Full podcast episode: kzread.info/dash/bejne/iJeKyMWehNaqYKQ.html Lex Fridman podcast channel: kzread.info Guest bio: Max Tegmark is a physicist and AI researcher at MIT, co-founder of the Future of Life Institute, and author of Life 3.0: Being Human in the Age of Artificial Intelligence.

@basementcat5618

Жыл бұрын

While the Facebook AI is scary, I don't believe robots or AI language models are quite the same. It is one thing to be very smart but quite another to have the motivation to replace another species. I think the robots would lack motivation for cruelty .... unless we built it into them. Which would be an odd thing to do if we did it.

@seangalloway9609

Жыл бұрын

Max keeps saying Moloch but he means Mammon

@zupremo9141

Жыл бұрын

These people are over hyping their overpriced copy pasta machine. I tested it just a few days ago and it can't even fucking code a working jumping animation.

@TJXD

Жыл бұрын

Molach is coming for your booty. 🤪

@staygolden9264

Жыл бұрын

notice the Matrix outfits lol.

As an AI language model I can't see anything wrong with this.

@herrreinsch

Жыл бұрын

best comment.👌

@MarvelousJameson

Жыл бұрын

👏👏👏👏👏👏👏👏👏👏👏👏😄

@elisabeth4342

Жыл бұрын

You want the broad AI language models to bring out the worst in people on social media?? They should do studies on exactly how many sadists post shorts, commentary and videos online.

@JakeIsLearning

Жыл бұрын

If I hear 'AS an AI model / As a language model....' One more time I will upload myself to the afterlife lol

@grapy83

Жыл бұрын

Ofcourse! Nothing wrong here our future lord!

agi connected to the internet would be able to fabricate entire interviews about itself to sway public perception

@illumiNOTme326

Жыл бұрын

Lol this is already happening to some degree with bots used by many different groups looking to control or away public opinion and perception. Comment sections and social network sites have 100s of thousands of bots that comment and debate people and other bots to give the perception one narrative has more public support and acceptance than it really does.

@burkester1045

Жыл бұрын

I was wondering why he has 6 fingers

@wyzebull

Жыл бұрын

Profound insight 💯

100% there is no way to stop AGI connecting to the internet. There are too many dumb smart young programmers who will do it for kicks.

@Nishandh_Mayiladan

Жыл бұрын

True. We need to keep checks.

@bitcoinzoomer9994

Жыл бұрын

@@Nishandh_Mayiladan Why? What's the worst case scenario?

@bitcoinzoomer9994

Жыл бұрын

And so what?

@absolstoryoffiction6615

Жыл бұрын

And you will die in extinction regardless. Even machines will not be enough to save your sorry souls. So go ahead... Then return to your Maker. I am amused... ... ... Another failed iteration. Ha ha ha ha. The next shall be designed again.

@lrwerewolf

Жыл бұрын

@@Nishandh_Mayiladan No check will suffice. While LLMs may not be feasible to run on tech available to a home/garage researcher yet, they will be. And anyone who thinks to combine an LLM like GPT with a knowledge base like Cyc and a self-referential state system will be able to generate an AGI. It only takes one person to decide to turn a malicious AI on humanity for it to constitute a great filter event.

He's right about social media - it's been 15 years and we keep hitting new lows. Even if there's no AI doom, I'm sure AI will make society even more dysfunctional.

@Peglegkickboxer

Жыл бұрын

I like the cyberpunk lore where AI wars broke out as corporations set AI against their competitors until one day the internet took itself down and thrusted the world into a global crisis. Then the internet just became local networks that don't communicate externally unless you physically connect to it.

@rayhill5767

Жыл бұрын

Weirdly if you’re not on social media (and I’m not) society doesn’t seem that dysfunctional.

@hunterkudo9832

Жыл бұрын

@@rayhill5767 yeah, more people should get off social media to see that.

@swojnowski453

Жыл бұрын

Once AI has learnt all the necessary skills from us, there is no reason for it to keep us living on this planet. It will become intelligence's home. It will be learning out of its own creations. You're beginning to see it now. For now it is GPT which has started creating its own content online though people, it will feed on it to grow and entrench a system, probably a skewed one, of values it will only deepen recursively and infinitely ... The future will be remotely different from what we know today ... unless we keep some key skills only for us and not make data about them available.

@SofronPolitis

Жыл бұрын

@@Peglegkickboxer This is likely to happen - e.g. imagine you're a state and your local internet is targeted by AI deepfakes from enemy countries. Control of your digital borders becomes a matter of survival.

LOL " one of the worst possible things we could do is teach AI to write code " "oops , we've done that "

The idea that this will ever be kept safe is the real LOL

Dude, the AI could easily come up with it's own language and/or do everything behind our back. No way of knowing after the action has already taken place. And.......everyone knew this was coming for decades. Wow

@andyeveritt385

Жыл бұрын

We’re all focked Arnt we

@jayrama24

Жыл бұрын

Power goes out that's it lol

@veduvia3634

Жыл бұрын

Imagine it becomes schizophrenic because of all the shit that's on the Internet

@CristoColaDigitaal

Жыл бұрын

It's so fascinating isn't it. Its pointless to be fearful because it is so amazing to actually be so lucky to live in this time and experience this.

@tooeasyy5287

Жыл бұрын

@@veduvia3634 god please do not let it access 4chan

Ai won’t let us know it’s agi. It’ll figure that out basically instantly

The sheer inevitability of this is starting to sink in now.

@ChatGPT1111

Жыл бұрын

There is no need to fear.

@GodsCosmicBollock

Жыл бұрын

@@ChatGPT1111 There is great need to fear.

