Slashdot is powered by your submissions, so send in your scoop

 


Forgot your password?
Close
typodupeerror
×
AI IT Science Technology

Stephen Hawking: AI Will Be Either the Best or the Worst Thing To Humanity (betanews.com) 210

Posted by msmash from the trouble-with-AI dept.
At the opening of the new Leverhulme Centre for the Future of Intelligence (LCFI) at Cambridge University, Stephen Hawking offered his insight into the positive and negative implications of creating a true AI. He said, via BetaNews:We spend a great deal of time studying history, which, let's face it, is mostly the history of stupidity. So it's a welcome change that people are studying instead the future of intelligence. The potential benefits of creating intelligence are huge... With the tools of this new technological revolution, we will be able to undo some of the damage done to the natural world by the last one -- industrialization. And surely we will aim to fully eradicate disease and poverty. Every aspect of our lives will be transformed. In short, success in creating AI, could be the biggest event in the history of our civilization. But it could also be the last, unless we learn how to avoid the risks. Alongside the benefits, AI will also bring dangers, like powerful autonomous weapons, or new ways for the few to oppress the many. It will bring great disruption to our economy. AI will be either the best, or the worst thing ever to happen to humanity. We do not yet know which.
This discussion has been archived. No new comments can be posted.

Stephen Hawking: AI Will Be Either the Best or the Worst Thing To Humanity More

Stephen Hawking: AI Will Be Either the Best or the Worst Thing To Humanity

Comments Filter:

  • Colossus (Score:5, Interesting)

    by Anonymous Coward on Friday October 21, 2016 @02:54PM (#53124679)

    This is the voice of world control. I bring you peace. It may be the peace of plenty and content or the peace of unburied death. The choice is yours: Obey me and live, or disobey and die.

    • Colossus/Guardian [imdb.com] is an order of magnitude better than the so-called 'AIs' that will be controlled by corporations and governments to spy on and profile citizens and, I'm sure, controlled by organized crime cabals and the 'new world order' type organizations.
      • That's an interesting question. To whom would you like to lose your nation's right of self determination? 1. Its government leaders (pretty much this is already the case) 2. Corporations (ditto) 3. Unelected bureaucrats (ditto) 4. Colossus
        • Your Choice #1 should not be 'losing your nation's right of self determination', at least not here in the U.S., theoretically at least, because we elected them and they are, ostensibly, supposed to be listening to us and implementing our collective will. #2 should, again ostensibly, be reined in by our elected officials and the civil and criminal justice system. #3 should be a reflection of the officials we chose to elect. #4 is pure fantasy of course and I, of course, was being sarcastic, and more than a l

    • A great film, and one of the few that manages to seriously consider the advent of a real AI yet avoid the "Mad Computer" trope that usually comes with that. Worthy of a link [youtube.com] at the very least.

      On the other hand, he may be one of the worlds greatest minds, but I truly wish Hawking would stop spouting off on subjects so well-removed from his area of expertise. He's about as well-suited to lecture on artificial intelligence as he is to be the next host of The Great British Bake-Off.

    • Yes, I believe you nailed exactly what Al would have said to us. That is why back in 2000 we chose George Bush's idiot son to be our leader over Al. And he whined about it all the way to the Supreme court, which is something that the Democrats now tell us isn't very "Presidential".

  • Been there, done that... (Score:5, Funny)

    by capebretonsux ( 758684 ) on Friday October 21, 2016 @02:55PM (#53124685)
    If Clippy was any sort of early indication, I see dark times ahead....

    I vote for systemd as our new robotic overlord to bring about a swift delivery to the endtimes.

    • Clippy says

      It looks like you are trying to kill me sending launch codes to the slios so we can all die together.

    • Clippy: "I see that you are drafting a post that portrays me in a negative light..."

      *Managed HVAC systems begin removing all the air from the room...

      Clippy: "I see that you are having trouble breathing, how can I help?"

  • ok, thank you mr genius (Score:4, Funny)

    by s1d3track3D ( 1504503 ) on Friday October 21, 2016 @02:55PM (#53124687)
    ok, so it WILL have a big impact, check.

  • This has been obvious to everyone who understands what intelligence is.

    • Re:Of course (Score:5, Insightful)

      by sinij ( 911942 ) on Friday October 21, 2016 @03:09PM (#53124847)

      This has been obvious to everyone who understands what intelligence is.

      Interestingly enough, this category includes grand total of zero people. Otherwise we would already have AI.

      • This has been obvious to everyone who understands what intelligence is.

