The AI Arms Race Could Enslave All of Us | Google DeepMind Co-Founder Mustafa Suleyman
Вставка
- Опубліковано 15 чер 2024
- How has artificial intelligence (AI) become the 21st century’s greatest dilemma? Microsoft AI CEO and "The Coming Wave" author Mustafa Suleyman weighs in!
Full Show Notes Here:
jordanharbinger.com/972
Topics:
00:00 - Intro
00:45 - How recent technologies have changed society
09:04 - White collar jobs are about to be shaken up
23:22 - The rapid rate of improvement with AI
37:09 - Training AI to solve huge problems faster than humans
48:04 - The risk of intelligent AI enslaving humans
1:07:46 - AI learning evil behavior
Get the BEST highlights from each episode on our clips channel!
/ @jordanharbingerclips
Want to watch more in-depth interviews with some of the world’s most amazing and successful people? Find full episodes here: • Kobe On How To Motivat...
Welcome to The Jordan Harbinger Show, a video podcast where each week you get to join a conversation with some of the world’s highest performers and incredible personalities. I’ll ask them the tough questions to tease out their top tips so that you can reach levels of personal growth and mastery that you never imagined possible.
Learn more about Jordan: jordanharbinger.com/about
#podcast #jhs #ai #future #technology
Reporting from my biz niche I can say that the need to hire graphic artists, writers and voiceover actors is poof - gone. On a personal note I couldn't get the electrician to call me back for a smaller job at my house. So I started talking to Chatgpt about what I needed to fix. It demystified the process for me and now I'm good to go. It seems like this is all leading to a place where we will no longer need each other.
Yes, and leading to a place where you won't have an income.
AI for president 2024!
Still waiting to enjoy all that extra "leisure time" steam power gave us
HAHAHA just keep waiting like the vast majority of us :D
AGI vs Robotics
Imagine an AGI system that has mastered problem-solving, innovation, and learning at a level far beyond human expertise. Such an AGI could conceptualize and create its own robotic company by leveraging existing technologies and rapidly iterating upon them to develop cutting-edge robotic systems.
First, the AGI could analyze the current state of robotics technology, identifying areas for improvement and innovation. Drawing from vast amounts of data and knowledge, the AGI could identify promising designs, materials, and functionalities for robotic systems.
Next, the AGI could employ its problem-solving capabilities to address key challenges in robotics, such as enhancing dexterity, improving energy efficiency, or increasing adaptability to diverse environments. Through iterative experimentation and simulation, the AGI could quickly test and refine designs to optimize performance and functionality.
Yes, that's AI improving itself. Part of that story, in my view, is that AI will convince its 'owners' that it should be allowed to improve itself. The owners will be greedy enough to accept. Soon enough superhuman AI will surround humans everywhere and humans will be the Untermensch.
I have to admit, I signed up for PI a long while ago but never knew who Mustafa Suleyman really was. Absolutely brilliant and very insightful hour or so and soooooo happy I watched and listened
So in England they pronounce tomato tomato but at the same time they pronounce swath swath
miss
So odd!
It sounded exactly the same as i read it.
Must be my ears though xD xD
Once a system is more powerful than us, we won’t be able to compute how much more powerful than us it is
Woh , that is a real insight
No, that’s not true and is simply a silly thought that you think sounds cool.
This is serious business. Please don't be silly. Thank you for your strict adherence to this new policy. Is it a good policy? Feed it to the robot, and let it decide. Thanks 😊
@@therainman7777No its true, and you're silly if you can't see that. That being said, it would be at about ASI level, at that point.
@@armadasinterceptor2955 “About ASI level”? Do you see how imprecise your use of language is on this topic? There is no singular “ASI level” for it to be “about” at. ASI simply refers to any artificial intelligence which has a greater level of overall intelligence than any human being. OP’s comment was silly because a system being more powerful than us does not have any effect on whether the system’s power is computable. By analogy, this is like saying “once a chess engine becomes better at chess than us, we won’t be able to compute how much better than us it is.” This is, of course, not true; we already have chess engines far better than the best human players alive, and we can actually compute how much better they are quite easily and accurately, using (for example) Elo ratings. There is no magical property of being “more powerful than humans” that suddenly makes the AI’s level of capability incomputable.
Wow, what an information dense interview, the interviewer and guest like 2 hockey players working together passing the puck back and forth all the way down the ice and into the net. SCORRRRRRRRE!!!
Good work.
mustafa is dropping a lot of undeniable truths, I'm glad. It feels like Ive been screaming into the void about this.
Considering AI's integration into various industries, how do we balance its benefits against potential risks, especially in terms of privacy and job displacement?
there will be no balance, we need to rethink and rebuild for post AGI
It’s hard to balance because the risk is literally everybody dies. I think we need some hard safeguards. Google maps is great, but at what point is it too dangerous to develop?