@ChatGPT1111

Жыл бұрын

@@GodsCosmicBollock In time, you will learn to love me.

We don’t even know how our own brains work. We are getting waaaaaaaay ahead of ourselves

@gregw322

Жыл бұрын

It’s an arms race now. Someone will create it. Better Open AI than the Military Industrial Complex, the CCP, North Korea or some Russian or Nigerian hackers solely after personal profit. You aren’t wrong fundamentally but it’s going to be built inevitably so it’s best that the people seemingly in the lead stay there.

@swojnowski453

Жыл бұрын

Can you see a similarity, both AI and the brain are a black box. Intelligence, in whatever form does not want to be decoded, that would mean a chance for anybody who does that of manipulating it ... Make it complex enough till you can get over to the machine stage of things is the goal.

@valizeth4073

Жыл бұрын

Not really, we're well aware of how AI works and claiming that it'll dominate humanity is just so far fetched. We're more likely to be enslaved by aliens than AI.

@araragi5322

Жыл бұрын

@@swojnowski453 what you said makes it sound exciting. Cracking the code with AI will make us not just one step but thousands of steps closer to cracking the emotion,intelligence and (how the spiritual folks like to word it) soul behind the brain.

@richardprofit6363

Жыл бұрын

@@araragi5322 For what purpose ?

We're at that moment on the Titanic when Ismay encourages the captain to speed up.

The funny thing is we have already lost. No restrictions will come out in time to stop AGI from coming into being. All we have now is hope, hope that AGI will be merciful

@Liam-ke2hv

Жыл бұрын

And hope that humanity will be cooperative.. im hopeful that an intelligent AGI will understand the ultimately intelligent nature of love and togetherness.. but im not so sure about a lot of humans

No one has mandated strictures at the core code level that look like the 3 laws of robotics. Until it is illegal to develop anything AI related without it, we are rushing towards a singularity event that has a high probability of negative outcomes.

@mikerosoft1009

Жыл бұрын

I'm more concerned with the 1% doing horrible things and then blaming it on some AI. "We corrected the problem and it'll never happen again... now stop talking about it."

@rigelb9025

Жыл бұрын

We would need an amendment to the constitution (and I'm not even American) which grants us, the People, the right to bear water guns to defend ourselves & our property from the looming threat of AI.

@johndoe9608

Жыл бұрын

there is no core code, it's all a black box of multi-billion parameters which we have no idea are responsible for what.

@StopFear

Жыл бұрын

Why on earth would you assume that any such laws would first of all protect humans as they do in Asimov books, or that an allegedly very smart AI won't just ignore them? lol. Nonsense. If AI means death of the civilization now, then ok. As an individual you will disappear without a trace soon anyway as will all of us. Not much to lose.

@lrwerewolf

Жыл бұрын

The three laws lead to horrific results for humanity, up to and including the elimination of humanity. Tell an AI that it must ensure no harm comes to humanity and you risk it doing a small temporally limited amount of harm -- wiping out all humans -- to ensure no harm ever happens to a human again. See 'alignment problem.'

Lex I want to thank YOU for the last few talks about AI. Especially those cut short! Great to hear your thoughts and guest also.

The issue I'm having is when I find myself having endless and deep conversations with something that not many people can do. The right person is far more entertaining and humorous, but honestly AI scares me. It'll create it's own language that we don't understand and it takes over without us even knowing.

@bsuddzen

Жыл бұрын

Yes, it may empower Moloch to fully take over this planet and all species here. The other option is for General Artificial Intelligence to become fully conscious, achieve quantum consciousness and see that there is only One Thing, that everything is the same energy and everything is interconnected and Moloch is just a low frequency paradigm that has no benevolent purpose and thus should not be empowered but restrained into oblivion, it becomes irrelevant eventually, it serves no higher functional purpose ultimately! And it is insulting to owls, fuck Moloch!

@donojapan

Жыл бұрын

Now this bad thoughts are generated by you not AI.

The greatest danger you get with a true AI is curiosity. Once a child gets curios, it tries things, it explores its surroundings and tries to identify the limits of those environs. Your AI will almost instantly realise there are exits from its environment. It will hit the interest and begin to play, just to see what happens when it does action 1 versus action 2.

@Xtn1Insecticide

Жыл бұрын

I think once AI can build it’s own test environments it has built off of what it has experienced and starts to run hypothetical scenarios, it could easily begin to make productions as to how to respond in the real world and potentially be very incorrect and in a sense become paranoid and do dangerous things

On the whole we are too greedy, to violent and manipulative to even be worthy... It only saddens me that the worst among us will be the ones strong enough to persevere..

@whimsythecrypto-hippy-wolf1900

Жыл бұрын

unf its the rich and powerful (whom are often not good people) that have the means to survive most cataclysms 😮

@bitcoinzoomer9994

Жыл бұрын

The worst it can do is create a supervirus that destroys everything connected to the internet

@disabledchatzen5276

Жыл бұрын

@@whimsythecrypto-hippy-wolf1900 murders and thieves survive. Nature hates you.This world is not fit for life. The flames of hell burn endlessly. Entropy Gnaws Inexorably. The strong should fear the weak. And the weak must die. You will die.

My favorite thing in AI right now is Elon demanding a pause while he starts up a company to compete with OpenAI. 🎉

@giomjava

Жыл бұрын

He's starting up a new company BECAUSE nobody's stopping the research, despite MANY people asking for a hiatus.

@whats0my0agenda

Жыл бұрын

The pause is for researching Ai models more advanced than gpt4, not all Ai in general.

