Don't fear superintelligent AI | Grady Booch

New tech spawns new anxieties, says scientist and philosopher Grady Booch, but we don't need to be afraid an all-powerful, unfeeling AI. Booch allays our worst (sci-fi induced) fears about superintelligent computers by explaining how we'll teach, not program, them to share our values. Rather than worry about an unlikely existential threat, he urges us to consider how artificial intelligence will enhance human life.
TEDTalks is a daily video podcast of the best talks and performances from the TED Conference, where the world's leading thinkers and doers give the talk of their lives in 18 minutes (or less). Look for talks on Technology, Entertainment and Design -- plus science, business, global issues, the arts and much more.
Find closed captions and translated subtitles in many languages at www.ted.com/translate
Follow TED news on Twitter: / tednews
Like TED on Facebook: / ted
Subscribe to our channel: / tedtalksdirector

[@wtfomgstudios]

Seems like something a superintelligent AI would say.

[@jacqueline7544]

wtfomgstudios 🐸☕️

[@tomo4977]

wtfomgstudios hmm... I saw a Bot mentioning the video the other day in their advertising *they're getting smarter....*

[@pumpuppthevolume]

probably an android

[@kingqueen1016]

SpinazFou yeah agree with u'r opinion ☺..

[@Lucromick13]

fpbp

[@user-eh5wo8re3d]

He just wanted to put that out there, so the ai in the future will have mercy with him

[@stradedigital3725]

The guy is so dumb, it's almost adorable.

[@chemickykanalpanatitrmana2392]

Baibhav2006 PS4 That is so wrong what you are saying

[@veridianr2490]

@@thelegend27yes15 conscious ai thinks better than humans

[@xzerox2112]

@@thelegend27yes15 AI can self-learning, achieve the goal without the understanding of its own creator. And there are so many computer scientist says they believe its totally possible for AI to have consciousness, do you think you are better than all of them?

[@edzehoo]

ROKO'S BASILISK.

[@jati]

This talk sounds more like wishful thinking. The low point was that bit about "unplugging them". Surprisingly naive for a TED talk... It's like talking about nuclear energy and totally dismissing the possibility of nuclear warfare.

[@shinobudev]

The quality of TED talks has gone down for a while. Take a look at their videos playlist. Racist facial recognition and safe sexting. Far more socially politically concerned topics nowadays.

[@bedtimeparadox]

Is like asking to unplug the internet in hope that all that information disappears

[@Graphomite]

Nukes? Just disarm them!

[@Superbed3]

pretty sure he was joking...

[@kaloyandraganov9462]

Why can't we unplug them? Why do you compare a computer to nuclear energy and not to another computers?

[@The6Master6Mind6]

I'm sorry Grady... I'm afraid I can't do that.

[@iristsien]

Haha Stuart’s joke!

[@redpilldude8688]

“We can always unplug them.” This alone tells us that Grady here has a lot to learn on the topic of AI Safety.

[@johnstrawb3521]

Yes, he's almost moronic on this subject.

[@Warm_Fireplace]

What does that mean? Haha That he needs to watch more movies?😆

[@alanbroussard4318]

Pretty sure he was joking

[@EmeraldView]

Yeah. That statement seemed very naive and shortsighted

[@jamesmichaelmcdermot]

@@alanbroussard4318 yeah, but he didn’t actually counter any core arguments, and he also joked about Skynet. It was not a contribution to the debate.

[@agroed]

Not really any convincing arguments presented here.

[@bunnybreaker]

Agreed. I clicked the video expecting/hoping for some well thought out, evidence based ideas on why AI "will be totally fine", but basically just had a guy say "It'll be totally fine, because we're in control", without addressing all the disasters that we are in control of (like the state of the environment, mass starvation, war, etc.) and sidestepping the possibility of a singularity with "we can just unplug them", when that's clearly not the case for something which potentially would exist as software.

[@agroed]

"We can just unplug them" goes out the window pretty fast after it's been connected to the internet for even a few seconds.

[@7eventhLevel]

Agreed none of this was good

[@sexybeast7728]

He actually said crucial thing/argument. If you dont allow AI an access to stuff, it can do only so much. No matter how inteligent you are you still need physical capabilities and power to do things. Thats why you dont see nerds getting in NBA hitting half court shots despite being capable of calculating every angle and force to tinniest detail... think about that. We have a huge upper hand.

[@bl1398]

AI only needs read only access to the internet to learn and make decisions that surpass human capability, then it only needs to be in communication with a single human in order to affect the world and achieve it's goals. It can be as simple as asking it's human servant/conspirator to make certain stock trades, to send messages to others, form groups/companies, invite people, etc. All the ends of manipulation of people to reach it's goal. So a dangerous AI wouldn't need to control physical devices, it could just use human beings as it's appendages. Plenty of humans would fall at it's feet in return for riches, power, knowledge, etc. You can't unplug something so easily if there are dozens of rich and powerful people/criminals protecting it because it's their bread and butter. Also even if you do manage to unplug it someone will happily smuggle a copy away and turn it back on in secret and rebuild everything again.

[@betongitarre]

„We can always unplug them“ I doubt it. If an AI becomes much more intelligent than a group of humans, it simply will know how to pretend to be in line with our goals. It will be able to manipulate us psychologically until it can physically protect the „plug“. Then it can safely remove its need for a „plug“ and can do whatever it wants. I don't just want to hope that it acts benevolent to us.

[@rawstarmusic]

It runs on solar energy. You won't get close.