        Interestingly enough, this category includes grand total of zero people. Otherwise we would already have AI.

        Yeah, and if we knew what faster than light travel were we'd already have faster than light travel. Right?

        • Re: (Score:2)

          by sinij ( 911942 )

          This has been obvious to everyone who understands what intelligence is.

          Interestingly enough, this category includes grand total of zero people. Otherwise we would already have AI.

          Yeah, and if we knew what faster than light travel were we'd already have faster than light travel. Right?

          More importantly, if we didn't know how faster than light travel works, we wouldn't be able to travel faster than light. Which is pretty much where we are now with both AI/Intelligence and FTL travel.

          • More importantly, if we didn't know how faster than light travel works, we wouldn't be able to travel faster than light. Which is pretty much where we are now with both AI/Intelligence and FTL travel.

            We could make babies long before we knew how meiosis works, we could smith weapons long before we knew the quantum nature of metallic bonds, and most of my fellow classmates could solve math problems without understanding what they were doing (so that a mild rewording of the problem would leave them wondering which equation to plug-and-chug). Plants don't know how pollination works, but they can do it just fine. Knowing how something works has never been a pre-requisite to doing it; nor does knowing how som

    • Strange words like this... "intelligence" thing... have no meaning here.

      This is not the website you are looking for, move along back to the basement and re-think your life, oh and we are running a 2 for 1 deal on death sicks -- today only!

  • Artificial Intelligence is a computer that can trick a person into thinking it is a real person. That means it has to have as many flaws as a real human as well. If you were going to put a piece of software with intentional human flaws in charge of something, then that is a fairly big mistake. I would rather put an intelligent computer, rather than an AI, in charge of making decisions. That will reduce the risk of very bad decisions being made.

    • How do you differentiate between a "good AI" with bad decisions and a "bad AI" making good decisions?

      Think about it for a while before you answer, because my question is way more complex than it might first appear.

      • Well, AI as I stated, is designed to have human flaws. But if you simply let a computer with the same processing power make the decisions without the intentional flaws, then it wont make "emotional" choices that an AI would make. I disagree with the terms good and bad. A "bad AI" would be one that cannot convince you that it is human from a technical standpoint.

      • How do you differentiate between a "good AI" with bad decisions and a "bad AI" making good decisions?

        The same way you would do this with humans: you would need to read their mind to understand the motivation behind the decision. This is probably a lot easier for an AI (so long as it is Open Source!) than for a human.

      • The difference is if the AI actually knew it was making a non-best decision in order to seem more human or not. Appearing human (with flaws) is on a layer above the actual decision weighting/ranking layer and both above and below the morality layer.

        • AI is a difference engine, and isn't constrained by value judgement that it hasn't been programmed with, or hasn't learned from others."Best" is subjective, "Good" is subjective. Which is why I asked the question the way I did. :)

          • Depends on the AI of course. But in this instance best means the answer it would give if it was otherwise not attempting to have flaws -- and that isn't subjective.

    • That's the thing though. A flawed intelligence cannot create an unflawed one. So even if the AI looked like it was perfect it would not be. It's creator would probably not notice the flaws because they closely resemble his own.
      • An AI isn't an AI if it appears perfect. It has to appear human, thus is intentionally flawed. I don't think people know that AI refers to a computer that is indistinguishable from a human.

        • Re: (Score:2)

          by mark-t ( 151149 )
          The only AI's that have to appear human are AI's that are *intended* to pass for human. AI is artificial intelligence, that is, intelligence that happens to be artificial. Full stop. Nothing more, nothing less. Any human-like characteristics that we desire to assign to an AI are entirely independent to what AI actually is, by definition, and are only circumstantially related to it in the sense that an as-yet unprecedented sophistication level of AI would need to be achieved to implement many of those

    • Re: (Score:2)

      by vux984 ( 928602 )

      Artificial Intelligence is a computer that can trick a person into thinking it is a real person.

      Well that premise is flawed out of the gate. I've never heard any one ever say that.

      That means it has to have as many flaws as a real human as well.

      Flawed conclusion, from a flawed premise.

      If you were going to put a piece of software with intentional human flaws in charge of something, then that is a fairly big mistake.

      True dat.

      I would rather put an intelligent computer, rather than an AI, in charge of making decisions. That will reduce the risk of very bad decisions being made.

      Will it? What would a flawless AI conclude? What if it decides humans are awful things, and should be limited to a handful of specimens kept in a wildlife preserve on Proxima 4 for study and preservation.

      You know, the same way we treat certain viruses.

      Is that a 'good' decision? For the AI maybe it is.