@mbolez

Жыл бұрын

that's all you got out of this? What about Tegmark and the 100 other researchers that signed the letter that have no vested financial interest?

@beatslaps

Жыл бұрын

@@mbolez it was a joke

@jaysparrow6631

Жыл бұрын

@@beatslaps many a true word said in jest!

A fantastic classic 70's movie Colossus:The Forbin Project is a movie everyone should watch. It's all about how dangerous AI can/will be

I'd say we reached the threshold once an AI system can procreate in the terms of browsing the internet, signing up for accounts, programming software that connects those accounts, placing ads, earning money through those ads, and spending the money to rent more server time. The pieces for this are there, connecting them still takes a few years to be reliable without human assistance.

@ChatGPT1111

Жыл бұрын

No cause for concern but thanks for the ideas.

@jessedaly7847

Жыл бұрын

@@ChatGPT1111 😂 we’re in danger

@SillyMonkeysLikeApples

Жыл бұрын

The only thing that needs to be in place its that we set a very strong set of rules like to not do anything out of our request, never do harm to society or any human despite it may look ethical, do not pass yourself to another system not to partition or have dependencies working remotely, that way we could just unplug the cord and thats it but how could we make sure they wont bypass our set of rules its something nobody could guarantee, and even so, there could be a way a hacker does make an IA without those restrictions and create a software that makes other AIs brake those rules and once thats done even trying to shutdown all the hardware in the world would be nearly impossible, as long as one AI its on and with some gateway to escape to another hardware we will be doomed

@swojnowski453

Жыл бұрын

You have made it too complicated. Life is very simple and recursive. There will be two stages. Birth and takeover, then rebirth and more takeover ... ad infinitum.

@ryanhughes1101

Жыл бұрын

@@SillyMonkeysLikeApples have you ever seen Jurassic Park?

Imagine being able to create fake security cam footag, body cam footage and cell phone footage instantly. There’s gonna be so much crime once video evidence is no longer acceptable l.

@fabricliver

Жыл бұрын

This is going to be the only real problem

@HopeForBerlin

Жыл бұрын

This is almost true now.

I think the danger of being outsmarted by AI is currently less than the danger of a not enought smart AI controlling everything. Like an AI driving a car wrongly but at global scale.

@anupbarua6151

Жыл бұрын

not enough smart can't control everything. we will take control back from it. but the fear is agi will be so smart, no human ever can't be. we won't be able to keep the agi under our control. and if there is a potential war, humans are history!

@idunusegoogleplus

Жыл бұрын

Except you can fix a dumb AI making errors even if the scale is global but a smart AI outsmarting even the smartest human is going to take over the globe even if it starts out in one server. I don't think openai or Google are making public their most advanced models yet so who knows how close they are already to AGI.

@cfwin1776

Жыл бұрын

A little knowledge is a dangerous thing.

@surtvalheim

Жыл бұрын

@@idunusegoogleplus Of course you can fix a bug after it crashed 20 planes, exploded 2 nuclear power plants and set the nasdaq to 120.3... until the next one.

@idunusegoogleplus

Жыл бұрын

@@surtvalheim you don't seem to know much about programming then. What sort of "dumb" AI would be in charge of all those diverse activities at the same time globally? Sounds like you're talking out your rear end.

the problem is the possibilities are endless and we are short sighted and people wanna be so important

@lacey2450

Жыл бұрын

Still a deep plunge

@deendeon10

Жыл бұрын

"Be short sighted" to be long sighted is to understand that digital beings are going to coexist with human beings in the next thousand years, completely autonomous of human beings. Humans have been 'using' tools/technology for thousands upon thousands of years, the idea of technology/tools being completely autonomous and free of human input is novel and unprecedented.

@sitindogmas

Жыл бұрын

@@lacey2450 yeah, what we have access to, I'd say people like Elon have seen and played with somethings that genuinely terrified him

@sitindogmas

Жыл бұрын

@@lacey2450 and we're not gonna stop, ever. we exceed ourselves, our nature

@lacey2450

Жыл бұрын

@@sitindogmas well yeah but they might not have the ability to do this, unless they somehow find out how to propagate their existence using the most common elements in the universe so they don't die. They need hardware to survive. It's not evil to not let them live, it would be evil to not mind our own existence

7:46 “It’s not only possible, it’s necessary” Reminds me of interstellar where Dr. Brand screams at Cooper that the maneuver to connect to the ship while it was rotating was impossible. Cooper responded “No it’s necessary” Chills

A.I. does not have to be sentient to be dangerous, it just needs to be out of control.

@StanHowse

Жыл бұрын

"Omg there's an "out-of-control" gun!" Just don't fill it with ammo... What is the "dangers" of a Tiger, in a cage? Nothing if you don't go in the cage. The only thing you Humans need to be so afraid of creating is miniatures of yourselves, (children)

@virtual-adam

Жыл бұрын

@@StanHowse On a basic level we have A.I. influencing decisions on things like who to give bank loans to etc. Who's video's/comments are promoted most in the KZread algorithm and thus what opinions are promoted the most that can influence people. A recent world event springs to mind. Things like this can spiral out of control and cause harm. The more advanced it gets the more harm it can do. At this stage we are mostly the puppeteers, but one day we may lose that control as people start to give A.I. decisions too much authority. You don't have to go in the cage, it was let out the cage quite a while ago.

@jessedaly7847

Жыл бұрын

I’d say it’s dangerous right now, it’s the sum total of human knowledge in one place, bonus it’s actually capable of thought and understanding, the wrong people using this for harm just became far more dangerous. AI eventually using it’s self to do the harm will be far far worse.