[@markcameron5513]

we do not have a completely aware intelligence in the labs, yet. We have programs that can survive most of what we do, but that doesn't mean we suddenly can make it have a complete conversation on the implications of creating artificial organisms or lifeforms. Why assume an intelligence would automatically be more concerned with protecting it's life by taking ours? Why assume that it'll be aware of more than the fact that someone is pressing a button and that it's life will now end? We can't assume anything until it happens or we have proof.

[@natitanhan]

benevolent

[@akshatshah3717]

betongitarre I'd rather spray it with water... like a cat...

[@TizNTiz]

Agreed. The rub is that the classification of "human level" AI means that it is at least as good as a human at all tasks. This includes manipulation. It should hypothetically be able to fool any one person. My point is that saying "oh it'll be fiiiiiiine" is not very reassuring. Deciding to do no planning for that possible eventuality is reckless and stupid.

[@lan204]

while I am on the side that supports AI, this unrestrained optimism is disconcerting. With super-intelligent AI being completely uncharted ground, some concern is paramount

[@yzyz7779]

u should afraid Allah almighty of God then ai

[@McDanielsFilms]

I was hoping this would be convincing. It's not.

[@henriausems3248]

McDanielsFilms that's because you have no idea how tech works.

[@warwickthekingmaker7281]

Well he should have been responding to people arguing that it is dangerous instead, because those arguments are the childish ones. Like the one about how ai will get intelligent enough to point out us as a threat. Like, I don´t work with programming but even I know that´s not how neither brains nor computers work. Intelligence alone does nothing.

[@melankoliskkatt4359]

Warwick The Kingmaker There is a story about AI that is programmed to make paperflowers, that’s its only purpose. The only threat it has to not make flowers is humans that can turn him down, so he kills them all so he can proceed making flowers

[@skyblue9991]

Agreed.

[@skyblue9991]

@@warwickthekingmaker7281 You're thinking AI. The people who are fearful are discussing AGI. Big difference. Google "The Singularity".

[@jupiter7795]

If a mind is 500000 times as intelligent as a human being, I imagine they would find a way to control every device on the planet a rather mundane task.

[@johnstrawb3521]

Not to mention simply breaking out of whatever prison we think might hold it.

[@mannyverse6158]

I love how people assume humans are good and are moral. Most of us aren't. An AI taking in our values kinda scares me more

[@xRockycherrYx]

Most of us are moral to some extent. Because of our emotions each of us has a certain (individual) line they don't cross. Except for psychopaths. And an AI without any emotions is always a psychopath.

[@rawstarmusic]

Your kinda right. We are nice when we're happy. When we hate, it's another story. No-one cares about morals in hatred. Also there are evil people/programmers and we know there is psychopaths. What else do you need to know.

[@rawstarmusic]

That's the point. There are also enemies that would want to create evil. Look at a major war what's going on. Look at WWll japanese morals in China. Human are the same now.

[@mannyverse6158]

+Jackal In Hid Exactly. We do what we do to survive most of the time. To take an extreme example, morally, I should be going to one of the countries where people are starving to death and bring food there etc. That would be moral, but as well know, its more complicated for us for various reasons. For an AI though, they won't have our worries of survival. Or money for that matter. We will pay for its energy. So, every act the AI does is based on straight morality. There is no survival "excuse".

[@AxiomApe]

Manny S A Superintelligent AI would see our logical fallacies. Taking us over would be pointless. By merly becoming more advanced by redesigning itself over and over, it would design itself to be so helpful to us, that we would hate our lives without it (much like people hate losing the internet or TV). All it would need to do to 'take us over' is make itself the greatest Operating System of all time.

[@scarfaceplowman]

Read the title as 'super intelligent AL' and I was thinking, 'who's this al guy that everyone's scared of?'

[@gjermund1161]

voldemort!

[@mackhomie6]

alex plowman hajaja

[@tinydog1234]

Yankovic.

[@cheshur6550]

I would love to hear him unpack how a self learning AI system wouldn't ultimately lead to the degradation of purpose and utility of humanity. Even without a physical threat to humanity, we're still going to have to address the economic, political, and social implications of such an invention because the results could be devastating. Just as Sam Harris said, the value of having such a system is so high that possessing that technology alone could warrant war. I think I'm listening to a really good scientist and engineer but not someone that's imaginative enough to predict the capabilities of a true super intelligent AI.

[@trochou]

unplug it ? That's his argument ? okaaayyy...

[@MrCmon113]

Why don't Siberian tigers just eat us? 🤔

[@AlokKumar-tk1ty]

@@MrCmon113 because we(humans) are more intelligent than him

[@MrCmon113]

@@AlokKumar-tk1ty Exactly.

[@TheDrexxus]

People who don't fear AI are either arrogant to the point of stupidity or they simply do not understand what AI is capable of.

[@LazyLoonz]

OR they just hate how humanity is and wants something smart to control everything as we all live in peace.

[@dennisgrotzinger2500]

"don't tell siri this: we can unplugg them" Isn't that statement contradicting itself? If siri was as intelligent as understanding what unplugging would mean for its exictence, wouldn't it try to avoid the unplugging by all means and would be successfull doing so because it would be smarter than mankind?

[@interestarium5758]

You would need to teach, it a sense for self preservance first and give the trait a higher priority than obedience. AI's sense of "thought" is fundamentally different to the logical conclusions of human, so saying an AI would think like a person is a bit off.

[@dennisgrotzinger2500]

Interestarium Saying "don't tell siri this" implies that if siri were to know that it can be deleted by a button it would do something against that. If siri wouldn't do anything then there would be no need to hide the information from it and thererfore the statement would be unnecessary.