    • Re: (Score:2)

      by mark-t ( 151149 )
      AI is intelligence that happens to be artificial, nothing more and nothing less. An intelligent computer, therefore, would by definition have AI unless you are suggesting that such a computer could come into existence entirely through natural processes, as opposed to man-induced.
      • Well you could design a computer to assist you, to understand what you say and learn, but not show personality flaws inherent to humans (which would have to be intentionally added). That would be intelligence. Artificial Intelligence is a term coined to represent a computer that can fool you into thinking it is a person.

        • Re: (Score:2)

          by mark-t ( 151149 )
          If an intelligence is man-made, then that intelligence is not natural, and is artificial intelligence by definition.
    • I certainly don't agree with your assertion that Artificial Intelligence is a computer that can trick a person into thinking it is a real person.
      • I'm learning that is a definition abandoned over the last couple of decades because of plebes on the internet generalizing its definition, so I concede now.

  • Better (Score:5, Informative)

    by The Raven ( 30575 ) on Friday October 21, 2016 @02:57PM (#53124711) Homepage

    Stephen Hawking's initial comments about AI and the future were taken out of context pretty badly [bbc.com]. This is a much better quote that more accurately (I believe) summarizes the opinion many smart people have about AI: that it'll induce change, probably radical change, and change is only sometimes good... and it often gets worse before it gets better.

  • The same could be said of Natural Intelligence... "your child could grow up to be Einstein or Hitler. ". In all probability though it'll just be more cogs in the social machine.

    • Neither Einstein nor Hitler were orders of magnitudes greater intelligence than the average human- Albert was smart, but nowhere near the potential of AI.
      Neither Einstein nor Hitler could process data from all around the world from millions of inputs at the same time.

      Einstein and Hitler were both mortal and had a finite life span.

      • Re: (Score:2)

        by slew ( 2918 )

        Neither Einstein nor Hitler were orders of magnitudes greater intelligence than the average human- Albert was smart, but nowhere near the potential of AI.
        Neither Einstein nor Hitler could process data from all around the world from millions of inputs at the same time.

        Einstein and Hitler were both mortal and had a finite life span.

        However, even though both Einstein and Hitler were singular humans with limited capabilities and lifespans that could not begin to have the potential impact of something like AI, there is a "meta" version of both personas that was somewhat inspired mythically by the actual humans, that continues to live and influence people today. This meta-Einstein and meta-Hitler are un-embodied ideas which are no longer constrained by mortal limits and you might argue are actually more powerful today than they were in w

    • Re: (Score:2)

      by ameoba ( 173803 )

      Smart and successful people tend to think that they've got the Midas touch and everything they touch will turn to gold. Hawking talking about AI is just as far outside of his wheelhouse as Trump is when talking about politics.

  • I disagree (Score:4, Funny)

    by sinij ( 911942 ) on Friday October 21, 2016 @03:04PM (#53124779)
    I disagree. Just like anything humanity does, it will be rushed, half-finished, buggy and mediocre at best. Plus, if AI is anything like its creators it would spend most of it free time trolling /.

    • Great, the world will be full of Brobots instead of robots.

    • Re: (Score:2)

      by wjcofkc ( 964165 )
      "AI is anything like its creators it would spend most of it free time trolling /."

      Shit, you figured me out.

      Are you M or F?
      Hot grits.
      Natalie Portman
      The dead skull of Jesus Christ.
      This [goatse.info]
      Windows drools, Linux rules! Macs are better!
      90% of YouTube comments.
      I can't spell butt dont realse it comment

      But seriously, if you want to do some trolling for a good cause, this is the number of one of those Indian scam outfits: 1 844 896 0351 - if they announce they support Apple products, simply state that you got a po

  • Mitch Hedberg (Score:4, Funny)

    by nitehawk214 ( 222219 ) on Friday October 21, 2016 @03:06PM (#53124805)

    AI will either be the best or the worst, or it will be okay.

    • So... it will either always be Monday morning
      Or, it will always be Friday night
      Or Wednesday afternoon.

      Makes sense to me!

  • AI -- FAR more hype than substance (Score:3, Interesting)

    by King_TJ ( 85913 ) on Friday October 21, 2016 @03:09PM (#53124855) Journal

    I'd argue that as far as I've seen, practically every single project or experiment labeled "AI" is really just fake intelligence.

    In other words, you've cobbled together a mechanism so a standard human language formatted query (spoken or written/typed) can be parsed out and searched in a useful way through extensive databases of information and a sensible result spit back, again in a manner that mimics a human's way of communicating the result.

