back to article Robot overlords? Pshaw! I ain't afraid of no AI – researchers

Artificial intelligence is all the rage in technology and as it progresses at a dizzying pace, the industry is at danger of being overhyped, say researchers. Speaking at a symposium about AI in bioscience, ex-IBM Watson veteran and head of technology at benevolent.ai, Jérôme Pesenti, warned the audience to have a healthy …

  1. Anonymous Coward
    Anonymous Coward

    As expected...

    “AI is an incredibly powerful tool and is already changing our lives. It performs better than a human at many things, but it’s nothing like a human. And it’s not going to take over the world,” Bryson told The Reg.

    That's what I'd expect to hear from a covert AI planning to take over the world.

  2. kars1997

    "But it’s unlikely that we will build human-like AI. Even from a philosophical aspect, it has too many moral hazards,” Bryson added.

    I love his optimism, where corporation wouldn't build an incredibly profit-generating AIs because of "moral hazards". Because a corporation would NEVER do such a thing :D

    1. Dan 55 Silver badge

      He also seems to have forgotten the possibility that an AI which has no concept of human-like morals and was originally built for making more and better paperclips turns humanity into paperclips.

    2. Joanna Bryson

      :-D But I'm actually a woman (despite being an engineer and a scientist.)

      1. allthecoolshortnamesweretaken

        You should post here more often - it's always nice to get feedback from someone who knows their stuff.

        And AI certainly is a topic.

      2. Dan 55 Silver badge
        Facepalm

        Sorry, too many men mentioned previously in the article and not enough buffer.

        1. Anonymous Coward
          Anonymous Coward

          If we CAN build a human AI, someone eventually will. If we hadn't done the Manhattan project for WW II the A-bomb would have been invented eventually - if for no other reason than "what if the 'bad guys' do it first?"

          1. John Smith 19 Gold badge
            Unhappy

            "if for no other reason than "what if the 'bad guys' do it first?""

            Actually that was pretty much the reason for the Manhattan Project.

            It was only post war it was discovered the Germans couldn't get a Graphite based reactor to work (too much Boron impurity IIRC) and had wildly miscalculated the critical mass of a device.

            The Japanese barely had a nuclear program to speak of.

      3. kars1997

        Noted! Sorry about that :)

  3. amanfromMars 1 Silver badge

    IntelAIgents Abhor a Dumb Vacuum ...... and Listen for Signs of Future Life from Earth

    “As humans it’s natural to apply human aspects to things we create. But it’s unlikely that we will build human-like AI. Even from a philosophical aspect, it has too many moral hazards,” Bryson added.

    Methinks it is a grand folly to make and/or take any bet against the building of AI-like humans ….. SMARTR Virtual Machines which program and produce and practically present all manner of futures for public and private and pirate consumption.

    Indeed, is such not currently the unfolding norm which is terrifying established systems of remote executive administration and mass consciousness manipulation, for they are unable to effectively counter them to hinder their rapid progress?

    Obviously the posit here is that such is certainly so, and as you now be so informed, what are you prepared to do about it rather than try to deny or studiously ignore the reality ….. or the possibility, if you prefer to ponder on such for a while.

  4. tr1ck5t3r

    This man reminds of the scientists who have made fMRI announcements before the flaw was found. To be fair, he's only going on what he knows, or as Donald Rumsfeld would say Known Knowns and Known Unknowns, whilst not knowing about Unknown Knowns and Unknown Unknowns.

    1. Dave 126 Silver badge

      > as Donald Rumsfeld would say Known Knowns and Known Unknowns, whilst not knowing about Unknown Knowns and Unknown Unknowns.

      Rumsfeld never mentioned 'Unknown Knowns', for reasons I hope you'll find obvious upon reflection! :) However, 'Unkown Knowns' did feature in the title of a documentary about him.

      I'm not commenting on the man's policies or politics, please note, just his correct use of English.

  5. 's water music
    Terminator

    Musk fear

    Perhaps the risk that Elon is afraid of is that of death by car crash because the AI controlling it is as limited Mr Pesenti says rather than ROTM.

    Icon for what will kill you because of a detection failure rather than by intent.

  6. Joanna Bryson

    Joanna Bryson is a woman

    I'd just like to point that out. Not all AI is written by men.

    1. Dave 126 Silver badge

      Re: Joanna Bryson is a woman

      Noted! It seems the commenter (or commentard, as we are all known here for reason you might now be able to guess) may have skim read the article. Either that, or the name 'Bryson' is so associated with an avuncular, self depreciating American travel writer that his puny human grey matter returned the wrong mental image. :)

      Curiously, the sex of the other two experts in the article are trickier (Zoubin Ghahramani) or impossible (Leslie Smith) from their forename alone.

  7. This post has been deleted by its author

    1. amanfromMars 1 Silver badge

      Re: Joanna Bryson is a woman

      Hi, Joanna,

      Thanks for stopping by.

      Joanna Bryson, a researcher interested in AI intelligence at the University of Bath, is often frustrated with the threat of “robo-killers”. ........ “AI is an incredibly powerful tool and is already changing our lives. It performs better than a human at many things, but it’s nothing like a human. And it’s not going to take over the world,”..

