Catch up on stories from the past week (and beyond) at the Slashdot story archive

 



Forgot your password?
typodupeerror
AI

Are Companies Overhyping AI? (hackaday.com) 179

When it comes to artificial intelligence, "companies have been overselling the concept and otherwise normal people are taking the bait," writes Hackaday: Not to pick on Amazon, but all of the home assistants like Alexa and Google Now tout themselves as AI. By the most classic definition, that's true. AI techniques include matching natural language to predefined templates. That's really all these devices are doing today. Granted the neural nets that allow for great speech recognition and reproduction are impressive. But they aren't true intelligence nor are they even necessarily direct analogs of a human brain... The danger is that people are now getting spun up that the robot revolution is right around the corner...

[N]othing in the state of the art of AI today is going to wake up and decide to kill the human masters. Despite appearances, the computers are not thinking. You might argue that neural networks could become big enough to emulate a brain. Maybe, but keep in mind that the brain has about 100 billion neurons and almost 10 to the 15th power interconnections. Worse still, there isn't a clear consensus that the neural net made up of the cells in your brain is actually what is responsible for conscious thought. There's some thought that the neurons are just control systems and the real thinking happens in a biological quantum computer... Besides, it seems to me if you build an electronic brain that works like a human brain, it is going to have all the problems a human brain has (years of teaching, distraction, mental illness, and a propensity for error).

Citing the dire predictions of Elon Musk and Bill Gates, the article argues that "We are a relatively small group of people who have a disproportionate influence on what our friends, families, and co-workers think... We need to spread some sense into the conversation."
This discussion has been archived. No new comments can be posted.

Are Companies Overhyping AI?

Comments Filter:
  • I think they're dumbing down the term just as how 'Hover Boards' were dumbed down from their original concept.

    I'd consider them more voice activated assistants from the consumer ends. From the machine learning part it's just heuristics on a pile of data to find patterns.
    • So use VI (Virtual Intelligence) for these lesser systems? Worked fine in Mass Effect.
    • AI has been overhyped since the computer in Willy Wonka refused to tell where the golden tickets were...

      Now that it has been monetized, the people selling it are going to hype it up like everything else that gets sold for profit.

      Also, of course, "Machine Intelligence" isn't the same thing that we consider human intelligence to be. And, no, they're not likely to go SkyNet on us (anytime soon), but the Flash Crash of 2010 was a small taste of how AI can, and does, affect our lives, and as AI gets more integr

    • What you call "True AI" is what Science Fiction Literature calls True AI. (What Fiction calls True) While in Computer Science AI is about solving problems much how an organism would, other then having all the steps hard coded in, it would have a "simpler" method where it can pick up and adapt to new forms of input and realize that it needs a different form of output.

      My AI Professor (Dr. David Anderson (A name to be a victim of AI for sure)) was studying Diagrammatic Reasoning. Where if a computer was presen

    • by elrous0 ( 869638 )

      Every time I see what passes for "AI" today, I think of that scene in Weird Science [wikipedia.org] were Wyatt shows Gary that the only woman his computer is capable of creating is a "5th grade slow learner, boring dipshit." Well, 30 years later and poor Alexa/Cortana/etc. still haven't even gotten close that low standard.

    • by Xest ( 935314 )

      That's a bit like saying we don't have true physics because we've not yet invented FTL space travel.

      The fact that something is understandable is neither here nor there, all AI requires is that something be sufficiently clever to be deemed a basic (even if entirely explainable) intelligence.

      AI is really just the branch of computer science aimed at producing algorithms that result in an outcome that is intelligent in appearance. Yes, the end goal is a real, actual consciousness that's as intelligent or more s

    • by gweihir ( 88907 )

      Indeed. Calling these "AI" is basically a marketing lie, nothing else.

  • by Anonymous Coward on Sunday September 24, 2017 @07:43AM (#55253689)

    This has nothing to do with Betteridge's Law. If that applied to any question, the answer to any Ask Slashdot question would also be no. That's absurd. This headline is asking your opinion of whether companies are overhyping AI. Betteridge's Law does not apply here.

    Ian Betteridge observed that sometimes journalists who hadn't adequately researched a story and couldn't confirm the story would still run with it. To avoid printing false statements, journalists would write their headlines as questions about the facts. The classic example was "Did Last.fm Just Hand Over User Listening Data To the RIAA?" where the headline is asking a question about the facts. The headline insinuates that the answer is yes, without directly saying so. The journalist doesn't have the evidence to be confident it happened but ran with the story anyway. It's poor journalism and basically a form of clickbait. That's where Betteridge's Law applies, where the question can be answered with 'no' instead of assuming that the answer should be yes. It's observing that the journalist isn't confident about what the facts are, so the reader shouldn't be, either. Betteridge's Law is a criticism of reporting unsubstantiated stories.