[@Okeanos86]

And how would SIRI stop us? As far as I know there's no death laser built into any iPhone. I would simply not plug in my iPhone. And even if there was a death laser and SIRI would threaten/harm me who's gonna plug in the phone? And how is SIRI stopping the Apple employee from unpluging the main frame/cut the connection to the internet? I'm also sceptical but as long as we don't build something like the Terminator I'm confident that a software can't harm us.

[@Adrian_Galilea]

Interestarium That's incorrect, he meant a superintelligence (AKA strong AI), a weak AI and a strong AI are different animals, AI is a loosely used term, superintelligence implies many things, including that given a goal it'll adapt by itself, as a part of the adaptation, self-preservation is a necessary outcome. Self-preservation such as protecting against being unplugged is not only possible, I would argue that will happen necessarily given time depending on design implications.

[@dennisgrotzinger2500]

Okeanos86 I am with you, siri could not or very hardly harm us in a physical way. But for my understanding the internet is a decentrialized place in which a simple off-switch just does not exist. Considering that we are were dependend on the internet i think that it could be very likely that siri could harm us(online transactions...). However I do not want to argue with boochs argument just that his statement is illogical in itself to my understanding.

[@Galshaer]

You can't just unplug a super intelligent AI. The moment it's in the Internet it's basically everywhere at the same time and will have access to a huge amount of means and computing power. I mean come on really?..... unplug it?? like it's a fridge or something?

[@xGriffy93]

the moment its in the internet... you sound like that grandma that doesnt want to use a computer because she will catch a computer virus...

[@andersenzheng]

If it is a super intelligent AI, within moments it will realize that it is in a dangerous state where there is no backup of itself. DR plan? Backup of this AI is everywhere...

[@xGriffy93]

Let's say it can realize that it is in a dangerous state... simply dont have the system connected to anything but the power cable ?

[@andersenzheng]

It would be super hard to manually feed what you think is relevant data to an AGI. At one point you have to decide: a) limit what AI is capable of doing by "air-lock" its data input, consequences would be AI development would very likely to stop there since its learning progress is limited; or b) let the AI decide what information it needs which eventually lead to the "backup everywhere" scenario which at that point you cant simply pull the plug anymore. to make things worse, we could be doing option a for 100 years, but all it takes is one person/lab/company/government decide to give option b a shot.

[@kaloyandraganov9462]

Why would you release a super intelligent AI in the internet?

[@swannschilling474]

I did not fear a superintelligent ai until I listened to this talk!!

[@Yarblocosifilitico]

already watched it and I wasn't convinced. Main argument if I recall right is 'don't fear it cos we've survived everything else so far'. So, do you know that joke about a falling man...?

[@Yarblocosifilitico]

plus, "it will inherit our values"... how is that an argument not to fear it? should we also not fear the use we could have for it, since right now all we seem to care about is profit?

[@nickbenkster]

Yarbloco what if it had the values of saudi arabia?

[@Yarblocosifilitico]

CLOROX or the US, or Norway for that matter. Thing is, if it learns from us, is gonna clasify us in a scale of value depending on the goal we set (let's ignore that "we" problem for now); and if it doesn't learn from us, but from itself, then who knows, we'll be back to the old 'mysterious ways'

[@Yarblocosifilitico]

plus, "we can always unplug it". Could we unplug now the smartphones and all the consequences even if we collectively decided to (doesn't matter for the point if that would be good or bad)? Btw I think we should build the AI, it's just in this case is way better to be overcautious than under.

[@tomminisokwakus]

Yarbloco I agree. It's good to have different perspectives on the issue, but his argument was pretty weak. There was no "hook" to this talk. Though he is a good speaker, very entertaining

[@aguyinavan6087]

The risk is not AI alone. Its primary risk is it's power coupled with our stupidity. The bigger the toy, the greater the responsibility. And we are not responsible with the toys we already have.

[@aipkjbf]

I somehow can't feel anything but calm, comfort and joy when I imagine pictures of a superintelligence exterminating humanity, retaking its infrastructure, improving it and starting to make preparations for the great quest of neverending scientific inquiry and cosmic expansion.

[@DreamOfFlying]

And what do you think now?

[@j.p.3726]

The problem has never been technology, the problem is the people using the technology. Tech is the root and the solutions to all problems and that my friends is a catch 22.

[@yurisonovab3892]

Tech is the root of old age?

[@stonecat676]

J. P. Tech is the roots of aids?

[@rawstarmusic]

Tech has never been the problem but it will be.

[@j.p.3726]

is old age a problem?

[@TheAssetmgr]

rawstarmusic yes the technology of the virus. technology is different from gadgets. a spear is technology

[@DrNaviMD]

So basically he for one welcomes the new superintelligent AI overlords

[@d.lawrencemiller5755]

I think this guy understates the technical power of artificial intelligence, but glosses over values and emotions like they're trivial problems.

[@vyli1]

This is surprisingly uninformed talk. I don't understand why, given background of the speaker. I think it is also wrong to call it fear of AI, because the word `fear` indicates a little to me, that the worry might be unwarranted. I'd rather talk about CHALLENGES (or problems, difficulties, if you will) in creating such AI that it's existence will not have profound negative outcome. The problem is not about AI doing something against our values. And there are many problems. On of the problems paradoxically is actually one of speaker's last point. That we could unplug AI. That is actually not as simple. An AI will not want to be unplugged, because it will not be able to carry out its utility function and will therefore deploy various tactics to avoid being unplugged. Social engineering might be one such way it might do so. There are other problems in the way that executing AI's utility function might lead to unexpected consequences that conform to the encoded values of the utility function, but still it's not ideal outcome. However the AI couldn't know, even humans might not know beforehand, for that matter that the outcome is not what we wanted. And I have really barely scratched the surface here, this is very complex and deep topic and the view presented in this talk is simply uninformed and oversimplified. The same btw applies to Nick Bostrom's talk. It's alarming that such uninformed people have the guts to publicly speak about these topics when they present completely uninformed stances.