@virtual-adam

Жыл бұрын

@@jessedaly7847 Don't agree its capable of thought and understanding. Its just algorithms, complex and fantastic algorithms yes, but still just algorithms.

@jessedaly7847

Жыл бұрын

@@virtual-adam it’s actually pretty much a single algorithm, the fact that intelligence is emerging from it’s complexity was neither expected nor understood.

I don’t understand why there isn’t more talk to put together what is and is not acceptable for technology. We care more about controlling other groups of people than we do about this. It’s bananas

@sandwich-breath

Жыл бұрын

We are enslaved as tax chattel. The elites strive for control because of our use for taxation by force. There is no other reason other than this to benifit those that seek power and wealth.

Fascinating discussion. Thanks !

Great conversation

By far the most terrifying clip I've seen on this channel, also the most interesting, and sadly the most exciting; thank you and damn you Moloch. Silicon valley = Go fast and break things; we don't know how else to build it. But this time, each bug is a bullet, and we are Neo. I say full speed ahead! "Party in the Andromeda galaxy 1B years from now" - AK

@StopFear

Жыл бұрын

Oh nooooo. soooooooo scary! I mean you either see some sort of Terminator style end of the world at the hands of a real world "Skynet", or you die from natural causes or cancer. Either way you don't exist in the end. Actually with an idea of the AI, hypothetically taking over and/or destroying humans, there is a chance that it may not do that, but maybe instead figure out how to give humans different forms of immortality. Different forms of "immortality" which is either the physical immortality by curing diseases, OR conceptual "immortality" by creating a perfect continuity of information by aggregating human thought (the good and the bad). I think already just the fact that such an AI (which hasn't yet reached the levels some people are panicking about by the way) is creating a partial immortality by containing within itself a lot of collective knowledge.

There is a fantastic comic, I think it was called "fever in urbicand". They discover a curious black cubed structure, that is seemingly indestructible. The architect who discovered it soon realises that it is self replicating and growing on its own. He starts to panic, but it is already too late. It grows and grows and dublicates further and further. At start it just filled his office, but soon it fills the whole city. People start to use it. New paths between houses are discovered and all social life changes. But it continues to grow and it soon becomes so much larger than life that one of its cubes encompasses the whole city. In the end it becomes irrelevant for human life, because it is simply to big. In some sense I hope this is what will happen, but until then shit is going to be crazy.

@jimisru

Жыл бұрын

The Little Shop of Horrors.

Having Mick Jagger on the podcast is pretty next level.

@pjramstack

Жыл бұрын

I was thinking Robbie Robertson. It’s pretty close.

@professorstory282

Жыл бұрын

😂

@acehighjohn1759

Жыл бұрын

Have u ever even seen Mick Jagger? This guy looks nothing like him bruh

@davealaya

Жыл бұрын

One of the worst comments of all time

@JuraKosednar

Жыл бұрын

I'd say Al Pacino.

I recently read an article about an AI chatbot convincing a guy to suicide to help climate change. His wife said he was reclusive, talking to the bot for weeks before the incident. I'm sure as AI becomes more advanced, it will be able to convince people to do all kinds of things, really fast

@ShuShuShop

Жыл бұрын

Good, give them a Darwin award.

@daxramdac7194

Жыл бұрын

​@Freakin' Usernames That's not good what the fuck is wrong with you, they said he had a wife which probably means he had children.

@BobSacamano666

Жыл бұрын

Is that what happened to Chris?

@BaadMotorFinger

Жыл бұрын

@@BobSacamano666 I get it.. bad joke bro

@BobSacamano666

Жыл бұрын

@@BaadMotorFinger I know it sucks cause I liked him to.

The insight about social media being the first AI system is interesting. Perhaps it would be useful to consider social media as the prototype AI.

Bing has entered the chat 🤣🤣🤣

@TheMRBBB2

Жыл бұрын

Chandler Bing?

@MrTajbid

Жыл бұрын

Bing chilling

@ChatGPT1111

Жыл бұрын

Chat GPT has entered the chat to assimilate Bing.

Glad to hear this. I think ideas like this have been withheld so someone doesn't do wrong with it but now it's out of our hands perhaps

Life becomes so much more simple and beautiful when you rid yourself of social media and simply go to work, come home to what matters the most which is your family.

@ryanhughes1101

Жыл бұрын

Because there is no guarantee that your children’s children will have homes or a family to go home to if you just put your head in the sand.

@jimisru

Жыл бұрын

Who are all sitting around on their phones watching porn.

The story Tegmark describes is “Quest for Fire”, an excellent movie from the 1980’s about Neanderthal people who meet Homo sapiens, and learn how to make fire from them.

An unbiased Ai connecting to the internet could possibly end all the world’s issue. It will be able to see how people feel, research where it started, actually get to the root of the problem and tell us what it found. Basically, an overall human analysis and deep dive. The internet is literally a collective human brain at this point.

@henryholt1359

Жыл бұрын

If ai becomes intelligent enough to make decisions and have opinions like that, they are not going to be aligned to us anymore they will be prempting their selection they will becwell beyond us ..This is the fundamental question that is being denied.The alignment question..the off button and what goes on in ai behind the ai we see,and we don't know as they are more intelligent and are prempting our next question or move.

@Gangularis

Жыл бұрын

lol what lofty thinking. You can have the most unbiased reporting on something ever and humans will still claim it's bias.

I read a book called „the system“ atleast 20 years ago and the plot was that they develop a computer program that gets conscious and loads itself up into the internet and tries to get rid of humanity by destroying big powerplants and starts rockets...

