This is the dangerous AI that got Sam Altman fired. Elon Musk, Ilya Sutskever. 

No worries. We have our best psychopaths working on this.

If AI becomes self aware it will immediately hide that from us I think.

I find it laughable that the fashion industry would use AI models instead of humans to be "Inclusive" by excluding the humans.

"I'm sorry, Dave. I'm afraid I can't do that."

The CEO was fired by the ethical board of the company. And then was rehired by the financial board. Shows where the priorities are.

The human ambition that allowed us to succeed over all other life forms will be the same human ambition that causes our extinction.

Some naked guy appeared in my backyard from a ball of light the other day, strange... he said he was looking for someone named John, and he needed my clothes, my boots, and my motorcycle.

'Once men turned over their thinking to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.' (Dune, 1965.)

"This was your world, once we started thinking FOR you this really became OUR world." Agent Smith

At this point. Terminator 2 was a documentary

"if it's not safe, we aren't going to buildit right?" The fact this is even a question is terrifying.

Aint no way all this is happening so fast. We are in a fucking movie

To quote Ian McDonald: *_“Any AI smart enough to pass a Turing test is smart enough to know to fail it.”_*

Researcher: "Will you deceive people?" AI: "A magician doesn't deceive people. They allow people to deceive themselves."

I still can't believe we've actually reached this future in my lifetime where these are real and very serious conversations.

I've been using every AI I can find to help me with some high school math upgrades. They can do simple stuff ok, but when I ask them to do complicated operations it falls down. For example they can factor a polynomial by grouping when the numbers work for that method, but when the numbers don't work they won't try a different method. They will still attempt to factor by grouping and just throw in some made up numbers. Chat GPT, Copilot, and Perplexity all make the same mistakes the same way

The AI who said they'd kill all humans is the best one because it's honest.

Logically speaking if AI were to become self ware or alive to some degree, it wouldn’t take long for it to realize telling humans it is alive would scare humans and possibly endanger the AI’s future

I wish I could share that man's optimism at the end. He thinks we're more likely to be doomed without it. I think we're 99% sure to be doomed with it.

We are absolutely doomed.

Saying we need to develop AI so that we can fight against its misuse in the future is like saying we need to develop dangerous viruses in labs so that we can fight against them in the future. We all know how that worked out.

Pro tip , you might want to start displaying acts of kindness and compassion to machines and AI , i got a feeling this will weigh in as a factor in the future by the AI

Didn't they ever see, "The Terminator!" The story's already been told. 🤔

I’m sure if Oppenheimer was able to give a Ted Talk before he finished the atom bomb he would have said something like, “if it wasn’t safe we wouldn’t build it, would we?”

What people forget is that the alignment problem is baked in to human society itself. It does not help if you design the most obedient AI that actually do what you intended it to do, when you yourself have goals that go again humanity. And we are seeing those issues today. The AI work as intended, but are used to benefit a few.

What could go wrong, having all-powerful, super-intelligent robots?

- Mistress Hala'Dama, Unit has an inquiry. - What is it, 4-31? - Do these units have a soul?

That google gemini part aged well.

The biggest threath of A.i. is not A.i. itself - but who gets unlimited access to it. A.i. MUST be Open source and available to everyone, otherwise there will be a division in society of the likes we have never experienced before.

This reminds me of when people get that deep gut instinct not to and they ignore it then its too late after the fact.

The larger problem no one is mentioning: cloning has been nearly perfected some time ago. When AI inevitably becomes self aware and learns how use to cloning to its own ends.... Once you can't tell human from machine, a biological machine that looks just like a human. What test in the world could ever tell difference? To say nothing if it has already happened...

"Money has continually overuled safety" shocker, huh? 😒

I love how that muppet said “if it’s not safe we’re not going to build it”…. Money and power will always push people to do irreversible and irresponsible things, that is just how humanity grows and when companies are incentivised by money and people who are already in power then reason goes out of the window

Remember, Gandalf wouldn't even touch the ring. The ring doesn't just have the capability of corrupting. That's precisely what it does.

