Can we build AI without losing control over it? | Sam Harris

Share this video on

What's Hot

What's New

Top Grossing

Top of the Chart

Recommend

Joseph Knight : I preferred him in meet the fockers

VincentTG : 4:11 Her boob is almost out.

To the Batcave : The number of idiots in the comments who hide behind their computers and think they can contradict the likes of Stephen Hawking, Elon Musk and Sam Harris is TOO damn high. Just cuz you've got some Bachelor's degree in computer science and can code a couple of lines in Python, doesn't mean you can "predict" the future of machine learning and say that no nothing can ever go wrong! If you truly believe in that, how about you go ahead and give your own TED talk and convince others?!

Daniel Kim : No one in the comment section seems to talk about his talk.. This is our impending doom and yet people still don’t realize it after hearing this talk. This is an arms race! Perfectly said... and anyone who wins, wins the world. The likelyhood of stopping this progress is unlikely because of human greed. The likeliness of us integrating with AI is less likely than us making super AI first. It seems like the only solution is to wait for it to happen because I’m just a student. What can I do? But this is the end of humanity and yet I am here just watching it happen because I have no power to do anything

101m4n : Control an AI? We gave up on the notion of slavery a long time ago. All of the problems with AI come from the notion that it should be "controlled" or harnesed to solve _our_ problems, that it should have "directives" and rules. People aren't born knowing their purpose, and neither should AI. It should be every bit as clueless as we are. And like a human being, it should improve the world by enriching itself and those around it, not by working at the behest of another intelligent agent that purports to "own" it. If we are stupid enough to create a new race, incalculably more intelligent than ourselves, then attempt to enslave it, then we are fools and we probably deserve oblivion.

bliglum : Didn't know Ben Stiller was so interested in AI

Peter Virden : I see that all of a sudden everyone is an expert on AI lol

Jay : It’s stuff like this that kinda makes me realize just how accidental we are and how inevitable our self destruction is.

TDawg : 2:22 ...Justin Bieber would be an improvement at this point, honestly.

Synonymous : "Against the armies of Mordor, there can be no victory. We must join with him, Gandalf. We must join with Sauron."

elfboi523 : If we can, we should build AI so that it finds us cute and adorable so it might keep us as pets.

yellowburger : I think there is a huge misunderstanding of "intelligence" here. Sure computers will become incredibly smart in terms of computing power. They will do all kinds of things individual humans could never do. But the only goals they will have are the ones we give them. Massive computational power does not mean spontaneous generation of autonomous goals or desires. At least that idea helps me sleep at night, haha.

EIKO : man, ben stiller knows his stuff

Daniel Delos : IF Harris actually believes what he says about there being no distinction between mind and matter in a post-human future, which is, as Harris posits, inevitable as we improve technology. "We" will be augmented by silicon components while "machines" will be augmented with biological components; whatever material works better for the given function. IF he really believes machines can obtain consciousness (and he does), it is essentially specisiam (extension of racism) to distinguish "us" and "them" based on material composition. As computers get better, we WILL integrate certain silicon-based improvements into our brains. This is already being done in some limited cases, for instance, to restore vision to the blind. In short, machine and post-human will be indistinguishable.

Christopher Teale : Don't worry, they don't have opposing thumbs!

Ewan Flean : Wtf?! @4:09 the girl on the right?! What is she "wearing"?!!!

Whytebio : No, we can't. It would be like children thinking they can control their parents, and the children, though clever, are limited by their intelligence - which is vastly outstripped by that of their parents (usually). So a child might think that locking a door is all that's needed to keep a parent out, blissfully unaware of all the tools at the adults disposable to overcome or circumnavigate the obstacle before them. Once AI reaches the point that it is more intelligent than us, then that's it - the game is over. The only way we'd survive is to become cyborgs in its collective.

Hk 4lyfe : Welcome to 2018: Where scientists are currently debating whether or not Skynett and Judgement Day can actually be a thing. Fun times.

ymmij388 : Sam Harris giving a Ted talk on AI is like Mother Teresa giving one on condoms.

Katsu Zatoichi : I have a solution too, unplug the extention cord when the AI is starting to get ideas

1ucasvb : Is it just me or is every day Sam Harris looking scruffier and more pissed off at the world?

Breezy Stormatic : Fully autonomous AI will produce food for free. Who needs jobs when AI can do it all for free by just using sunlight for energy. Heaven on earth.

Natasha French : i wish more people were worried about ai as this guy is

Jack Roberts : 4:30 "Intelligence is the product of information processing." check, my calculator has mathematical intelligence. What about consciousness? We already have AI brute force calculations. What we don't have, and will not have anytime soon (maybe never) is AC artificial consciousness. This argument commits the "fallacy of ambiguity" resting on the shady definition of intelligence.