    The headline here isn't asking a question about the facts. Instead, it's asking a question of opinion, specifically whether businesses are overhyping AI. Betteridge's Law does not apply here. It does not apply in situations such as this story.

    If you're going to mention Betteridge's Law, please understand what it actually means. It doesn't apply to this headline or story. The headline is a question to solicit your opinions and encourage discussion.

  • Are they doing anything with AI today that couldn't have been done in the 80's? They're creating rules today, they could have created rules in the 80's. I think people are just running out of ideas more than anything.
    • by qbast ( 1265706 )
      Quite accurate image recognition for one. Example: https://azure.microsoft.com/en... [microsoft.com] .
      • But that's not due to AI. That is due to advancement in sensors. I asked what they are doing differently with AI itself.
    • Computers have won against humans at playing Chess and Go for example. The list of tasks that you can do now with a computer continues to increase. But is it an universal solution? No.

      • But a person following the same rules as a chess beating computer would still beat a human player, albeit 100billion times slower.
  • by Jim Sadler ( 3430529 ) on Sunday September 24, 2017 @07:49AM (#55253705)
    Defining thought and original thought is a very complex issue. But to give a very simple example a common chess machine can be overwhelming against human players. And yes the program functions with a list of rules and values. But the telling point is that machine may well play a unique, winning game. To my way of thinking that is original thought and intelligence. There are alos electronic circuits that have been created by computers that are totally inscrutable to humans.
    • Knowledge, intelligence, cognition, and thought are all distinct things. I studied AI as an undergraduate, rule-based expert systems to be precise. Those are very much considered a form of AI yet most people do not think "expert system" when they hear "AI".

      The examples you cite might or might not involve original thought, but then it might appear that they do to a human user.

      I think part of the confusion is that in the field, AI can mean lots of different things, while in popular culture people think of

    • That's baloney they had a team reprogramming the machine between matches.
  • they are.
  • by hey! ( 33014 ) on Sunday September 24, 2017 @07:55AM (#55253723) Homepage Journal

    Well I'll be damned.

  • Conscious thought doesn't have to come into play. A connected "AI" with bad data can cause damage as well
  • by fygment ( 444210 ) on Sunday September 24, 2017 @08:21AM (#55253777)

    Just wouldn't happen.
    Right?

  • Eliza: How do you feel about companies over hyping AI?
    ?
  • I would say the biggest issue is that there isn't an agreed upon definition of what constitutes "AI". To get an idea of this, technically from the field, OCR is "AI". Yet commonly I think you'd find people expecting some sort of robotic terminator-like lifeform that is capable of all things human, including emotions (if limited). When you have that sort of range... I mean I could make a case that genetically engineered oranges are "AI". They are computing sweeter flavor (and radioactive cancer).

    From my

    • I would say the biggest issue is that there isn't an agreed upon definition of what constitutes "AI".
      Actually there is.
      Visit an university, or their relevant web sites, and the definition(s) are right there.

  • by Hognoxious ( 631665 ) on Sunday September 24, 2017 @08:49AM (#55253851) Homepage Journal

    Not at all. No, no, no, no, no. Not even a little bit round the edges.

    Oh, alright then, yes.

  • by tomhath ( 637240 ) on Sunday September 24, 2017 @08:54AM (#55253869)
    Most of the AI hype is coming from people who are paid to write tech stuff. They write about whatever they think will get them page views; AI is a hot topic so AI it is.
  • Quantum handwaving (Score:5, Insightful)

    by Kongming ( 448396 ) on Sunday September 24, 2017 @09:06AM (#55253915)

    Worse still, there isn't a clear consensus that the neural net made up of the cells in your brain is actually what is responsible for conscious thought. There's some thought that the neurons are just control systems and the real thinking happens in a biological quantum computer...

    Penrose, is that you?

    Seriously, there is no evidence for any kind of "quantum consciousness", nor any convincing theory as to why a neural net would be insufficient to produce consciousness. I suspect that the main attraction of this idea is that it is a non-religious excuse for believing consciousness to be magical or special in some way.

    • by ceoyoyo ( 59147 )

      Can't be. Penrose dropped that hypothesis. It was a cool idea in the 80s, but not so much anymore.

    • by gweihir ( 88907 )

      No, actually there is zero evidence in the other direction too. Physicalism is just a fundamentalist quasi-religious belief, it is not grounded in fact. In fact, what you people would need to do is to stop claiming that neural nets are _sufficient_ to produce intelligence and consciousness, until you have some actual evidence for that. All you have at this time is unproven assumptions (a very beloved "proof" technique in religious circles) that basically say "everything is physical, hence consciousness and

    • Seriously, there is no evidence for any kind of "quantum consciousness", nor any convincing theory as to why a neural net would be insufficient to produce consciousness. I suspect that the main attraction of this idea is that it is a non-religious excuse for believing consciousness to be magical or special in some way.