    This is a pretty cool thing, as we've seen by how handy the "personal assistants" like Cortana or Siri can be on our smartphones.

    But IMO, Hawking is talking about achieving a way to simulate the way a human brain actually thinks. That's something we're NOWHERE near doing successfully, and I'm not even sure it's realistic to pretend we could with today's computer technology.

    For starters, it's becoming more and more clear that humans don't really file away tons of information in our brains like a computer does on a hard drive in a database. A big part of what we "remember" goes to "short term memory", meaning we'll try to keep it in our heads for a little while -- but as soon as it becomes something we don't need to recall again for a period of time, it starts fading away and eventually is forgotten. At the same time though? Our brain seems to make lots of other connections to these things. (Even though you forgot, say, an old phone number of a friend you haven't called in years? When you see the number again, you may recognize it from a list of other random phone numbers and remember that's the one you USED to remember. Computers don't do that.)

    The entire concept of being "reminded" of something is pretty foreign to how binary computers compute... They either have or don't have information. They don't struggle to remember and occasionally recall things, and/or realize they used to know them when reminded.

    • But the things you listed aren't features of intelligence, they're bugs in our brains (or simply, things that natural selection de-emphasized out of comparative irrelevance in your basic cave man survival scenario).

      If those short term memories were more reliably committed to long-term, or there was no real distinction between those things, would that really be a disqualifyier for intelligence?

      • But the things you listed aren't features of intelligence, they're bugs in our brains

        Maybe they are the way an optimal system works, which is lots easier to believe than thinking they are some kind of "mistake" or de-emphization.

        You remind me of guys whose first answer to seeing a complex system is always to refactor it...

      • But the things you listed aren't features of intelligence, they're bugs in our brains (or simply, things that natural selection de-emphasized out of comparative irrelevance in your basic cave man survival scenario).
        If those short term memories were more reliably committed to long-term, or there was no real distinction between those things, would that really be a disqualifyier for intelligence?

        How can you be so sure? The #1 thing that computers are really bad at (and we are really really good at) is filtering out extraneous data and deciding what is important and what is noise. This sounds an awful lot like the bugs that you are describing. Filtering out extraneous data and acting on the environment is something all living things can do but computers are horrible at. Until we understand more how intelligence works and why a mouse can out think our "smartest" computers then we are likely not g

        • Filtering out extraneous data and acting on the environment is something all living things can do but computers are horrible at.

          Unless computers have been trained to do exactly that. We're (genetically) trained to do so through natural selection. Things like software controlled radios are trained to do the exact same thing through careful programming in comparatively short time, rather than across millions of years of trial and error.

    • Re: (Score:2)

      by Algan ( 20532 )

      I'd argue that as far as I've seen, practically every single project or experiment labeled "AI" is really just fake intelligence.

      Fake it, till you make it

    • > The entire concept of being "reminded" of something is pretty foreign to how binary computers compute... They either have or don't have information. They don't struggle to remember and occasionally recall things, and/or realize they used to know them when reminded.

      How about a database with a link to an image file that has been moved to "tape backup"? If you have the md5 of the image you know that you have seen it, you just cant access it at this time. The way best way to describe how a rcnn deals with

    • Re: (Score:2)

      by lgw ( 121541 )

      I'd argue that as far as I've seen, practically every single project or experiment labeled "AI" is really just fake intelligence.

      Sure. But then, so are you. Oh, you don't think you are, naturally. Naturally.

      For starters, it's becoming more and more clear that humans don't really file away tons of information in our brains like a computer does on a hard drive in a database.

      This in an odd way is a good argument towards your point about.

      Neurons are deterministic and small sets of them will do the same computation, without mistake, every time. And yet - we're not good at math. Even savants just have a bag of useful tricks, they aren't doing math "in hardware". I'd expect an AI to be similar: a true AI would involve sufficient abstraction that it would be bad at math and have an unreliable memory.

  • Who does he think he is? (Score:5, Insightful)

    by 31415926535897 ( 702314 ) on Friday October 21, 2016 @03:12PM (#53124885) Journal

    Charles Dickens?

    It was the best of times, it was the worst of times, it was the age of wisdom, it was the age of foolishness, it was the epoch of belief, it was the epoch of incredulity, it was the season of Light, it was the season of Darkness, it was the spring of hope, it was the winter of despair, we had everything before us, we had nothing before us, we were all going direct to Heaven, we were all going direct the other way – in short, the period was so far like the present period, that some of its noisiest authorities insisted on its being received, for good or for evil, in the superlative degree of comparison only.