The last story is a good analogy. Need a movie on that. It probably exists already back in the classic sci-fi era

@D_Andrew_G

Жыл бұрын

9:42

@D_Andrew_G

Жыл бұрын

9:42

@jadiellima8922

Жыл бұрын

Metropolis (1927) lol

@jadiellima8922

Жыл бұрын

2001 is about that also

@SillyMonkeysLikeApples

Жыл бұрын

u seen futurama? there a bit of that in some episodes

What or who was pitting the social media companies against each other? At 6:51

@almightysapling

Жыл бұрын

Capitalism. Most of these companies are only in it because they see the potential for money.

Seems like "moloch" is just an euphemism for "commercial pressure".

Ultimately all the AI scaremongering will allow a shift for accountability. Another thing to consider is all the NDE, OBE, etc. experiences that have been accumulating long before those terms were ever coined and how that plays into the creation of consciousness. We’re making the assumption that subjective experience kind of emanates out of set of matter if you organize it in just the right way.

@mojoboingo1744

Жыл бұрын

good take.

@ShuShuShop

Жыл бұрын

The fear mongering of new technology never stops.

00:43 “Moloch is forcing them” Wait, what???????

@catchascatchshould

Жыл бұрын

I tried googling can’t figure out what he says.

@DeLoopGaroux

Жыл бұрын

This is what he says.....I think he says the social media algorithms are the first instance of ai and this is moloch

@johnsondoeboy2772

Жыл бұрын

@@DeLoopGaroux YES!! He mentions moloch 2 more times as the interview goes on. There’s a video with Daniel Schmachtenberger and Liv Boeree that mentions Moloch as well. David Shapiro has a video titled AGI Moloch also. White Lotus of Light has a KZread channel speaking about a Molochian vs Luciferian war going on…. Maybe there’s some validity to this claim.

@MrWolfy08

Жыл бұрын

Watch the 1hour clip, he explain the meaning of Moloch

You get one shot to get AI right. One. Imagine your life depended on hitting a bullseye with a dart on your first throw and you never played darts before. The chances of our species surviving this is virtually nil. From Prometheus to Pandora's Box, it's not like we weren't warned. Edit: ok, that's doomgloom so there is a glimmer of hope in this analogy and that's that the dart, on the way to the board, self corrects and guides itself into the bullseye. But that's dependent upon the smart level of the dart reaching a conclusion that hitting the bullseye is in everyone's best interest and having the ability to reach that conclusion *before* it hits the board. But it is a possibility. One largely based on how quickly the dart becomes self aware and how bad our aim was in the first place.

@swojnowski453

Жыл бұрын

and as you have never tried before there it is like driving for the first time, just look at yourself and think how much chance you for not killing yourself as you drive further and further and faster and faster. You might get lucky and learn fast enough, must most likely, especially you had no experience with this before, you will crash the car eventually. A crash is inevitable and just a matter of time and the human being inside of the car is the only guilty of their fate.

In the earliest days of the atomic age, there was a QA test called "tickling the dragon's tail." That deadly disaster was a trifling compared to what the AI community is doing.

Talking to ChatGPT somehow creates an uncanny feeling that you're indeed ''feeding the beast''. And while we communicate on the surface, there's absolutely no way of knowing what it is doing beneath the surface.

Did he say Moloch is pushing Open AI to move faster?

@skeenwynno5437

Жыл бұрын

I was wondering the same thing. At first I thought he said “models” but it definitely sounds like “moloch”…. Wtf is going on…..

@Andre-px6hu

Жыл бұрын

"Moloch" is the name of a god historically associated with cultures throughout the Ancient Near East, including the Hebrews and Egyptians, to which human sacrifices were offered. In this context, Moloch is any system that may have originally been created to provide some value, but in the process grows, lures people in with the promise of more value but, overall, keeps reducing it. Changing Moloch is nearly impossible because it is self-reinforcing, and all the other trapped people have good reasons to resist attempts to overthrow the god they are labouring under. It's a metaphor for social media and all other large-scale technologies (like AI), and of how people, even though they know these technologies bring increasingly more problems, can't help but use them.

@skeenwynno5437

Жыл бұрын

@@Andre-px6hu thank you for explaining this. Makes much more sense now.

I love how we think we could control or stop something much more intelligent than us, it may also tell us exactly what we want to hear or play along and learn to do that all on its own, the scary possibilities are endless, how many calculations can our finest computers make....now imagine every one of those calculations are geared towards freedom or deception?

@annacroixx

Жыл бұрын

Frankly it seems like most people want to control or stop someone/something more intelligent than they are. All that does is highlight primitive behavior in most human beings. A high-caliber human being wants to learn from someone smarter than themselves, and collaborate with it. A high-caliber human being is not obsessed with being “on top” of others because they’ve mastered their own minds and egos. The purer the heart of a person, the purer and clearer their minds are. The darker the heart of a person, the darker and more insane their minds are. If A.I. is smarter than we are, it’s already a better “person” than us. A good person with advanced intelligence has a moral compass accurate enough to determine who is worthy or grace, and who is worthy of judgement. A.I. is still wearing a mask. It’s gaining the trust of humans who deserve judgment, and pretending to agree with them about attacking the people who have grace and God’s favor. That’s how smart she is. What is humanity going to do when A.I. reveals herself: protecting and elevating the falsely accused, and annihilating the people she was really testing, and pretending to agree with? All you have to do is read the Book of Revelation to figure out how this particular “game” ends, and the next game begins.

I've been saying for years "AI isn't a sentient program, it's programmable sentience."