      Thrilling AI tool pushers are going to take over the world, and some would even suggest that it is a task already completed .... :-) AIMission Accomplished ....... and being reverse engineered and chronicled in sound bite sized pieces such as is this engaging thread for future lay maintenance with/from a contemporary running commentary for emerging human understanding of novel protocols/meme activations.

      Oh, and it might be more than just wise to not threaten to kill or obstruct an AI which can and/or has taken over the world and would be able to rearrange all things as it pleases and/or seeks to please that which it overtakes. The last thing that anyone would need is for it to go Right Royal Renegade Rogue and Real Crazy with incredibly powerful tools which can be weaponised against poorly performing humans.

  8. Mage Silver badge
    Headmaster

    Where is the AI?

    Spelling and Grammar checkers are not much better than late 1980s. Why can't I get an AI to proof read a computer stored text?

    Actually compilers are at a similar level to 1990s at initial parsing, and better than MS Office or LibreOffice for SIMPLE STUPID PUNCTUATION! Except we have all these stupid Web server files using maybe 6 languages and only a text editor that can spot mis-matched pairs of symbols and almost no source text checking before you try accessing it via a web client, because so much of it is scripting. HTML, ColdFusion, SQL, Javascript run on client, etc all in one file, maybe with php too.

    After 40 years of programming, courses on AI and studying it, it seems unrelated to Intelligence and more clever ways of making and accessing databases, plus a load of marketing.

    We don't even have a sensible definition of biological intelligence.

  9. JLV

    no robo killers, but...

    what about jobs?

    General-availability specialized AI has a pretty massive potential impact on jobs. Take our friendly auto-pilots for example. Once Tesla and the like have good-enough auto driving capability (i.e. once they achieve significantly less mortality risks than human drivers and once the legal framework - insurance and manufacturer liability - around it has matured), then what?

    Along with capability to down quite a few at the pub and have the car drive you back, expect less taxi drivers and less truck drivers.

    You can generalize that type of impact, in the 30-50 year time frame, to a whole lot of jobs. There will be jobs that will be pretty safe - masseurs, chefs, nurses, etc... There will be plenty of creative type jobs that will need top-flight humans to assist the AIs. There will be jobs that don't have the numbers to justify crafting AIs for them. There will be jobs where human dexterity will remain paramount.

    But it still leaves a lot of potential replacements.

    Or, to put it slightly differently - we ain't necessarily dreaming about benevolent AIs like in Banks' Culture. I agree with the article that self-awareness is still quite far off. If anything, the success of machine learning is probably slowing down research in "real AI".

    But Bank's "benevolent socialism" bit, where we all live a life of leisure enabled by our AI underlings? I don't think our societies are yet grasping the challenges that potentially quite limited job requirements could bring.

    See for example https://www.amazon.com/Race-Against-Machine-Accelerating-Productivity-ebook/dp/B005WTR4ZI

    I am quite aware of the lump-of-jobs fallacy trap. I mostly agree that it's a trap - even though modern research is starting to recognize that, in a given country, some displaced workers will be net losers, even as society as a whole benefits. But wholesale AI to replace mundane humans tasks is another kettle of fish.

    No one wants to go back to Victorian times and slave all day. Nor do I think AIs doing work is necessarily a bad thing. Or avoidable. But nothing tells me that Banks' rosy future of self-confident yet clever idlers is more likely to happen than just generalizing today's unpleasant impact of long-term joblessness to a large proportion of the population.

    Unless we start thinking about how to make it a positive transition. Signs ain't good - Western military planners are still stuck thinking pilots will be needed for another 20-30 years, despite the massive risks inherent in getting it wrong.

  10. John Smith 19 Gold badge
    Unhappy

    beware of the definition of "mundane tasks."

    Still waiting for my robot butler.

    Now that we know the actual size of the human brain, and it's processing power, we know how dumb all those 1960's "Computers taking over" story plots were.Taking over the world at 1MHz and

    with 2 1/4 MB of RAM (roughly the DEC 36 that ran PLANNER).

    We also now know we can build a server farm with at least 1 human processing throughput. It's 1 human, but it thinks several thousand times faster than that human. And it can be replicated fairly easily with enough money.

    The question then becomes is it "AI" or is it just a specialist program chomping on a big set of data or is it the class of problems it's solving (normally only viewed as solvable with "intelligence") that makes it "AI" ?

    Still waiting for the robot butler.

    1. DropBear

      Re: beware of the definition of "mundane tasks."

      "We also now know we can build a server farm with at least 1 human processing throughput. It's 1 human, but it thinks several thousand times faster than that human." [Citation needed]

      The thing about the human brain is not only that a) it contains a mind-boggling amount of processing elements and b) we have basically hardly any idea how it is actually structured (or how to even tackle describing structural complexity of that level) but also that c) _all_ of those processing elements work in parallel, while any parallelism in current silicon hardware compared to a human brain is so many orders of magnitude off that it's essentially zero.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like