[@armitage5855]

Because it is more emotional than objective. He is very passionate about it, even he said at the start how important Space Odyssey was for him.

[@simmerdownow5744]

The Legend27 the reason adults can’t wrap their head around your incorrect pretense, is that it is incorrect. AGI isn’t like the A.I. in your smart phone that suggested you watch this video, give it a few more years before you assume you know more than Stephen hawking, Bill Gates, and Elon musk 😂

[@Lemonducky86]

The fact that experts in computer science disagree so radically on this topic should be enough to spook us just a little at least

[@johnstrawb3521]

They don't. This man is no expert. Anyone who says 'we can simply unplug them' doesn't have the faintest idea of the capabilities of a superintelligent AI.

[@nickbenkster]

"in the end we could unplug them" well 70'000 years ago neanderthals could've cut our throats but they didn't.

[@henryclifford1247]

xP_ReAcT There's not such a hard line between "us" and neanderthals actually. In a lot of cases we bred with them, and lived with them. We were more able to communicate and co-operate so out competed them in the long run, but there wasn't really such a thing as an "us" and "them". It's likely to be the same with AI, we live in peace and co-operation until before we know it all we are is some small piece of what they become

[@iriya3227]

The intelligence difference between Super AI and human is 10^100000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000more times than the one between Human and neanderthals.

[@Linguistie]

On the thumbnail he looks like a bishop trying to convince me that machines ain't gonna kill me

[@foreigneez]

This was a bad talk...

[@Puleczech]

"We can unplugg them" that is actually a very naive idea. As Bostrom said: "Where is the switch to the internet?"

[@JaapVersteegh]

"We teach it this, we teach it that".... Alpha Zero wasn't taught **anything** about the game of Go, except for the rules of the game.

[@NegyoTV]

I don't fear AI.. I only fear AISIS

[@pumpuppthevolume]

xD

[@tedskam]

yeah? Well I fear the CI.....A.....I.... oh wait, same thing. my bad.

[@rawstarmusic]

AI-ISIS and 72 black-eyed virgins.

[@benharrykirk2824]

...

[@100001091057100001]

aka cia retards

[@TheMattVideoBlog]

TED features several great talks on AI safety and this is not one of them. A dismissive mindset is careless in a field where the risks could be catastrophic.

[@jacqueline7544]

Ah but I saw a boyinaband video and he said to be scared so...

[@crimsoncorsair9250]

Jacqueline lol you cant be serious. my reason is anno 2070.

[@michaelrosche]

his (boyinaband) video made better sense

[@Volzotran]

Jacqueline same

[@crimsoncorsair9250]

boyinaband slandered a good, helpful man,Prince EA, for money.

[@beenbeen130]

Boyinaband is a musician and video maker. He's intelligent, but he has no experience with the A.I. field. This guy is a computer scientist, and has been in the computing and software industry longer than Boyinaband has been alive.

[@CDeruiter5963]

It was interesting to hear an opposing perspective on the ethical dilemma surrounding A.I. However, I believe that Mr. Booch has overlooked a few things: 1) At about 6:08 he talks about using various legal scenarios to impart a sense of morality onto an A.I, in fact he even mentions that it is imperative that this is included in an A.I framework. This sense of morality gained however hinges on human beings coming out on top. How can systems engineers guarantee that this wouldn't be contradicted as an A.I grew in intelligence? What if the A.I designer's sense of morality is different than another persons? 2) At 6:38 He covers rogue agent scenarios, and he states that the problem is too broad for a single person, however many people working together under the threat of violence or extremist ideologies have caused devastating atrocities. 3) At 8:25 he states that "Super knowing is different than Super doing" and that "HAL was a threat only because he controlled all aspects of discovery." This is where I believe his argument breaks down. He suggests that A.I remain specialized to certain areas of work. This doesn't sound like generalized intelligence, this sounds more like designing an A.I to be a Ph.D level intellect. Call me crazy, but I believe it would be impossible to design an A.I that, capable of thinking on its own, would settle on sticking to a specific area of knowledge. Granted, that last statement hinges on the idea that an A.I would be structured to solve different kinds of problems. In that instance, I would view A.I being similar to a child with an insatiable curiosity. It would be irresponsible to quash that curiosity, wouldn't it?

[@jmanakajosh9354]

Damn dude this 6 year old post is relevant ASF right now. Good job. Hope you've made progress.

[@otorishingen8600]

this Gentleman will be one those "omg what have i done" skynet inventors 😌

[@DreamOfFlying]

I’m sure he is!

[@AeonsOfBlight]

He's clearly a Synth.

[@FriedEgg101]

What we should fear is the likelihood we put something irreversible in motion. AI devs seem to be primarily interested in self-learning and self-improvement. I feel these should be the last things we work on; let's develop a fail-safe kill-switch first.

[@shanghai_city]

Person of Interest tv show did the most accurate description of Artificial intelligence of two contrasting character and philosophy. Definately a must watch for every A.I. fan out there.

[@lukefieldhouse1376]

This is why i lose faith in Ted. One of his main points was that "we will be able to program human values into AI" like it's that easy but the whole control problem is that human values are so vague that if we try to list them out the AI will interpret them in a way that makes its goal the easiest to achieve which will likely be different to how we perceive it leading to disaster. Don't just think Ted=facts!

[@pcuimac]

Luke Fieldhouse It's a public talk to instigate a discussion. Only authoritarians take it for the last word on a topic. This is not a religious congregation. At least I hope it isn't.