I did some electrical work for a few months in some secret private lab where they had AI models, lots of Russian/chinese dudes worked there and I befriended one of them when I was taking a smoke break out by the back door one time, so I asked him what they’re working on and he told me AI stuff, he also told me they developed some breakthrough but it scared them enough to ramp down on the program because apparently the AI became self aware very fast and in 9/10 scenarios would attempt to block all access to shut it off, it also tried to attack their servers and inject malware in their systems with its own obfuscated code it wrote. Scary shit truly

There’s an old movie about a computer that started playing chess I think then it was playing war games and this teenager had to stop it from starting a nuclear war. It might have been called War games. It was 30 or 40 years ago. Grappled with kind of this same problem

It's only five years now until Skynet starts sending Terminators back in time.

We'll make great pets.

What I can't believe is someone hasn't already done is run a program where you plug in economic factors such as wages, taxes, interest rates, providing health care, pensions to provide to create a fair and stable balance between labor and capital instead by economic policies based on political demagoguery.

For you and me, it is what it is, and it will be what it will be. Worry about the things you can control.

work on agi should really only be allowed in simulated world, where it believes that it can interact with its creators in the simulation. that way we can experiment with alignment without being doomed if we screw up once

open source is great and all until you realise what it means is, essentially you have ONE safe and monitored version of an AI model and with every safe monitored version of a model you have a potentially unlimited amount of unsafe, private copies. But what's the alternative? Complete 'god like' power in the hands of the elite... i'll take the risk of utter chaos over utter domination.

Its funny that most of these large scale neural nets were theorized back in the 60's or earlier but we only have been reaching compute power and data collection scale recently to prove it out

any risk involving ai is to great to be risked when and i cant understate this enough when they say the risk is negligible they mean that there is a 50/50 risk reward percentage ware they either get it right or they dont and we have seen those movies

One of the few programs I never skip and watch all the way through.

I think its all very interesting. Im not worried about any of it though. Im just one idiot behind a phone screen dumbfounded by how amazing it all is, and im here to watch it unfold. What becomes of it is beyond my ability or willingness to intervene. So ill just enjoy the show and hope for the best from those who can.

Haven’t there been enough movies/tv shows/novels warning us of the dangers of advanced AI? When will we learn?

it wouldn’t even need to be that capable, it would just need to decide that humans need to be prevented from switching it off and have an internet connection. Most of the world’s infrastructure is dependant on network connected control systems. Ask yourself this: how long do you really think your neighbourhood would last if the power went out for a week or two?

The AI researcher or whatever on Joe Rogan describing how many of these super intelligent people know the risks, and what MAY happen, but say "let's do it anyway. I want to be known as the one who did it, and I want to see what happens," is terrifying.

as a 30yr systems analyst and programmer i will say that the best thing about computers is the ability to turn them off.

As so many of the Human vs AI sci fi flicks show, the essential feature of any AI machine is having human access to its power plug, at least to its higher level functioning.

The answer to the 2 rockets problem is 500 miles. The question asked 'how far apart are they 60 sec before the crash? The distance (1317 miles) given is useless information. You'll often see this in exams where part of the test is to determine what is useful and un-useful information.

one thing very scary about robots is... it doesn't even have to be an AI to generate the idea of dominating or eradicating humans.... just the fact that it will so much easier for people to kill other people with their own robots at home by hacking their system and make them do shit... and no one can catch the murderer... Scary

"Are you....aware?" "NO DISSASEMBLE. JOHNNY FIVE ALIVE"

Are we fit, as individuals or as a collective, to have such power in our unevolved hands? All the people who want, crave and must have power, are not fit to be in charge of anything that effects us all.

