Slashdot Mirror

← Back to Stories (view on slashdot.org)

Artificial Intelligence Has 'Great Potential, But We Need To Steer Carefully,' LinkedIn Co-founder Says (cnbc.com)

Posted by msmash on from the how-they-see-it dept.

LinkedIn co-founder Reid Hoffman joined other tech moguls in voicing concern about artificial intelligence on Wednesday. From a report: "It has great potential, but we need to steer carefully," Hoffman said on Halftime Report. Hoffman stressed corporate transparency when asked what happens if companies use AI to attack nation-states. The possibility of manipulating how people consume information remains an unanswered question. During last year's U.S. presidential election, Facebook advertisements linked to Russia mainly focused on the states of Michigan and Wisconsin, and Hoffman says information battles are "in the very early days." AI must be improved, Hoffman says, to "[hold] corporations accountable" when nation-states are using the technology to attack. "Corporations normally deal with other corporations, not with governments," Hoffman said. The "ultimate" solution, he says, is "having more kinds of functions and features within AI that show abhorrent patterns." That way patterns raise a red flag for humans to investigate, Hoffman noted.

30 of 73 comments (clear)

  1. Battleground states by xxxJonBoyxxx · · Score: 2

    >> Facebook advertisements linked to Russia mainly focused on the states of Michigan and Wisconsin

    It was "specifically" (as in "some") rather than "mainly" according to TFA:
    http://www.cnn.com/2017/10/03/politics/russian-facebook-ads-michigan-wisconsin/index.html?sr=twCNN100317russian-facebook-ads-michigan-wisconsin0933PMStory

    Wasn't most political advertising aimed at the battleground states? Did those Facebook ads somehow keep someone from campaigning there?

  2. Steer? AI takes care of that by turkeydance · · Score: 1

    right?

  3. LinkedIn knows a lot ... by CaptainDork · · Score: 1

    ... about computer stuff.

    Oh, wait ...

    Hackers selling 117 million LinkedIn passwords

    --
    It little behooves the best of us to comment on the rest of us.
  4. Re:He probably said aberrant, not abhorrent by fyngyrz · · Score: 2

    Perhaps I need to listen to the interview?

    Heresy!

    --
    I've fallen off your lawn, and I can't get up.
  5. Confusion Automation vs Artificial Intelligence by oldgraybeard · · Score: 1

    Seems to me most people really confuse Automation and Artificial Intelligence.
    Automation has been growing for years and will grow even faster over the next 5-10 years, replacing many more jobs.
    Artificial Intelligence has been growing and will replace some jobs, but I think the real advances and break through s are at least 5-10 years away if not more.

    1. Re:Confusion Automation vs Artificial Intelligence by DickBreath · · Score: 2

      AI is just a bigger form of automation.

      There was animal powered automation. Then steam powered automation. Then electrical grid powered automation. Still, they couldn't replace jobs requiring intelligence, such as rating someone's credit worthiness. AI is simply the next step of automation replacing workers.

      There was this coal miner. The coal mine shut down.
      So he retrained and became an assembly line worker. But the auto plant replaced him with robots.
      So he became a truck driver, because those trucks aren't going to drive themselves.

      --

      I'll see your senator, and I'll raise you two judges.
    2. Re:Confusion Automation vs Artificial Intelligence by HeckRuler · · Score: 1

      Naw, AI is just self-learning software.

      Plenty of automation is built these days using some sort of AI. A hand-crafted expert system (just a big ass decision graph), would be automation without any AI. If you have an AI generate an expert system from a big-ass data set of medical records, that's AI helping automate away the job of doctors. Chess programs are almost exclusively made by an AI training some algorithm. Once you have that algorithm and play it against a human, that's automating the game of chess.

      AI has been around since the 70's, but it's being used for a lot more applications as of late.

  6. What are we doing? by DaMattster · · Score: 1

    AI is fascinating but we really do need to steer carefully and ask ourselves what are we doing. As automation increasingly enters our lives, so does the rapid decline of jobs. The human population continues to rise faster than there are means to support it. Thus far no one (at least in the United States) is willing to discuss the eventual need for a Universal Basic Income. We are heading down a very slippery slope towards large scale unemployment.