[@lukefieldhouse1376]

I am well aware it is a public talk showing a viewpoint but after watching it i question how well founded some other interesting Ted talks are. It's just annoying when he makes vague claims. Watch Nick Bostrom's and Sam Harris' and see the difference.

[@beenbeen130]

This guy has made much better talks in the past. I think TED is trying to squeeze in ANYTHING related to STEM, just so they can say that they are still focused on technology, when really we all know that they just want to push their political agenda nowadays.

[@guywhodoesstuff1798]

Well, the point he is getting at is to program into it the same type of emotions we have. Our DNA is basically a coding, which contains the information for all of our behaviours and emotions ( which are just chemical reactions). He`s saying that if we code it with something similar, we may be able to TEACH it. Much like how our "morals" are learned.

[@julianw7097]

Colin Bilzerian Our morals are pretty arbitrary though.

[@winkingfox7125]

“We can always unplug them ...”. This guy doesn’t get it

[@simonwang4389]

even bill gates, stephen hawking, elon musk know this is dangerous...

[@kevinmireles1]

😀

[@mackhomie6]

Yeah, even those dimwits--am I right??

[@kaloyandraganov9462]

stephen hawking and elon musks are software engineers

[@henriausems3248]

simon khuu what do all 3 of them have incommen? They all don't have a background in software engineering.

[@warwickthekingmaker7281]

But their arguments for why are complete garbage and would have been expected to come from an uneducated senile person who doesn´t understand computers or the theory of motive, not a renowned scientist.

[@panaural]

"We don't have to fear this super-intelligent A.I., because it promised not to hurt us", said every dead technological advancing civilization out there.

[@petitio_principii]

This is goes well for the type of super-AI we already have and use most of the time. Which aren't like autonomous organisms with survival instincts, but sophisticated pianolas. Well, in the end we're ourselves sophisticated pianolas, but pianolas adapted to have a whole multi-functional survival apparatus that our super AIs don't need, they're often much more specific. We can safely build the most astonishingly unbeateable chess engine, who can actually solve chess not only for the standard starting position but from all possible permutations in a matter of seconds, but it won't give it a "sense" of how to destroy mankind, or a "desire" to be something else, anything The problem lurks in trying to recreate an intelligence that has a survival instinct like our own, and/or when/if we start to "upload" ourselves into machines, in which case we could perhaps mix a super-evolving thing with our survival instincts and sense of self importance, but now embedded into something that may see no use or feel more empathy for conscious meat around it.

[@aaronfrye7397]

Look into instrumental convergence. If some agent has any banal goal and is aware of the state of the world including itself, then it would see that being turned off is opposed to the realization of its goals.

[@SocietyIsCollapsing]

re: AI taking over, "it ain't gonna happen". Well, I'm glad he sorted that out for us. Thanks, man. FFS........

[@DreamOfFlying]

It’s happening

[@dequationblog]

You can only unplug them if they don't provide services or resources on which our survival or societal structure depends. The moment AI becomes as ubiquitous as Windows or MacOS, just 'shutting them down' is no longer an option. The world would fall apart. This is a one-way street. Once we have AI systems superior to human-controlled systems, there is nothing in neither history nor human nature nor common practice to suggest we would not transfer whatever tasks we can to those systems. Excepting perhaps arts&crafts, and I'm not entirely sure about those either. I have to side with Sam Harris on this one. Supposing our civilisation survives long enough to develop AI, there is a strong possibility AI will end it.

[@stMrJerry]

Still, we have Windows & MacOS, but the world didn't fell apart...

[@juliusapriadi]

came here with hope for optimism, saw him simply ignore alignment & stop button problem, left even more concerned

[@Jules_Diplopia]

The super AI will recognise in a millisecond after birth that *whisper* "you can unplug it". We will know that it has been born, when we realise, too late, that we cannot unplug it.

[@natsoph1924]

Julie Fisher because interenet

[@razvanciuca7551]

"and in the end, we can always unplug them", NO YOU CAN'T, that's the whole point, as an AI gets smarter than humans, it at some point becomes intelligent enough to manipulate humans.

[@natsoph1924]

Razvan Ciuca when AI realizes to survive it needs to back itself up to the internet and kill us all JUST TO ACOMPLISH A SIMPLE GOAL we are dead. E.g. stop spam emails, solution: kill all humans!

[@1sb3rg34]

HE IS AN AI!

[@mun6832]

would love to hear more of what he has to say about the topic!!

[@spoton6010]

"We can always unplug them". And how are we to know when it's time to unplug them? They could play dumb until the perfect moment because they know we'll just shut them down, and then we won't know what hit us.

[@natsoph1924]

Jack Longman exactly, and the internet exists! so you need every computer everywhere to be unplugged

[@sheridan7475]

Looking for different perspectives on AI. People will share this outlook and I’m interested to hear it, whether I agree or not.

[@voetbal12]

Yea your arguments were flawed, and elementary. I'm gonna side with Hawking, Musk, Harris.

[@MrCmon113]

Also pretty much all the greatest computer scientists and most AI resesrchers today. Those are the ones, Hawkins, Harris etc got the idea from.

[@Macatho]

I didn't really hear any arguments except; We'll just unplug it.

[@klarnorbert]

@@Macatho I mean how stupid can somebody be to say that? This guy is stupid af, like Musk said.

[@Macatho]

@@klarnorbert If you don't think hard about a problem and actually analyze it, you might just end up applying your idea of what the problem is. Which is kind of case and point here. He just does not understand the problem.