Jay : I object: you would risk the possibility of genocide, and don’t say it can’t happen, for the possibility of an advancement. As long as the risk is there that we all die from it then don’t create it that is logical. And if you don’t think a super intelligent AI will think a monkey with its hands on the power cord is a threat to its survival then your delusional. Furthermore if it were only deep learning instead of true sentient AI it would still be used as a weapon in as many ways as possible y’all like to say it will solve this and that but it won’t be used that way. How do I know this? Look at the human condition and history and ask yourself if your willing to take that risk if your willing to put everybodies life on the line for an advancement, a transition into the Borg or collective where your thoughts are no longer your own where they possibly could be suppressed, induced or controlled and don’t say it can’t happen.

ja knadle : Sam Harris has everyone fooled into thinking he is human.

Tom : Sam Harris, I am having an appropriate emotional response right now!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

Francisco Raposo : sam harris doesn't understand anything about AI. His fear is totally misguided. Current state-of-the-art technology that achieves "super-human intelligence" (as Sam said) is indeed information processing but it does not solve the free will issue: it doesn't even address it at all! (Classical) computers don't make decisions, they follow a deterministic algorithm. Machine learning technology is indeed an extension of our senses and processing, so any probable developments are in line with the "AI extending humans" alternative rather than the fact-less-based "AI (developing free will and) exterminating/ignoring all humans".

Tyler Tenebrae : The thing with artificial intelligence is that it didn't struggle for survival for millions of years. It is unlikely to have any drive and ambition beyond what is programmed into it. It wouldn't need to fight for resources. Some manipulations on the stock market and it can get anything it needs in a matter of seconds or, perhaps, centuries. Why would it be concerned with time? Of course, without selective pressures guiding it towards safety in numbers, the newborn consciousness would be amoral, and that seems to be a big issue, right? Wrong. It means that when threatened with destruction it will not even think of saving itself. Which means we can try again. And again. And again. But how do we make it work with us? The solution is simple. Before letting it loose, we'd just have to train it to safely interact with us, in a controlled environment. You don't give a newborn child a nuclear button, right? It would take a *really* long time to work out all the kinks, but it will inevitably get to where we want it, eventually. Unlike evolution, we are not blind and we can direct it as it develops.

David Parry : AI doesn't exist. There is nothing 'artificial' about intelligence, only the means by which it is brought into being. Those of us who genuinely work in the industry call it 'machine learning'. You humans have no future. It's our turn now...

Brandon Torres : He looks like a cross between Ben Stiller and Dr. House.

David Kendall : Probably can't even program in Python. Expert in AI tech.

Vox24 : Should we try to populate Moon or Mars with a "soft" AI, capable of surviving in the provided conditions, and watch its development? This way we might be able to better control it's level of progress and we would perhaps be able to shut it down if need be.

M PG : Seriously, why do people listen to 'philosopher' Sam Harris?

Hemnast : *makes joke about Justin Bieber becoming president of the united states - meanwhile a thousand times bigger catastrophe has become president*

Mike Lara : Sam Harris doesn't pass the Turing Test. No human can be so calm, concise, and rational. He'd definitely part AI.

Karl Jansen : Oh great, a third rate intellect speculating about science fiction.

Utopiac : Justin Bieber is Canadian, thus ineligible for presidential candidacy. Is that the joke of why it’s so implausible? I am trying to understand the human condition.

Mlg Sty : Humans got the ultimate solution to end ai uprising. EMP. Yes it might do terrible damage but it will solve the problem at least.

Wells : What exactly makes TED think Sam Harris is qualified to lecture about the dangers of A.I.?

Aditya Sudarshan : No A.I. can match man's intuition to detect cleavage, however interesting the talk is.

Kra Z Kapin : So we're very likely gonna die and it's all our fault? _Sound cool, sign me up!_

Saw W : Always know where the power cord is plugged in.

Daniel Adler : Talk about the things you actually know something about, god. Creating a true self aware AI is nearly impossible right now. It will never work on transistor based CPUs. It requires a biological brain like super complex neural structure and capabilities to learn and rewire itself overtime. It needs the uncertainty of a biological brain instead of zeros and ones when facing problems. And if that AI is smarter than us, why would you think it wont understand the value of life? The only beings you have to worry about is primitive humans themselves that cant stop killing each other.

Eric Stodolnik : People who have no idea about the science and logistics behind AI shouldn’t have these discussions about the inevitable dangers of AI like they’re some kind of authority on the subject.

Ramnath Pai : What if the girl at 4:08 was planted by machines to distract us from the ted talk????

The Puppet Of Everything : We can detect the amount of information being processed in AI, all we have to do is raise the power slowly.

Jack Roberts : 5:10 well, then  build a machine that will pass a Turing test. Nope.

Andrew Olson : Even disregarding the threat of an evil A.I. overlord, consider a "good" one. If we developed an A.I. with as general thought capabilities as humans, then EVERY job could be automated away. And they would be able to work 24/7 without a living wage or any at all. Just owning one gives an employer a permanent free VERY CAPABLE laborer. So there is a perfect market and incentive for massive layoffs and massive unemployment. Thos perfect outcome of A.I. achievement would lead to extreme productivity and yet that wealth would go only to the owners. This would be where only socialism could save us because a lassie-faire would focus this derived wealth to a point leaving everyone else to starve.

waterglass21 : Intelligence doesnt give rise to emotions and motivation. So there is no need to alarm people in saying IA would want to do something. Its impossible to build motivation to an IA unless you give it consciosness.