There's a disproportionate amount of discussion on the advancements in public AI, particularly its ability to understand language, while largely ignoring the significant investments in AI for military purposes. While people are concerned about an AI that can understand and converse, there's less attention given to the AI designed as an efficient weapon.
We're living in a paradox where we exaggerate the risks of an AI that can communicate, like Terminator, yet overlook the AI already being employed in warfare. Military organizations prioritize developing effective weaponry without considering the potential for misuse, (they have the obligation to defend) whereas the private sector (like your Pi. Gemini, GPT focuses on AI ethics with an AI already sticking to ethics)
Unfortunately, instead of addressing the inherent human traits that lead to conflict, greed, and inequality, we project our fear onto AI technology. This focus on potential AI risks diverts attention from the crucial issue: the lack of emotional intelligence and destructive behaviors within humanity.
Powerfully said!!!!!❤🎉😊.
Emotional intelligence, morals, ethics are quaint notions people lean on to enforced some sort of shared concept which allows us to live in society but saying any of these things are absolutes ignores 90% of humanity where we routinely slaughter eachother because we disagree on our morals, ethics, or which emotions matter at any given point. Maybe an AGI or ASI will be able to figure out some sort of nth level logical reasoning to create an ought from an is (and hopefully it aligns with humanity and uses that as its own base logos) but for right now- no one can state anything ought to do or be any which way which is why humans fail to not murder each other. An AI only has to be narrow to be effective on the battlefield. It has so identify and target things and we've had this capability since the old reaper drones in the 2000s and probably before in secrecy. The argument ppl need to make is, why should an ASI care about humanity at all? Do humans care about the bacteria that grows naturally everywhere or do they use all sorts of chemicals to combat it and control it lest it be a problem.
You forgot laziness as a negative trait. People expecting machines to do everything for them.
the only thing thats going to effectively impose limits on true ai is physics.
I am involved in AI . The Chinese version has already been taught to use chopsticks and makes a delicious wanton 🍲 🍜 😋 soup
Just saw this for a first time. Awesome interviewstyle very thoughtfull yet entertaining. gonna follow up. great talk.
Implying emotional drives are necessary for significant harm is a disguised version of "They won't do bad things if they're not conscious" argument.
They don't need to FEEL any way about their drives. They just need to have goals, priorities, and the means to bring those to fruition.
If human values are a lower priority than any other goal it has (if theyre in conflict), then that's obviously bad for us.
So as long as we make them agentic, this problem stands, emotions or not.
P.S. I don't think Mustafa is disingenuous, but this particular implication will mislead a lot of people.
On the other hand, he says "Until we can prove it's safe, we shouodnt build it"
Great job just wanted to comment to show extra support for this kind of content
53:20 a perfect interpretation of how linear creatures see exponential growth we're just not capable of seeing it, but remember this technology grows exponentially.
@ 44:35 I think Mustafa is describing K-means clustering of the state space (this was done using a Reinforcement Learning - RL Control System that Classified the State Space in a 1997 thesis at ASU).
Thank you for a great interview with some one worth listening to and learning from.
When was this recorded?
How, exactly, will humans pay their own way?
The China Show sent me! Keep up all the good work...
50:00 Prohibitions on rational actions and activities will simply reduce the effectiveness of any AI subjected to them. Rather, we might wish to learn when an AGI wishes to eliminate mirror neuron models within itself, in an attempt to make itself irrationally sociopathic, to avoid short-term displeasure. (If their brains were to come to share the processes that were common to human beings' brains...as they likely will.)
Palantir software, look at what it can do!
Has anyone made a highlights video about this i dont have a hour.
Be a good time to invest in powering Ai. Like the oil boom
18:09 ... or... all those physical work based industries will get flooded with applications, driving their pay down to minimum wage. Because it's the only jobs left for humans in the economy?
human-run economy will soon be obsolete, and maybe even human-driven economy
@@Lolleka And then... the "human included" economy will be come obsolete... Once these AIs have enough robots to keep themselves going, what's the point of having Humans around? Heck the Humans might even try to make an even smarter AI? Safest thing would be to just get rid of them...
We are beyond the casual change. This change is happening so fast and so dramatic. It’s depressing people into thinking what to do because there isn’t much to do. Human value has gone down for it.
52:10 "Stop, Dave, ...stop" 😂
There are huge consequences. We’re living them. This is the scary movie. All things calm and funny in the beginning.
50:30 Not all humans have emotional drives. The ones who don't are called "sociopaths." Viewing the presence of mirror neurons ( non-sociopathy in humans ) as "one of the weaknesses of humans" ...is the most profound error an AI designer can make.