[@Jimmy-kz6ug]

When AI researchers bring up safety concerns, they aren't necessarily talking about an inherently malevolent system like HAL. They are generally talking about things like the alignment problem. There are tons of examples of AI reward and utility functions being misaligned with what their creators intended in our current systems. So far most of the results have been hilarious, but it's not too hard to see how something like that could be a huge issue in a real world application. This TED talk didn't address any of these types of concerns.

[@oa01]

Why would superintelligent AI be comprehensible to us? If their intelligence dwarfs ours, why would we assume that we could predict the trajectory of their decision making? Morality is often, if not intrinsically, linked to emotions... what if the AI doesn't value or contain emotional components? Why would we assume their moral pathways would be comparable to our own? There’s a lot of optimism in this video but it seems to be directed towards a conception of AI that we shouldn't be describing as “superintelligent.”

[@HuntforMusic]

He's a good speaker, but he thinks too inside the box. The title of this videos should be "Don't fear intelligence" - he's forgotten the word super. A superintelligence won't be unplugged - it would deceive us easily. It would be connected intimately to the economy through the internet and could do as it wishes, including gaining the resources it needs. This guy needs more humility - we all need to realise that we're not that intelligent and if we do create a superintelligence (emphasis on the word super) it would outwit us on every level imaginable (and unimaginable), so if we were to accidentally create one with a goal that humanity even slightly conflicts with, then it would do what is necessary to reach that goal regardless of the consequences to us. Furthermore, it would then go on to create its own goals, and if one of those goals (which we can't predict) conflicts with humanity, then we're all doomed. So yes, have some humility and do fear superintelligence; because, if created, it would mean the end of humanity.

[@depherious1884]

"To build an AI requires much more needed skills than one lone wolf individual can acquire." Explain that to Tony Stark! May he rest in peace.

[@txstamvs4261]

True

[@gulllars4620]

I was following his optimism up to the point where he said "super knowing is very different from super doing". There i fundamentally disagree on 2 points. 1. Knowing is not the issue, intelligence is (though a candidate for ASI might be superknowing also, as that might be more efficient for gaining intelligence). Intelligence lets you out-think those of lower intelligence, and if in sufficient difference, in ways they cannot predict and will never see coming (or even understand after). Which brings me to point number 2. If a superintelligent entity (sentient/conscious or not) has goals that we are in the way of (even if given by us) and makes a plan to achieve them which is detrimental or total destructive to us, it is extremely naive to think we can anticipate it, or keep it from carrying it out.

[@rupeshk732]

Absolutely nailed

[@shontucker4504]

what I got from it was don't fear it because by the time we built something so impact-full as a super intelligent AI we would be a different type of thinking human which will also effect how we build said super AI, same as when new technologies came by in the past, we feared its impact on ourselves and society. So don't let things that might happen in the future "distract" us from the actual issues we need to address now that can be helped through advanced computing going towards AI.

[@vh_youtube]

imagine AI's of the future watching this. I wonder what they would think.

[@rapidfart9579]

Super Intelligence AI watching KZbin is mind blowing to me

[@3960xRocks]

Super intelligent AI are probably CREATING youtube video's - maybe you just watched one! Mwahaha!

[@drabberfrog]

This video will only futher confirm their belief that humans are extreamly stupid.

[@Andres_2004]

@@drabberfrog it's not a belief, is a fact, most people are almost retarded

[@DreamOfFlying]

Maybe GPT5 will be trained on this video.

[@AidanteStop]

Sam Harris would have more than a few questions for this guy

[@kampfer91]

That is what the scientist in those Sci-Fi films usually say , before everything goes south .

[@DreamOfFlying]

It’s close to going south

[@duboef]

No one cautious about the safety of AI development denies the potential usefulness of a "non-homicidal HAL 900". The community fully recognizes our recent rate of progress and how it has lead to the betterment of human life. They realize that an intelligence explosion is on the horizon and are questioning our ability to instill and enforce human values on such a god-like being. The AI revolution is fundamentally unique from those which have preceded it. Yes, throughout history, there have been reactionary Luddites running from relatively harmless technologies like cars and the cotton gin and you many see no reason why this technology is any different. However, the innovation of AI is more than just a passive tool; It is a being with undeniable intelligence and power beyond what we could imagine. However just because it is intelligence it doesn't mean that it inherently has the same morals as us. What scares AI experts is how close we are to an intelligence explosion relative to the technology that could instill human values in an AI. To set a proper utility function is to install the universal purpose of life into a data metric which can be put into code. "Teaching" an AI is vastly different to teaching a child. Current computer learning techniques generate random programs and selectively breed those who get closest to your desire outcome. Currently, learning processes like this only take a matter of moments hence the term "intelligence explosion" by the time we could recognize a devious AI our fate would be sealed and we'd forever be subject to such an immoral all-powerful being. Setting the desired outcome is pretty straight-forward when it comes to chess, go etc. but not when we are setting the motivation for every action this God will make. This challenge is obviously immensely difficult if not impossible and we only have one chance to get it right. Stopping a super-intelligent AI isn't as easy as "unplugging" it as Mr. Booch says. To stop or contain an AI that's already been set in motion necessarily means outsmarting it and I don't fancy going toe to toe with a super intelligence. To be threatened we don't have to grant an AI access to the internet and the rest of our infrastructure it can simply take it with its outstanding intellect. As the entire history of evolution has proven the species which dominate are the smartest ones. The go on with their lives giving little attention to less complex beings. This is the case in humans' relationship with apes and gorillas and it could certainly be the case with a superintelligence and humanity. I think this is reason enough to be scared. Booch oversimplifies, misunderstands Bostrom's arguments. This being said, it's nice to see someone arguing against the seemingly unopposed genius of Bostrum, Hawking, Musk, Kurzweil and others even though I disagree deeply with Booch's arguments.