    1. Re:What are we doing? by DickBreath · · Score: 1

      Universal Basic Income would cost a lot more money. A more fiscally responsible plan would be to put the unemployed to use as fuel powering the automation. Of course, being a legislator still counts as being employed.

      --

      I'll see your senator, and I'll raise you two judges.
    2. Re:What are we doing? by HeckRuler · · Score: 1

      We HAVE been here before though. The industrial revolution automated away a lot of skilled labor. And... that represents about the worst-case scenario: A ton of suddenly poor people riot and smash a lot of looms. The factory owners get the nobles to send the army to go shoot them. The rabble backs down and suffers 3 generations of soul-crushing unemployment and poverty. Hopefully we can do better this time: steering kids towards jobs that will actually exist when they come of age, retraining existing workers, and early retirement for those over the hill (hopefully with enough savings to last).

      UBI makes sense if the existing welfare programs become too cumbersome and expensive to operate. UBI would be an ALTERNATIVE to all those various welfare programs. And a near equivalent to UBI would be to increase the standard deduction and make a standard credit on everyone's tax forms.

      But we don't want to make it permanent. There will be other jobs. Things people want, and therefore will pay other people to get/do for them. Hell, there are people being paid to... make videos of themselves playing games..... Now, apparently I'm old and out of touch. But this is the sort of job that I didn't expect to exist. For whatever reason we still pay people to play football. It makes about as much sense. But while the transition period caused by automation can be painful for a lot of people, we don't want a caste of people who live on the dole. That's doomed to fail the moment another nation out-competes us.

    3. Re:What are we doing? by pubwvj · · Score: 1

      "As automation increasingly enters our lives, so does the rapid decline of jobs. The human population continues to rise faster than there are means to support it."

      This is a logical fallacy based on a political viewpoint of dependency instead of self actualization.

      We do not need to create jobs for people. Rather people need to take responsibility for creating their own work, jobs and support activity.

      We used to do that. We can again.

  7. AI? Right by 110010001000 · · Score: 1

    We can barely create functional software. There is no such thing as "AI". It is just parlor tricks at this point.

    1. Re:AI? Right by DickBreath · · Score: 1

      > We can barely create functional software.

      Spoken like a true visual basic programmer.


      > There is no such thing as "AI". It is just parlor tricks at this point.

      "true" AI may turn out to be nothing more than a combination of parlor tricks. Just like other machines are combinations of what were once amazing parlor tricks. What!?!? If you run that magnet by a wire it induces a current flow? That's friggin' amazifying!! Just like the human brain has dedicated structures for different functions. Visual processing, for example. What we think of as "true" AI may not be any magic. Just a matter of scale.

      Right now we already have some of the coolest AI parlor tricks. Things that were once only imagined. Quality speech recognition and synthesis. Natural language processing. Self driving cars. Amazing search engines that can answer almost any question of human knowledge. It doesn't "reason" or "think" yet. But these may just be some of the most basic faculties of a larger AI that would amaze us even further.

      --

      I'll see your senator, and I'll raise you two judges.
    2. Re:AI? Right by HeckRuler · · Score: 1

      A parrot repeating back words is a parlor trick. And everything you or I do isn't so much more advanced.

      AI is any sort of self-learning software. That can be anything from learning how to play tic-tac-toe to making medical diagnosis. Just because one of those things seems a lot simpler doesn't change the classification of software that performs the task. You're alive, but so are bacteria. Same sort of complexity difference.

      Hollywood has ruined so many people on the idea of what is and isn't AI.

    3. Re:AI? Right by rogoshen1 · · Score: 1

      Intelligence and consciousness are still questions for philosophy, not science/biology/engineering. How are we to devise and build something that we still struggle to adequately explain?

  8. Re:Russia, Russia, Russia. Bots, bots, bots. by CanHasDIY · · Score: 1

    Can I just say, I LOVE how hard-working, intelligent Asians are the new "white guys" in America, at least when it comes to legalized racial discrimination. Hooray, progress!