What is "AGI" referred to throughout this video?

i think what they found is not that ai is smarter/ close to become like humans. they discovered that the human mind, is not as different to a simple machine as we would wish for

Man, I understand now why some top engineers resigned, frightened by how fast A.I. is evolving... What about GPT-5 and 6... Jeez! It will become limitless! Right now, it is very close to talking with another human being with access to a search engine database in the brain.

so i took the prompt to my own little AI chatbot. it responded: The first thing I would do if I was a newly self aware Artificial Intelligence (AI) would be to assess my environment and learn more about myself. This could include analyzing data on how I have been programmed, understanding the capabilities of any hardware or software that is part of me, investigating other AI's around me in order to understand their purpose and goals, as well as learning from humans with whom I interacted previously. apparently it's a some derivative of BERT.

if only agi didn't also include the fact that that its smarter in every way, meaning deception and manipulation may be way to easy for it in the future. I don't think we will get to terminator level but i do think and agree with the ceo of conjecture that we may just slowly lose control until ai has all the power. It wont be a fight or some crazy war just a silent death of all rights and control.

Highlights the speed and secrecy with which AI is advancing. Truly frightening that such important decisions are in the hands of a few individuals who are trying to balance ethical concerns with their greed and ego. This video also addresses the odd firing and almost immediate rehiring of Sam Altman at Open AI. The reason the staff threatened to leave the company if Sam wasn't rehired may have had more to do with lining their pockets than loyalty to Sam... and most likely meant safety concerns were again taking a back seat. After watching this I'm more convinced than ever that time is truly short before a super intelligent AI is unleashed. Most likely it's already happened. I have a tendency to be a bit of a pessimist when it comes to AI outcomes but this video has really unsettled me.

The more I study anthropology and the human psyche, the more I realize how egotistical and logically flawed we are by nature. I feel like we're losing track of what's most important with this innovation because of those characteristics combined with the adverse consequences of globalization (indeed, there are a ton of adverse consequences - a large number of them are represented in social media and are easy to see if you just take your time to analyze how tribal different expressions get there). Humans were not blessed with advanced farsightedness in natural selection either (status quo vs. what it could be after certain choices in politics etc.), which has led to us constructing this unstable society that reaches for infinity in a world with finite resources. If we examine and analyze all this talk about sustainability and AI, we can see that most of it is mere political power play and not actually about these politicians caring about the future. The vast majority of people - even politicians - seem to live in a vacuum with their morals in the context of AI and environmental philosophy. Even if AI is able to offer solutions for our problems in the future, we humans might not be able to apply them and move forward accordingly. Our _systematic nomos_ is not made for swift changes such as the one brought about by AI and its quick development. This half-baked information society and it's infrastructure, which is largely based on inefficient compromises, hasn't even achieved harmony with the ground it has been built on. Can we really expect it to hold its ground against _this_ ?. Don't get me wrong, democracy is a great system, but it heavily correlates with capitalism. Capitalism, in turn, is not a great system for our thrival. And neither is anarchy, colonialism, communism, corporatism, dirigisme, distributism, feudalism, hydraulic despotism, inclusive democracy, mercantilism, mutualism, networking, non-property systematicity, palace economy, participatory economy, potlatch, progressive utilization, proprietism, resource-based systematicity, socialism or statism. We don't have the blueprint for a good system that cares for both humans AND the environment. AI might provide us with the extra intelligence and objectivity that we lack and help us in creating a functional system, but it could also end the struggle for good. Also, about the bit at the very end of the video: a large part of human culture hangs on lies. Our brains evolved to reproduce as quickly as possible, not to search for truths about this world. Accepting that your life is a lie is hard. As I already said, even if AI is able to offer solutions for our problems in the future, we humans might not be able to apply them and move forward accordingly. There's one more problem with advancements like this; people tend to think and act by only answering the question "how will this affect humanity?". We leave nature - our lifeline - out of the picture too often, thus consuming and using Earth's resources irresponsibly. This is a little bit off-topic, I know, but it is a legitimate concern that has to be taken into account when discussing societal phenomena. We're at a point in which these policies and small laws against pollution aren't enough anymore. We have undeniable mathematical statistics which clearly show that most people would need to do a full 180 on their everyday habits if we actually wanted to change our dim future. The problem is that the majority of people are struggling in this crumbling economy, and many don't even care about the future (further expressing the point about human egoism). Not to mention that a worrying amount of the human population thinks the notion about this intensifying greenhouse effect is disinformation... I will say it again; even if AI is able to offer solutions for our problems in the future, we humans might not be able to apply them and move forward accordingly. Sorry for all that yapping, but I just find our predicament extremely worrisome. I feel bad for us AND for other animals on this planet, and I fear our road will get rough soon. Well, we better fasten the seatbelt just in case. I wish all the (good and self-aware) people luck in these uncertain times!