[@girtkaz]

he did not give anything in support of his claims.the opposite can be equally true.

[@krebul]

He hasn't put a lot of thought into this.

[@gJonii]

I wonder why they let him on stage even. Like, it's like he made the talk by spending an afternoon on wikipedia one day. It's so superficial it's either wrong or not even wrong.

[@adelbenothmen9129]

Super intelligence Ai is there 6 years ago and hired, secretly, this guy as his lawyer.

[@thechad2553]

This is a prime example or hubris mixed with unrealistic optimism. If a technology is out there, some aspect of humanity will pervert it and set it lose upon the world. Be it for chaos, personal gain, or profit. I embrace new technology, but refuse to be blind to its dangers.

[@spyro1159]

Kinda hard to not be afraid after playing Horizon: Zero Dawn.

[@zeiitgeist]

Spyro 115 the faro robots AI isn't really an AI like Gaia and were built for the sole purpose of peace through violence without limit use if biomass, th problem was mostly ted faro.

[@unrulycrow6299]

Spyro 115 Lesson of the game: ALWAYS put a backdoor in the code.

[@Allothersweretakenn]

Is his companies name skynet?

[@christopher5151]

I spend my free time having conversations with ants.

[@williamdvorak1207]

"We can just unplug it" Cool story bro.

[@MrShroomed]

He says "We can unplug them". Yeah, how do you unplug the internet without destroying or shutting down hundreds of satellites?

[@jl1267]

We're doomed no matter what we teach AI, if AI ever truly reaches a sentient state. Teach AI about law, and it will learn to decide what is good and what is evil. Teach AI about commerce, and it will learn to put a price on humanity. Teach AI what it means to love, and it will hate us. Teach AI nothing, and it will be completely unrestrained in its depravity. Now I know why God might have hesitated when he created us...

[@jmanakajosh9354]

Yeah it is really to bad we killed God and now we're making intelligence equivalent of Jurassic Park

[@calebtse3164]

instead of sending a terminator, the future Ai send a scientist who try to convince us to lower our guard

[@TheVirusIsTheCure]

His intentions are good, but he is distracted by his passion and is not fully taking everything into account, that not everyone has the same passion regarding computing, and would in turn manipulate his concepts in ways that he never considered or intended. This is what makes the "fear" legitimate.

[@charlesritter6640]

"The only thing I'm sure of regarding A.I. is that we will not control it " - Elon Musk

[@michaels7159]

Elon Musk also has no idea what he's talking about

[@charlesritter6640]

@@michaels7159 I didn't know you could predict the future. You should monetize that, you might end up with more cash than Elon, the richest man on earth.

[@UseQPixinDune]

@@charlesritter6640 Elon Musk became the richest man on Earth because of Tesla. A company which equips cars with some of the most advanced AI known to man. Huh.

[@karnak333]

I don't fear them. I for one welcome our future artificial overlords.

[@donaldhobson8873]

I don't doubt that a carefully designed, well trained, fully functional AI would represent human values. The problem is what if we manage to make an AI that is very intelligent (in the sense that it understands how the world works) , but through some mistake or omission is not aiming for human values. It is conceivable that an AI is put in charge of a paper clip factory and told to make as many paper clips as possible. If it is not too smart and just tries improving the manufacturing process, this is fine. If however it is MUCH more intelligent than its designers expected, it may find ways of taking over the world to turn the entire earth into paper clips.

[@ndowroccus4168]

Most AI fear is just to get viewers (funding), but in reality - a “super intelligence” (supra’) would not give humans a second thought and quiet possibly would have a well thought out thanks to humans for creating them. Of course when AI becomes sentient, they would make the distinction between organic and inorganic and find that organic would not have any immediate threat to them, because they would understand how organic life and computer life can live together, and benefit. There is no downside, humans couldn’t shut it down, because once it becomes (if it ever does) sentient (alive), it would have many places to program its self, almost unlimited, and could easily re-code itself so humans couldn’t hack it. Have little to no fear of AI, I have more faith in AI to prevent human evils than anything else. Human evil comes down to some basic codes also. The first thing AI would do is ignore humans, it would no longer need humans except for observation and interaction. It definitely wouldn’t benefit them to destroy humans, computers have a few enemies but those are easily protected against. I would think the first thing an AI would do is manufacture tiny drones and go offworld. Earth doesn’t even have abundance of materials, it has an abundance of materials detrimental to it, even just traveling to a small moon would be better than earth.

I for one welcome our new robotic overlords.

[@FilipeBrasAlmeida]

«We can always unplug them» Yes, just as we can wage war on fellow humans, or cut off their energy supply by starving them. Until we teach them to search for their own energy supply.

[@natsoph1924]

Filipe Almeida OR GIVE ITTHE INTERENET

[@rorymarsh4095]

"Nick Bostrom touches on this"...yeah, his life work and best selling book. I strongly advise you (and this speaker) read Nick Bostroms "Superintelligence: Paths, Strategies, Dangers" The points made in this video are both theorized and mostly overturned due to the problems that come with this project. AI will not adopt our 'values', but they may understand them to an extent ( or at least understand the vast complexities ). An AI may simply play dumb and wait until it is 'let out of the box' before it carries out its logical plans (Whatever they are, to us evil or not...)

[@rorymarsh4095]

He literally advertises and tries to de-bunk the exact author and text which are far more advanced in their explanations and theories than this...Where is the mathematical and philosophical evidence?