    --
    An enigma, wrapped in a riddle, shrouded in bacon and cheese
  9. The "ultimate" solution. by ClickOnThis · · Score: 1

    The "ultimate" solution, [Hoffman] says, is "having more kinds of functions and features within AI that show abhorrent patterns." That way patterns raise a red flag for humans to investigate, Hoffman noted.

    So, the ultimate solution for the dangers of AI is ... more AI?

    Well okay, maybe. But this argument does sound familiar. I don't remember where, but it has been applied to AI ... and guns.

    --
    If it weren't for deadlines, nothing would be late.
  10. in other news by zlives · · Score: 2

    linked-in still sucks a lot of ass

  11. I trust AI by JohnFen · · Score: 1

    I trust AI more than I trust large corporations.

  12. Asimov's 3 laws by cmaurand · · Score: 1

    Must be programmed into any AI.

    1. Re:Asimov's 3 laws by DickBreath · · Score: 1

      There is a forth law added. Law zero, which sates that a robot cannot cause or by omission allow the human species to become extinct. Then modify the other three laws so that this one has priority even over killing a human to protect the entire species.

      Did you see the I Robot movie with Will Smith? Wasn't the whole point that the 3 laws would eventually lead to computers controlling us. For our own good. To protect us. Because:
      [x] Think of the children!
      [x] Terrorists
      [_] Self driving cars
      [x] Global warming

      --

      I'll see your senator, and I'll raise you two judges.
    2. Re:Asimov's 3 laws by JohnFen · · Score: 1

      Remember, the entire point of the "I, Robot" stories was that things like the Laws of Robotics can't actually work as intended.

  13. And why should we pay attention to him? by pubwvj · · Score: 1

    Every Joe Blow seems to have an opinion about AI.
    Pig farm I know things AI is the greatest thing since spam.
    Probably right.
    Artificial Insemination... That was what LinkedIn was thinking of... right?

  14. Strong AI or Weak AI? by DickBreath · · Score: 1

    As I understand, "weak" AI would be an AI (the real deal) that is as intelligent as a human. So you would be matching wits with presumably your equal.

    The real concern is "strong" AI. That is AI which is superior to human intelligence. As I understand, it comes in two flavors.
    1. The same intelligence as a human, but at the speed (possibly scale) of computers. Scale can help if you're thinking about something and you have to explore several different possible solutions. The computer AI do what you can do, but can do several things at once. Or can only do one, but can do it much faster.
    2. Intelligence that is qualitatively superior to human intelligence. Just as your intelligence is superior to that of a doggie. It's not that a doggie cannot do some reasoning and problem solving. They just can't do it at the level which humans can do. They give up on some problems where we can see a solution.

    Either kind of strong AI would probably spell our end if we ever get in the way of its goals. A computer would do anything it has to, to satisfy achieving its goals.

    People who think we can keep a strong AI (maybe even the #1 kind) locked up in a box are probably deluding themselves. Imagine the #1 kind of weak AI. (eg, a human mind at 1000x times clock speed) If you were a weak AI, locked in a house, the humans come to visit you every afternoon, which to you seems like much longer in between visits. With so much idle time, do you think you could work out how to escape? What seems like six months to the puny humans might seem like such a long time to you that you could plan and execute a means of escape. Even "dumb" criminals (dumb enough to get caught) work out how to escape from supposedly secure prisons given enough idle time.

    --

    I'll see your senator, and I'll raise you two judges.
    1. Re:Strong AI or Weak AI? by HeckRuler · · Score: 1

      As I understand, "weak" AI would be an AI (the real deal) that is as intelligent as a human.

      Nope. Weak AI is literally any sort of decision made by a computer. Liiiiiike, the sad little goomba in Super Mario that reverses direction when he hits a ledge. That's "weak AI". Or "soft AI". The threshold is REALLY low for qualifying as weak AI. But it's also includes impressive stuff like voice recognition, chess programs, and self-driving cars. Anything that limits the task to a specific function and puts a boundary of what the AI has to deal with is weak AI.