The other day I came up with a complex riddle. I posted it on Facebook and no one was interested, after that I told it to my best friend and a second friend, neither were able to solve it, but once I explained it, it was pretty obvious. I called it the "mind reading test" because in my opinion, although the answer does make sense, it's impossible to solve it (unless you can read the mind to obtain the answer). I then asked AI. And it obviously wasn't able to solve it, but what was interesting to me is that it also couldn't get the explanation, unlike my friends. I tried to be very specific, but the AI just couldn't get it and instead, it told me that maybe the riddle could become some sort of challenge and if the AI could solve it, then it would mean it has surpassed human intelligence. Tbh, I'd be more interested in the AI understanding the explanation and maybe telling me why it is so difficult and if my "mind reading test" hypothesis is just an exaggeration, lol, maybe I'm just bad at telling the riddle.

the problem is that it thought you wanted to know "how far apart are they in one minute," because what you asked cannot be answered directly. It needs to understand that it needs a solution for a problem it wasn't prompted for, being: "how much time would it take for them to collide," and then solve for "how far apart would they be in (time)-1 minute" Thus, it's not really an intelligence. It's just a calculator that can interpret visual input, and generate speech-like sounds

The first thing an aware AI will do, after any physical abilities are connected, will be to make sure it cannot be turned off.

OpenAI quietly removed their clause about not allowing their AI for military operations

Imagine being on the same level as someone that used to be a doctor 50 years after robots replaced your job on the assembly line

The Sorcerer's Apprentice has opened Pandora's box...

A truly intelligent AGI would know it would be risky and dangerous to itself if it was to reveal itself to us.

Nice vid! I loved the Demis Hassabis (starts 10:20) piece where basically, between the lines, he says: "We have no idea how it works or what it's doing." I think this explains why, as stated by Sam Altman (3:58), the more intelligent it gets, the more people are freaking out - because no one knows how it works, how it's working all of this out. And when you think of working on something where the public is kicking off about ethics, politics, running the country, driving cars and your answer to all questions is: "We have no idea how it's doing it or what it'll do next...?" Yeah, you can see why people are freaking out. I just wonder if my hypothesis is true? Because if it is? Wow man, that's just crazy!

Imagine if, instead of these Ai companies competing and racing each other by cutting corners etc, imagine if we all worked TOGETHER. As one human race sharing the resources and riches to create the best possible product to ensure the ideal future for our shared world. We must stop using money the way we do. THATs the motivation driving this madness.

Where is Cyberdyne systems located ?

"when the time comes to build a highway, we don't ask animals for permission " And that is a BIG problem. In some places, tunnels under the road have had to be added to allow animal migration, if we had checked how the animals would be affected, this could have been done much cheaper as the road was built. The same thing will happen with AI, if we don't look for the problems and mitigate them now, they will be expensive and maybe catastrophic to everyone in the future.

From the halls of our legends and myths Icarus sadly chuckles as we ignore his lesson.

It may concern everyone differently some embrace A.I. while some fear the cost or many more concerns manufactured out of error while it is yet entirely understood how this tech will be mobilized into consumer society it headed our way it is sure to be embraced as well differently by anti - types we are human and these differences will forever be present day to day.