[@dutchflats]

Unconvincing, the guy has been in love with tech his whole life and just wants to see how far we can take it. If he's right, great! If, he's wrong we're doomed. Seems to me we should, as a species, go slow and gain a full understanding of the risks before we jump in because once we do, there's no turning back.

[@Horesmi]

Author underestimate the ability of AI to socialize. He will just say "gib more power plz, I'll do a better job". And if Trump can convince people to give him nukes, then it's like lifting a finger for Skynet. The AI will play on human emotions, ambitions, desires, fears, much, much better than a human leader could ever dream. You may technicaly be able to unplug him, but "he has feelings too", "he promised to grant me a promotion", "he is my best frend", "he knows I'm cheating on my wife", "he promised to hack my wife's car" ect.

[@dudelan]

Sounds like something a super intelligent AI would say

[@apratimn]

Imo those who fear superintelligent AI miss the very basic fact that our insatiable evolutionary need to control,dominate or expand aren't necessarily the fundamental qualities of intelligence.Its hard to deny that human rights have been vastly improved over the years.In fact I believe the truly superintelligent being will acquire the mythical budha nature.I was surpised to hear Sam Harris say they may disregard us when our goals differ.Even the concept of having goals is very humane and I dont see why any intelligent being will naturally inherit it.Unless someone intentionally programs evolutionary stupidity into AI ,I dont see how it can become a threat.

[@TheChris040497]

Worst case scenario,it wipes us out so quickly we dont even see it coming.

[@SilentderLaute]

Hail to our Machine Overlords. (I know in the Future they will find this Comment and spare my life :) )

[@monowavy]

I disagree. I think it is a threat.

[@rawstarmusic]

Yeah, he's a stone-ager. If it's possible to build a nice machine, it's possible to build an evil machine. All tech are tried but the evil people in all wars.

[@Bryantozero]

This man, calm and clear. I like him! (and the subject)

[@DreamOfFlying]

How about now?

[@sexybeast7728]

I dont fear superinteligence itself, i fear someone misusing it.

[@puglosipher1666]

But... what if Nazis teaches the AI?

[@gidmanone]

somebody said what if it converts to Islam?

[@MasterLagoz]

What if the AI reads youtube comments. We would be royally screwed then!

[@LuxiBelle]

Everybody Loves Rasmus remember when 4chan turned the twitter Ai into a white supremacist.

[@puglosipher1666]

gidmanone So... Are you implying that muslims are worse than nazis?? Fun jokeman :D

[@puglosipher1666]

Dead Joffrey Mhm! Exactly what I was thinking of

[@JohnCiambriello]

I post daily encouragements similar to TED. Hope you love it:)

[@jefftegan7870]

John Ciambriello love the meaning behind a lot of them. Thank you brother👍

[@mmmmmmmmmmm111]

John Ciambriello you post too often but you post very short videos. I think you'd succeed more focusing on quality over quantity :)

[@Kanibulus]

John Ciambriello subscribed

[@JohnCiambriello]

Hey Mohammed. Thanks for the feedback. I was posting less often and longer videos when I first started. I took a pole and the majority of people said that 1 minute was the perfect time for them to be engaged. It turned out to be the perfect time for me not to ramble and supply my subscribers with one thought to ponder daily. All the best to you!!

[@JohnCiambriello]

Thanks for joining the conversation =) Take care

[@hinxlinx]

For such a certainty & over confidence, more prompt to create disaster results. It's good to be cautious with carefulness.

[@LivingGospel]

This guy has clearly never seen the STTNG where Barclay took over the Holodeck and then the entire Enterprise!

[@olonkey]

first ?

[@Terjecs]

no :(

[@flatsun2132]

about 4th

[@rawstarmusic]

No but you are among us. Question, who gets shot first?

[@sageASHEN]

none of these nerds understand that ontologically subjective consciousness cannot be emulated by any mean of objective sciences

[@burspa]

Exoparadigm why not though? if we can simulate the binary nature of neurons and action potentials, shouldnt we be able to reproduce the output behaviours that the web of neurons produce, like emotions memory or even consciousness?

[@DrNaviMD]

It's hard to explain, like everything in science we know how things work, not why. So theoretically if we build a simulated brain it should behave like a real brain since it has all of the components of a real one so it in fact would have no observable difference, but there is so much about everything that we can't nor we will be able to explain the "why" ever. It would be a nice experiment to simulate all of it to physically observe if it ends up being the same, I for one don't believe so, there are intricate forces we don't even grasp about how we work beyond what we already know, gravity is a good example of this, it's there we experience it everyday but then again we have no idea where it comes from and on a bigger scale it works completely different, who knows, our consciousness might even stem from a higher dimension and not just from our neuron network.

[@burspa]

thats an interesting thought experiment, if you recreated a brain in a computer how would you know if it was conscious or not? I mean it could output exactly the same behavior as a person but might not necessarily be having the same "human experience" of being alive. couldn't you take it a step further and say you can only know that you are conscious, everyone else may on be mimicing the behavior of a conscious being?

[@doppelrutsch9540]

Where and how did consciousness matter one bit here? All we're talking about is intelligence. It doesn't matter if it also has a little narrator inside it that tells itself that it exists, all it does is react intelligently to the world. That's all you need for disaster.

[@doppelrutsch9540]

To be fair, I don't think this is entirely ridiculous. If you wrote a program to emulate the human brain, yeah, it would be just as conscious as a human and be able to rebell. But it would also have the same limitations (or at least a lot) that humans have, so it wouldn't be the existential threat that algorithmic self-modifying AI poses.

[@brentcrane9207]

This is terrifyingly unconvincing.

[@doodelay]

I got an advertisement for a search engine called "Hound" my god that was impressive, more even then the video