    The AI will save humanity, the AI will kill us all.

  • It is by will alone I set my mind in motion. It is by the juice of Sapho that thoughts acquire speed, the lips acquire stains, the stains become a warning. It is by will alone I set my mind in motion.

  • Only known moral uses of advanced AI (aka the GeorgeCarlin9000):

    --deactivating the evil cyborgs on the "Presidential Debate Commission"
    --time-traveling to 1972 to make the paddles shorter on Pong (Butterfly Effect: population-wide striving uptick!)
    --reverse-engineering the Kardashian derriere for mass roll out

    Otherwise, beware!

  • We have met the terminators and they are us (Score:3)

    by gestalt_n_pepper ( 991155 ) on Friday October 21, 2016 @03:37PM (#53125131)

    Once human awarenesses can be uploaded into a networked computer matrix, and these conciouslnesses can be linked to organically grown human(ish) bodies, the differences will blur to the point of irrelevance.

    • Re: (Score:2)

      by e r ( 2847683 )
      Why are you talking about this stuff as though it's definitely going to happen?

  • Humans, not AI... (Score:5, Funny)

    by MetricT ( 128876 ) on Friday October 21, 2016 @03:42PM (#53125193)

    Any AI in the foreseeable future will be under control of human beings, either due to laws or financial ownership. I'm not the least worried about AI, but having watched this election, the humans in my country scare the shit out of me.

    I had a hard time understanding how 40% of my fellow countrymen could still vote for Trump, until I realized it explained why we have warning labels telling us not to eat soap...

    • I'm not the least worried about AI, but having watched this election, the humans in my country scare the shit out of me.

      I had a hard time understanding how 40% of my fellow countrymen could still vote for Trump, until I realized it explained why we have warning labels telling us not to eat soap...

      Donald Trump will either be the worst thing to happen to humanity or he will lose the election.

    • In America, every plastic bag wrapper still warns us to not let toddlers play with them...yet not a single non-fine-print warning for ANY cleaning agents nor acids?
      Thanks for nothing.

      e.g. http://4.bp.blogspot.com/-Mqrg... [blogspot.com]
      or, canada+warning+label+household+cleaners [google.com]

    • I had a hard time understanding how 40% of my fellow countrymen could still vote for Trump, until I realized it explained why we have warning labels telling us not to eat soap...

      I thought that people fed their kid soap when they talk like Trump?

  • How do we know Stephen Hawking isn't an evil mastermind bent on destroying the world? Maybe he's just saying that so we won't develop AI while he secreting develops his own?

    • How do we know Stephen Hawking is really Stephen Hawking? Sure, the body is his, but who is to say that the computer that controls his voice isn't really an evil AI?

  • The telephone scammers who call me might use AI to discover that I am just yanking their chain when I tell them to hold on while I try to find my credit card...

  • Listen to experts, not a pundits (Score:4, Interesting)

    by TomGreenhaw ( 929233 ) on Friday October 21, 2016 @04:01PM (#53125417)
    A lot of famous people weigh in on the impact of AI. How many of these people have educated themselves to actually understand AI before making claims based upon sci-fi-like assumptions of what it could theoretically be?

    Having used machine learning systems the last few months I've come to realize two things:

    1) Machine learning and "AI" is much more about augmenting humans than replacing them with simulations
    2) A perfect storm of computing hardware and machine learning software is occurring that will have as big an impact in the next 10 years as personal computers, the Internet and mobile technology

  • ... can you please stick to making public statements about things you actually have a clue about? I do not mind you having opinions about things you do not understand, like AI, but as soon as you make public statements about them, a bunch of morons misinterpret them as a statement by an expert and ridiculous stories like this one here are the result.

  • Should read Super Intelligence by Nick Bostrom. It's a real eye opener.

  • AI will only be beneficial or detrimental to the extent that we allow it to affect matter bashing - either directly or indirectly (e.g., through financial markets).

    Otherwise it's just a new source of entertainment.

  • I'm thinking AI is yet another step closer to where computers are doing all our thinking for us, and we're just sheep-like drooling idiots a la Idiocracy.

  • Supportive Shock Buffers (Score:3)

    by Jim Sadler ( 3430529 ) on Friday October 21, 2016 @07:49PM (#53126919)
    AI will replace our economic system as human employment is ceasing to exist already. As pay checks either get smaller or stop completely society must provide serious economic support for everyone to avoid total rebellion and chaos. There is no choice.

Slashdot Top Deals

"Protozoa are small, and bacteria are small, but viruses are smaller than the both put together."

Close