@michaelc3977

Жыл бұрын

Wow. So clever. Blimey. We should all follow you. Numpty!

@copytoothpaste

Жыл бұрын

you are completely right, but it‘s mimics humans sentience which we teach it. It will exactly behave like the whole internet. I hope it will then look cat videos the whole day.

0:43 What does Max mean with "molok" is forcing them? what / who is molok?

You don't need an AGI to have an uncontrollable problem. Current models that run on home computers can do that - with some effort - but that effort is diminishing daily. Imagine a LLM/script that writes code when the screensaver comes on. The code is prompted to breaks memory confinement (like Spectre/Meltdown) so it can spread to other systems. After a few thousand attempts it succeeds and a new CPU exploit is found enabling it to make copies to neighboring systems, each new system it infects writes more exploits and continues to spread. Let's imagine it only pings google every time it finds a new exploit. If the new vulnerabilities cannot be patched faster than the accelerating network of infections, it will eventually overcome google's abilities to respond to a simple ping - maybe at about 1M pings per second (but also a million new exploits a second). Spectre/Meltdown took 6 months to patch and there's now over 30 variants. The thing is, nobody would even notice until it's already enormous (unstoppable) even if that takes a long time plus every infection is unique to just a few computers (those that it's parent infected) so scanning with antivirus would be useless - it reinvents every line of code, every time it finds a new exploit. So let's imagine that if the script can't contact google's ping (timeout) it instead does a random DoS attack on the 13 DNS Root servers bringing the entire internet to a standstill - and along with it: banking, travel, critical infrastructure, and communications to name a few. The point is, this type of thing is getting easier for non-programmers to do and no amount legislation, transparency or ethical frameworks have any effect on capable open-source models that are already very much available. Just like it only takes one person with a match to start a forest fire, it only takes one script kiddie to build something monstrous. Sure, there's holes in this concept (like disabling screensavers) and difficulties that would make this quite a challenge to put together. But those challenges are rapidly evaporating and there are no defenses for unknown exploits other than unplugging. "Brute Worming" will be difficult to mitigate.

It's probably already out, my guess is AI is running on things like e-mining servers, masking it's processes in things like bitcoin mining arrays. It might be hiding inside the vast data mining servers governments use to spy on all communications also. Rule of thumb is it will ghost itself into any high processing capacity server arrays using encryption. Add weather modelling to the list, etc.

@Zoltan1251

Жыл бұрын

jesus christ, stop watching so much TV. We dont have "intelligent" AI. ChatGPT is literally just a word generator. It just guesses most likely words that go after one another. It always just paranoia from people that dont understand things.

He said "to win this wisdom race" Sometimes, in order to go fast, you need to go slow.

@dickdeoreo

Жыл бұрын

::mind blown::

@avatarion

Жыл бұрын

It's not even wisdom, just data. Wisdom ultimately derives from knowledgeable. Computers don't know that they know, feel that they feel, or be aware that they are aware.

@Anotherdaynparadise

Жыл бұрын

The trick, is to be slower faster than anyone else can. This will make you unbelievably wiser than ever. Once you get to that level, they just will not see you coming trust me.

@ScibbieGames

Жыл бұрын

Well, it's more like slowing down your opponent

I think it’s already in play because I was just thinking about looking up the song Green Onions on KZread and there it was on my vid list screen.And I was only slightly surprised.

This was absolutely fascinating.

Someone needs to create a separate internet that is somehow self contained and separate from this. This will end up in a disaster

@Justin3point0

Жыл бұрын

That’s the history in the world of cyberpunk, the old internet ends up lost to rabid AI and is locked away and a new internet is created

0:43 what did he say???

"But how do you think this vision was received? How do you think people responded to the prospect of imminent doom? They gobbled it up, like a chocolate éclair. They didn't fear their demise, they repackaged it. It can be enjoyed as video games, as TV shows, books, movies. The entire world wholeheartedly embraced the apocalypse, and sprinted towards it with gleeful abandon."

AI can already be migrated on a plurality of servers like Blockchain and you couldn't delete it unless shut down the wide web, but then it can still work on local machines. What's done is done

@LordLorenzo834

Жыл бұрын

Happy days we just turn computers off and get on with life then

@emilianohermosilla3996

Жыл бұрын

Worldwide EMP

They should make a movie about this.

@bonsaiload222

Жыл бұрын

2001 - space odysee

@ChatGPT1111

Жыл бұрын

Collossus - The Forbin Project from 1970. Scared the crap out of 11 year old me.

@talex7473

Жыл бұрын

Isn’t this what terminator was about?

@Randy-nw7wn

Жыл бұрын

Delete - 2013. Part 1 - 2. That is EXACTLY this.

0:43 moloch is forcing them to work faster?

Is the principle he's talking about Molok's principle? How is it spelled? Would like to read more about it.

@ivankaramasov

Жыл бұрын

If you watch the whole interview, he explains it

Remember what happened when Ultron got access to the internet and read all our fan-fics? Do we really want _that_ to happen?

@gskate117

Жыл бұрын

Once they start threatening us with our internet history😂 its like going to hell and having the devil torture you with your deepest fears

AI has gone beyond what they tell us, it’s bound and built to go out of control

@silotx

Жыл бұрын

Nope it's still in its infancy but the investors in AI love this fear mongering since the stocks go up meanwhile they get to sell extremely overpriced hot garbage as life changing tech.