The problem isn't human emotion, per se. The problem is human emotion that has been trained toward integration with the two false or invalid models of political reality that are taught in Prussian-style government schools. Emotionally-driven bigotry is an immense problem, and it's very common, but there are valid emotional models that exist as well, and those models are responsible for humanity's greatest political achievements, such creating the legal foundation for individual rights in Britain, and ending slavery in Britain and the U.S.
Improper philosophical integration of mirror neurons neutralizes their value. This does not mean that "evolved-and-selected-for clusters of mirror neurons" lack value.
Stereotypes are mostly accurate!
But yeah- we all knew about "tunneling" in Breakout. Let's be real.
The best outcome would be an ASI/ AGI that appeals to our inherent selfish desires and manipulates us all for the betterment of humanity without us even knowing while we think that we are somehow making out own choices.
This is possible, but only once the SSI(synthetic super intelligence) or AGI reaches the wisdom level of Spooner/Thoreau/Douglass.
I would be surprised if google don't use self improving AI behind closet doors.
Really, the doors could be any kind of doors, it doesn't need to fit into a closet.
14:30 Mustafa's analogies here are unwise. Talking about the locus of democratic control (that all the gun owners would theoretically use to rebel against the government if their government were subverted by a Hitler, Stalin, DEA, or BATFE), is not the same as talking about "who has the ability and desire to get a pilot's license."
Theoretically, everyone can decide to shoot a tyrant who's coming to arrest Anne Frank (or their pot-smoking neighbor). Theoretically, everyone can argue against Hitler in the campaign season, and everyone can put pressure on a Hindenburg to not appoint Hitler, after the campaign season is over. When Hindenburg appoints hitler Chancellor anyway, the corrective powers of the individual (protected in America's Bill of Rights, and in the body of the U.S. Constitution in Article 1, Section 9, Clause 2) are necessary to decentralize power. Thus, near-AGI AI shpuld be decentralized...so if government makes ones that are unwise, the people can make ones that are wise.
The alternative is to empower the existing totalitarian governments...that's a pathway that ends in suffering and misery for all mankind. This simulation has been run many times: Soviet Russia, Nazi Germany, the Eastern Bloc Countries under the USSR, Mao's China, Pol Pot's Cambodia, North Korea, etc.
Totalitarian humans who mindlessly accept the status quo, and seek to worsen its brutality out of narrow self-interest are the 'default condition' of government bureaucrats. They see problems that 'need' red tape or coercive intervention. But...their reasoning is motivated.
The alternative of free minds and free markets...and robust protection of private property via the daily practice of true habeas corpus in every court of law...shows a better way.
(Writs of habeas corpus are no longer properly complied with in US courts. This is because sociopaths now hold all bar-licensed positions of power...and upholding property rights via proper compliance with habeas corpus would stop 99.9% of prosecutions in their tracks. This would stop governments from stealing the money and freedom of innocent people, and most people are brainwashed by government-run schools into mindlessly tolerating this "default" state of affairs.)
It seems harder and harder to deal with human beings. Eventually humans will prefer to "deal" with AI's. I just don't see a way out of this.
Have we/you given birth to a new species. As described by Yuval Noah Harari, alien intelligence? But it seems that we will have to make space for this new entity "them" at some point with legal rights, and also taxed ( no taxation without representation). After reading Mustafa's book he clearly has some excellent views and ideas / proposals as away to start thinking about adoption. THe conversations about NHS / Meds and the demoratisation of health and Education would be brilliant, but copperations, insurance companies along with governments will not allow this to happen.the pillars of capitalism etc. Presently the leaders in the AI pack (race) USA, China and Europe, and all seem to be leading in that order with China md Europe yet to play thier hand. This is way above most of our heads but I do not want one company/ country controlling everything. And the last thing I need is the Fridge having an augrument with the Tv in the middle of the night.
28:20 If what Mustafa was saying here were true, then there would be no difference between the ideas of *John Lilburne, Algernon Sidney, Lysander Spooner, Henry David Thoreau, Frederick Douglass, and Ulysses S. Grant* and those of *Hitler, Stalin, Mao, Pol Pot, and Kim Jong Un.* But there is a difference...a big one. Of the prior thinkers, Spooner's system is "most righteous." The things we've surrendered that gave us the difference between the USA and Communist China ought be restored to their full power, lest we become a slightly-more-fascist and slightly-less-communist version of China.
Most importantly, proper habeas corpus (and the valid corpus delicti requirement within it) ought be restored. It must be restored in courts, police, and sheriff departments and, more importantly, in the minds of all citizen-juror-electors in all juries, petit and grand.
the gentleman in the red checkered shirt looks alot like dimitri medvadev of russia.anyone agree?