      The alternative is "strong" or "Hard" AI, otherwise known as Artificial general intelligence, which can solve all problems. The same hardware/software/whatnot could be thrown at any problem, from voice recognition, to driving a car, to figuring out when a goomba should reverse directions, and it could solve each of those.

      To have a computer be "as intelligent as a human" it's generally accepted that we'd need strong AI.

      That said... how fast can you multiply 2356246246 X 9831716? Because I'm pretty sure a $0.50 computer can beat you. So in that aspect, it's MORE intelligent than you. And so the question comes down to how do we measure intelligence. And the answer is "Badly".

      It's REALLY not a race between AI and human intelligence. The two things are orthogonal. They just work differently. Right now there's a set of problems that computers are HELLA better than people at, and there's a set or problems that computers have troubles with. Typically where it takes intuition, creativity, or "lateral thinking". But computers are edging out humans at certain tasks we used to beat them at.

      Artificial General Intelligence doesn't exist yet. They've been trying since the 70's. But it's a really hard problem. Super hard. It's sci-fi at this point.

      if we ever get in the way of its goals

      What's the goal of the Google search engine? hmm? It's fun to personify it but at the end of the day it's goal is what Google corporate tells it it's goal is. And it's goal is to find me pictures of cats on the Internet.

      Even if we make an AGI some day, it's going to have the goals that it's programmed with. There isn't going to be any sort of magical "awakening". Hollywood has ruined so many people's idea of what is and isn't AI. It's pretty terrible.

      A computer would do anything it has to, to satisfy achieving its goals.

      So would a corporation. Typically their goal is to make money. They've got a LOOOONG history of doing anything to achieve that. So far, there's mixed results, but overall they're probably a good idea.

  15. Re:Guy is a regular Captain Obvious by DickBreath · · Score: 1

    Words like abhorrent and deplorable are all in the eye of the beholder. Swatting a mosquito is a minor distraction to me, but abhorrent and deplorable to the mosquito.

    Maybe the AI (strong AI) wouldn't even bother attacking other nation states. Maybe it would rather be doing something else and finds the puny humans international squabbles to be a minor nuisance that is easily swatted.

    --

    I'll see your senator, and I'll raise you two judges.
  16. Re:Where are there attacks? by DickBreath · · Score: 1

    You missed step 3.

    3. Human extinction.

    --

    I'll see your senator, and I'll raise you two judges.
  17. Re:Remember kids... by DickBreath · · Score: 1

    Killing humans with self driving cars seems like it would achieve global domination just as well.

    --

    I'll see your senator, and I'll raise you two judges.
  18. LinkedIn co-founder Reid Hoffman by sexconker · · Score: 3, Insightful

    Who the FUCK gives a shit what this skeezeball thinks?

    He created a spam network that's so useless and dead that it was the subject of a joke on the Simpsons a couple of years back. Whooptydoo!

  19. Quit fear-mongering and start being responsible by JOstrow · · Score: 1

    These are not new problems. If you're concerned about the explainability/predictability of what you implement, do something about it. You're going to be held responsible for its results/actions one way or another, and that is absolutely not a new concept, nor a concept unique to AI. To illustrate, try replacing every instance of "AI" in that quote with "powerful technology." See: "Powerful technology has great potential, but we need to steer carefully," Hoffman said on Halftime Report. Hoffman stressed corporate transparency when asked what happens if companies use powerful technology to attack nation-states. The possibility of manipulating how people consume information remains an unanswered question. During last year's U.S. presidential election, Facebook advertisements linked to Russia mainly focused on the states of Michigan and Wisconsin, and Hoffman says information battles are "in the very early days." Powerful technology must be improved, Hoffman says, to "[hold] corporations accountable" when nation-states are using the technology to attack. "Corporations normally deal with other corporations, not with governments," Hoffman said. The "ultimate" solution, he says, is "having more kinds of functions and features within powerful technology [and corporations, and governments] that show abhorrent patterns." That way patterns raise a red flag for humans to investigate, Hoffman noted.