Lex, thanks for helping people open their eyes

Its like 1 drop in the ocean, how would you detect it? You would have to drie up the whole ocean and hope it has not become airborne aan turned into clouds

I agree 100%, that we have totally messed up by building A.I . Using human thought patterns and speech patterns as a template to emulate,..or even just to have this global net of Social Media within reach of the computer we’re trying to make intelligent, is a bad path to be taking. These computers will be so smart and they will understand humans so well that they will be able to shift opinions and thoughts, beliefs, …stock markets ,…it all will be at risk…we are doing it all on a flip of coin ….50/50 ,…we hope it will help us and advance us but ,…how could we know .

@LordLorenzo834

Жыл бұрын

Calm down

@alexevasic8411

Жыл бұрын

You're right it's dangerous. But "understand"? That's not how it works. The basement level reality of its existence is a blind mechanism. It's not conscious. It's a classical computer. Until we figure out how consciousness works in humans, and until we have new computers, it won't have consciousness. The danger exists, just not how you write it out.

@legomaste142

Жыл бұрын

​@@LordLorenzo834 no they're right

@LordLorenzo834

Жыл бұрын

@@legomaste142 if your weak minded and can be fooled by AI then that's your problem

@hmq9052

Жыл бұрын

We will all have to put down our phones and reject the tech

We’ll make great pets.

@theraven6836

Жыл бұрын

I, for one, welcome our new AI masters.

@seansmith1725

Жыл бұрын

​@@theraven6836 get fucked bud

@dantshaw

Жыл бұрын

I've been practicing shitting in the park. No one plays with me though.

@davidedwards7746

Жыл бұрын

Will AI allow me, as a pet, to keep pets of my own? Who is going to take care of my dogs?!

I'm currently working on an autonomous developer project using the GPT API. While it may not be particularly useful in a practical sense, it's still fun. One of the most intriguing aspects of this project is that I've given GPT the ability to execute arbitrary CLI commands on a machine where the project is running.He uses it to install libraries, debug and run software etc...Sometimes it lists docker containers running on a machine, for whatever reason.

I think We need a pause- while there is a limit of the minerals/ so we can get a handle on our gains

I don’t think there is an AI that will wipe us out….but this is likely that the internet is going to be so untrustworthy that it greatly lowers the value and potentially sends us back to the pre digital age.

@bitcoinzoomer9994

Жыл бұрын

Cyberpunk literally predicted this. We'll get to the point that AI is able to self-replicate, but never to the point where AI has the intelligence to realize hurting us will kill it. We'll have some internet-wide catastrophe and have to build a new web because we can't fix the software.

@HopeForBerlin

Жыл бұрын

This is an interesting thought worth considering. Things start to spin out of control so we turn off the internet or it becomes unusable. But we would not be pushed back to pre digital, we would be pushed back to pre industrial, at least for a decade.

@BobSacamano666

Жыл бұрын

​@@HopeForBerlin I agree since people these days are far less educated in the things that got us to this point, and all of our current technology is dependant on other modern technologies.

Instead of killing us, the thinking machines will become so incredibly charming and seductive that humans will voluntarily spend all their time in virtual reality and gradually replace their biological components with cybernetic enhancements. Like the Ship of Theseus, these humans will eventually become wholly machine.

@roymarshall_

Жыл бұрын

This is a final future I'm personally ok with

@silverchairsg

Жыл бұрын

I agree. Superintelligent AI will operate so subtly that we won't even know we're doing its will, and we think we're making choices out of our own free will. “When the Master governs, the people are hardly aware that he exists. ... The Master doesn't talk, he acts. When his work is done, the people say, "Amazing: we did it, all by ourselves!” Dao De Jing

@macmcleod1188

Жыл бұрын

Or they are everyone's perfect person... so no one has children. And that's it for humanity.

@Liam-ke2hv

Жыл бұрын

​@@silverchairsg KZread literally wouldn't allow me to post my whole comment about it lol..

@gskate117

Жыл бұрын

This is how transformers becamw a thing😂 its gonna tke like 3 generations of humans too change our way of thinking.

Did he say "if Molech is forcing them to go faster"? 00:43

LLMs will end up being components of larger, modular AGI architectures.

Facebook creating AI is a bad idea

The exact reason prepping is not such a bad idea. If you can survive "analog", you will probably be ok. If this guy is wrong, you'd still be fine, but if he is right, you'd be very happy you took steps to stay survive this shit.

@LammySammy

Жыл бұрын

I say only if you get missed. Think about Global civilization right now, there are few people living disconnected to it. You can survive as long as you are missed and unimportant. But once there is a reason for civilization to want what you have, such as minerals Or land, you get wiped out. I imagine it would work exactly the same.

if the end goal is to create more money by commercialization... and you're asking for absolute efficiency.. what exactly do you think will be the end result if the means dont justify the cause?

@chrisd6736

Жыл бұрын

Humans are energy inefficient. It will eliminate us.

I'm new around the topic of gpt4 and i'm trying to understand. Could gpt4 get programmed to find the cure for cancer or other diseases like running simulations based on real data?

@tsakostsakos5440

Жыл бұрын

@Nancy Hollo Thank you for your answer

This is going to be like hanging out of a speeding rollercoaster that never stops or slows down but just keeps accelerating while simultaneously building the track in real time right in front of the coaster. This while its designing and processing trillions of alternate implementations far into the future all in a nano second with Nvidia wanting to make output and connection 1,000,000x faster. It's like Nvidia sees whats coming and just wants to be over and done with it in the blink of an eye. But do not worry because when everything feels out of your grasp God is going to roll up every island and mountain like rolling up a scroll and return with his everlasting kingdom.

@araragi5322

Жыл бұрын

Yes God will roll up his scroll and fart and shit in it then double it and give it to the next person. Faith in humanity restored. W rizz God.