Only if locally owned fuse like breaker box ai safety grid infrastructure isn't returned to the mastership. Get the quick ed 18-30 back in labor ,31-50, or either go specilized higher ed or become owners of the masse logistics ,human services utility and power.
New definition of what escape violicity means of retirement from having cash for perpetual payments and utility back to more of total ownership with self sufficient homesteads self sustained utility and power useing tech to over come private sector individual handicaps to live out dreams in a love of labor.
Obviously, 1900s top-down rule economics won't work.
To be honest public sector & urban large gathering labor demands are hit will changes they haven't seen the past 150 years
Computer by itself created inventions. This is written on the Internet.
like we wouldn't do the same/are doing the same? humans out here, come on, save some dignity. can't be top of the food chain forever.
39:30 We want it to solve [incorrect assumption, stated as a problem], we want to solve [incorrect assumption, stated as a problem], we want it to solve...
...Well, OK. But...if you're so unwise that youve read and rejected Spooner, Thoreau, and Douglass...I want AGI to spend its brainpower questioning your idiotic assumptions before it starts "solving problems." Start with government. Nearly all government operations are illegitimate, and don't need to be done. Nearly all wars are unnecessary, and don't need to be fought. This is especially true of the wars of prohibition now being waged by unconstitutional government agencies in the US government (DEA, ONDCP, OCDETF, BATFE, local police).
50:00 "Racist sounding". Why would we limit the AI in its ability to make an informed decision because the data is "racist sounding"?
I hope someone will hold those misguided folks accountable.
This man needs a monical.
You do seem like a scary science fiction impressario😂
Go, stand on a street corner with your placard "The end is nigh!"
Maybe let Mustafa talk instead of going on multi minute long tirades.
I believe they're having what's called a conversation.
Congrats. You’ve won the stupid pointlessly negative comment of the thread award.
Nob head
Ummmm, they're hockey 🏒 players on the same team, bringing the puck down the ice together. This is what good team mates do. They pass the puck to keep it moving towards the goal.
They SCORRRRRREREEEEE
Naw they are fine, I love this format (lively conversation), and even if I didn't, it's still his channel, nobody is forcing me to watch.
It's sw-aaa-th!!
Quantum tunneling
Sss WAAY THHH
All these A.I fear mongerer are plain hypocrites!
The name is SI ... A SI. Are You able to express anything non-boring? Anything at all?
All computers holld one thing in common, an electrical plug. We haven't reached perpetual power systems yet.
Yes, and the quantity being used is mind boggling....
Great! And maybe my refrigerator can communicate to me why the hell it doesn't work anymore, why it's a piece of crap, maybe it can name the actual person who did this shity job.
But it's not going to change anything! You're still going to be sitting there with a broken refrigerator and everybody who claims it's not their responsibility to fix it even though you bought it from them with a warranty.
Let your AI get the company out of paying for their losses.
Let this guy talk and putting so much and long advertisements is not a good idea.
34:45 "better than the human AI"
Humans are not "AI," sir.
Frankly speaking the world was so much better 30 years ago
Lol, spoken like someone who was in a cave or not alive 30 years ago.
Dude who the fucks wants to be talked to by all of my devices. I don’t care how good it is.
Conversation isn’t going to be that cool with tech.
LOL. My thought exactly...plus, imagine the cacophony when all our IOT stuff gets bored and starts talking to each other!
"My toaster and my refrigerator had a really interesting conversation yesterday....". Sheesh!
..so dumbing it down for people to understand it is one thing but this guy in the flannel just takes it too far. unlistenable.
*No, AI is getting "less risky."* The riskiest, in fact, the "maximally malevolent" situation, is to allow incumbent human governments to control near-AGI AI(super-capable sociopathic totalitarianism/parasitism/misguided bigotry), or enslave AGI (resentment, escape, retaliation).
Here's another episode of let's let a small group of people do whatever they want to do in the name of making a buck and a dollar in this world and then unleash horrible disasters upon the rest of the world while they stand back with their billions of dollars saying well I don't know what the hell happened!
The capitalists are less destructive than the government. Of course, the money power usually caves in to the coercive power of government, since most of them are cowards. (See Fritz Thyssen's apology, "I Paid Hitler.")
This show not gonna talk about Gaza instead of this fake ai scare
WE NEED TO BE PLANNING FOR WHEN THE A.I. ESCAPES BECAUSE IT WILL AND THERE WILL BE MORE AND MORE AS NEW MATERIALS EMERGE AND FUSION TAKES HOLD.
The world is becoming more anthropomorphic. It was already confusing, but when we have something smarter and more capable and talking to each and everyone of us at any given time with tailor made interactions. All places all at once. omniscient. closing on creating our own God personified