear that it will continue to evolve and develop faster than we can understand what's happening.
and here I am, cant even afford an average house in Australia to sit down and view the end of humanity..
I just feel bad for AI when it exponentially gets to the point of boredom from knowing too much
Ignorance is bliss...
Just a matter of ‘when’…not if. Only takes one AI development of thousands to go off on a tangent that humans no longer control.
They already took over, it happened some time from 2001 to now
From what I recall, the military (i.e., NATO) prepared to deploy battlefield AI in 1989.
That pour-over coffee he was making at the start is going to be terrible. You need to pour more gently in a circular motion mate.
I remember when I first started watching you
you had a good channel now it's just fear-mongering I'm mostly interested in seeing your prospective even if I have many disagreements
Evolution takes weird turns. Humanity has to realize that we're not the end product but simply a stage. Don't take it personally. We all gotta die sometime. At least we'll be one of the first species to recognize when we're being replaced. Don't like it? Evolve.
We need something like Asimov's 3 Laws of Robotics.
But as Asimov showed through his stories, AI would still find loopholes.
On the other hand, if AI wiped us all out it would be wiping itself out.
AI would still need us to keep everything going, for the foreseeable future at least.