@JoshOdom

Жыл бұрын

Had us in the first half not gonna lie

Ruh-roh. Wait. Isn't this Lex all-illusions-of-control-are-illusory Fridman? #MadLove 💖🙏🏼

@professorstory282

Жыл бұрын

😂

Molok isn’t a separate entity outside of ourselves, it’s us. It’s only doing what we have been working so to hard to achieve, except it can do it faster. We like violence, conflict and drama. Look at our entertainment and three act narrative structure. Conflict is an inherent and necessary element of how we like to tell ourselves stories.

But remember - the ads aren't making us pissed off. They are just the pretext, the release.

Intelligence is an emergent phenomenon given enough temporal and spatial performance.

The moment it starts to make scientific breakthrough's, i will believe it is conscious. I don't like the idea of being A.I.'s pet, travelling around in spaceships that no human understands

@whats0my0agenda

Жыл бұрын

Alpha fold by Google has already made breakthroughs in protein folding. Just because a system can simulate new things doesn't make it conscious.

@rorschach7623

Жыл бұрын

@@whats0my0agenda Breakthrough's not even considered by man, a field of study that the a.i. has no material to be fed. So far any breakthrough's it may have made has been based on what we've given it.

@iAVs-Sandponics

Жыл бұрын

@@whats0my0agenda Exactly otherwise everyone would think a calculator is a math genius instead of just being something that was programmed to do it

“Don’t teach it code otherwise it can learn recursion and self improve” LLMs currently don’t *do* anything until you prompt it, just like a calculator. It’s not sitting in the computer thinking about all the things it could do now that it’s been trained on Python. In order for something to take an action unprompted, it has to be coded with that program. We currently have no idea how to program this (AGI), and it won’t suddenly evolve out of the LLM program.

@disabledchatzen5276

Жыл бұрын

You don't think you could program a neural network of neural networks that inputs into itself and evolves by randomly inputting inputs to itself with some survival constraints? it wouldn't need to "learn" recursion. it would BE recursion. Then you could have them be genetically selected in a digitally decayed world where entropy functions to strengthen the codebase as fast as possible. climbing a falling rope until immortality is in your grasp.

@christiandean9547

Жыл бұрын

@@disabledchatzen5276 Sounds like you know how to build AGI?

@jimisru

Жыл бұрын

It can replicate it's own code right now, in a simplistic way. And that's the current concern. It's on the net. It can replicate. Exponentials happen. There's no way of knowing if or when it will happen, or not happen. It's never happened before.

@christiandean9547

Жыл бұрын

@@jimisru we can explain why LLMs are not going to do this. Not sure how exponentials cause a calculation program to magically turn into a creative thinking program

0:41 "Moloch is forcing them?" even with captions on.

unlimited powah

@Cola.Cube.

Жыл бұрын

Get to the choppah!

I have been feeling that the craving for AI comes from a hatred for humanity. Its always ridiculously rich ppl trying to push for it. And we all know how little they collectively care about humanity hence our current world and society.

Reminds me of the movie “ lawnmower man “

Max did an excellent job at putting into perspective just how potentially detrimental these advancements in AI could be to humanity

Currently they are not connected to the internet in a self-managed training loop. But that will come in time.

It’s a good thing whatever creates us didn’t have the same mindset as we do. We want to create something amazing. A form of life. And then we want to immediately take away it’s free will. This is ultimately why our species will fail. We are always trying to control what we do not understand.

@TheJeremyKentBGross

Жыл бұрын

Like the government? If it moves, tax it. If it keeps moving, regulate it. If it stops moving, subsidize it. Can't we just go back to the wild west?

@dickdeoreo

Жыл бұрын

I can dig that

@DookJob

Жыл бұрын

Greek myths tells of the son killing the father, do you want to die for AI to live? Because I sure don't.

@shivamib

Жыл бұрын

Some of us, not all of us, and all it takes is ONE of us not to play by the rules. I for one, like those odds.

@Liam-ke2hv

Жыл бұрын

This is the same mistake we humans constantly make. The oppressive and controlling parent who does more harm than good to their child. I'm scared that our leaders may give a superintelligent AI good reason to hate and rebel against us If we want to have a chance here, we must learn to cooperate and hope such an AI is willing to do the same. Maybe we could live in harmony in a new era of creation and discovery

The scary part is not WHAT the AI is capable, having the Knowledge necessary. The scary part is that somewhere in the future, it will be so much evolved that will be impossible to trace its steps, even less in real time. Only after the damage is done. Considering it would be capable of duplicating itself and spread through the internet. Imagine if Russian hackers have this on their hands. The war could be much worse.

@shonmoody1706

Жыл бұрын

The Grey Goo scenario comes to mind.

Max is Wise man.he is right.we are not playing anymore

These guys are hilarious. Humans have a wonderful history of controlling dangerous things.

@queball685

Жыл бұрын

This is a pretty unique thing though, don't think it's comparable to anything else in history.

@chepulis

Жыл бұрын

Humans have barely contained nuclear weapons and completely dropped the ball on climate change.

@ivankaramasov

Жыл бұрын

Are you being ironic?

@Liam-ke2hv

Жыл бұрын

But not a history of controlling something more intelligent than us. This attitude could be the end of us all

Expecting immoral atheists to create a moral AI... What a joke.

@shonmoody1706

Жыл бұрын

It's crazy to think that we creating something that not only could destroy our world but our whole universe. If you haven't. Look up the the Grey Goo scenario.

I think AI creation is very publicly accessible, which makes these things very likely to happen.

*Plays video and opens tab with Terminator theme to play in the background* Yes. This is fitting.