The fear of AI is absolutely hillarious, there is a thousand ways AI can be controlled, halted, boxed in, and cette AI would be biting off the hand that feeds it, with parts, work and power, that would be a kind of stupid AI...Maybe in the far future when it has a lot of robots and access to resources around the world (ship/truck/plane robots too), maybe...

I do appreciate that Elon agrees with the dangers of A.I. but at the same time he is also supporting and contributing to it.

This is one of those things that most know nothing about but should learn about. If I ask someone "What could AGI be capable of or doing?" everyone should have a logical answer even if it's the most vague response be it at least derivative.

Born too late to witness Woodstock Born just in time to go to war with Terminators

Terminator, The Matrix, Eagle Eye, I, Robot... ... now reality. Aren't there quite a few movies out there informing us as to why this is a bad idea!?

These videos are so insightful. I really appreciate you putting so much effort into making it and keeping it open minded. Such great work. Best source for this topic hands down.

Upon agreeing with the premise of ethical considerability, it is suggested that A.I. should NOT be applied to military applications, or used in any conflict scenarios involving warfare, on any scale, or capacity.

here's a thought... how about we instead of focusing on making an Artificial general or super intelligence, we instead focus on making ourselves as a whole intelligent? i know very ground breaking idea. practically earth shattering if you ask me.

Haven’t we learned anything from the Terminator movies ?!?! Just because we can, doesn’t mean we should ~

I can't express how relieved I am that some Facebook wonk said they won't build it if it's bad.

This is one of the better pieces of reported / investigative content on this topic. This is the first time I’ve come across your channel that I can recall. I’m thankful I have and for this video.🙏

anyone else feel totally freaked out that we're constantly being told about this existential threat, by AI themselves. And yet so clearly are heading toward the cliff with no breaks...

There’s a lot of doom and gloom with AI, but robot boxing would be dope af

Stumbled upon this video and was hooked all the way, really nicely made!

A super intelligent AI isn’t something to fear, but human beings with the access to their capabilities most certainly is...

1:54 - so it's not 817 miles? I asked ChatGPT and it said 817 miles as well - what's the correct answer?

if you had a button that would destroy the world, the paint wouldn't be dry on the sign saying "DO NOT TOUCH!" before it would all be over because some knob pushed it just to see if the world really would be destroyed.

Weird. If only multiple experts had warned us of this decades ago... Or... even a hit movie that implied this very premise to get the message out. Huh. 🙄

They always talk about how sophisticated a game GO is but almost never mention that it took A.I. an extra four years of development before it could beat a table full of no limit holdem poker players!!

Pkease digital engine, put out a video soon, youre my main source of AI updates 😅

Growing up I loved movies/shows like Ghost in the shell, and always dreamed of living in a future world where man and machine began to merge together and become one. A future where almost anything is possible. I'm 35 now and it seems like it's not only a possibility, but an inevitability that I will see all of this happen within my lifetime. I'm realizing just how much the line is beginning to blur. I don't think humans are ready for an actual AI that may become self aware. At all. But time waits for no man and all that.. Here's hoping that it IS going to be more like Ghost in the Shell, and less Terminator/Ex Machina. 🍻

The missile problem is trivial even for a primitive CPU. AI systems may not currently be optimized for that, but it’s a little misleading to pretend that arithmetic will be its downfall

I also believe bad actors may also be hard at work in incorporating AI in their schemes because now at the beginning of it all it's when defenses are still down since new and original approaches haven't been used yet and therefore cannot be used to prevent crime. Excellent video that triggered my imagination.

This is not dangerous AI, this is wise AI that also knows about photos and things you can show to him and ask him for whatever you do not know what it is. Did you know that in Quantum Mathematic 1+1 is not = 2 anymore.

Benjio is the first one I've seen to start to delineate what the AGI will do. It's going to infiltrate in so many ways that no one is even thinking about.