      You said that using language. Language is the evidence.

      What is the sound of one hand clapping? Quantum handwaving.

    • There is some evidence that neurons are not important as people generally believe. https://www.scientificamerican... [scientificamerican.com]
  • A recent German article on a similar subject called out the current crop of AI and suggested to refer to it as "clever statistics" instead. IMHO this is spot on.
    • by ceoyoyo ( 59147 )

      I've also heard humans referred to as biased linear regressors. Also spot on.

    • by gweihir ( 88907 )

      I like that term. It accurately describes what is going in. Now, there still will be some stupid people claiming that human brains are doing nothing but statistics and hence intelligence and consciousness happen when you pile enough statistics on top of each other, but there is always an ample supply of smart-stupid people with selective blindness. At least psychology knows their number, namely that these people are not able to live with uncertainty, so they make up sophisticated-looking pseudo-explanation

  • by DontBeAMoran ( 4843879 ) on Sunday September 24, 2017 @09:19AM (#55253965)

    Is Slashdot overhyping headlines?

  • Remember how client-server computing was going to bring world peace?
  • Too much Hype.

    A deep neural network, a massive dataset brings you a statistical correlation where one is expected to be found and its called AI now?

    This is impressive in itself but even the futurists singularity proponents like Ray Kurzweil are not calling this 'AI' as 'The AI' for the singularity. Or true 'thinking' AI in the sense of human cognition.

    There is a massive gap in understanding the definition of AI, its a 'magic hat' term that is ambiguous and over-reaching. The progress should be appreciated f

  • The marketing department

  • AI has been overhyped since the late 60s, many times by people like Marvin Minsky, who ought to have known better.
    • by gweihir ( 88907 )

      Ah, Marvin "the idiot" Minsky. I am really glad he is not part of that conversation anymore. He has done immense damage to Science.

  • by Proudrooster ( 580120 ) on Sunday September 24, 2017 @10:27AM (#55254155) Homepage

    Hey Siri, "How many cylinders in a V6 engine?" .... let me search that for you. Seriously?
    Hey Siri, "How many doughnughts are in a dozen doougnughts?" let me search that for you.
    Hey Siri, "What is the nominal size of a 2X4 board?", it's 2x4=8
    Hey Siri, "What time is it on Mars?", I am sorry, I don't know where that is.

    So yeah, I am thinking AI is perfect.

    Seriously, there is no "I" in AI. There is no intelligence.

    • Those examples are why I call the technology behind Siri and Alexa "Artificial Stupidity".
  • "Are Companies Overhyping AI?"

    Not withstanding Betteridge's Law, the answer is "yes". Yes they are.

    Next clickbai- err, I mean "story", please.

  • Besides, it seems to me if you build an electronic brain that works like a human brain, it is going to have all the problems a human brain has (years of teaching, distraction, mental illness, and a propensity for error).

    If you created an electronic brain (of which no one is remotely close), the one advantage it would have is the ability to copy. That's the same advantage that Expert Systems have today. We don't have intelligent self driving cars but once we cover enough edge cases and the software controlling a self driving car becomes safer than the average driver then we can copy that to 100k other cars. We can also continue to improve it and then copy that improvement. The advantage that computers have is reliabili

    • It is artificial intelligence though it might be better if we used another word for artificial and called it "fake intelligence"

      They wanted to call it psuedointelligence, but it was determined that not enough people would know what that meant.

    • by gweihir ( 88907 )

      I Like "FI" (fake intelligence). Because that is what it is: It can fake being intelligent for a limited task. In a sense, it is really good automation. And such a thing is hugely useful, because we are not finding out that many tasks we thought require intelligence actually do not (like playing Go). Many of these tasks are accessible to fake intelligence.

      Of course, the other thing we find these days (even though many cling to a desperate belief it is otherwise) is that we do not even have a hint of general

      • we do not even have a hint of general intelligence in anything computers can do and we have looked really, really hard.

        That's because nobody is really pursuing general intelligence. I'm not sure we even have the technology yet to truly pursue general intelligence but I would likely start with studying the amoebas and then move on to ants. I read an article recently about how unlike the rest of your body, the cells in your brain actually all have different DNA. If this is true, then it means your brain is millions of evolving organisms all working together. You are basically emergent behavior of a large group of microorg

  • Yes!

    Yes they are!

  • This one:

    "Besides, it seems to me if you build an electronic brain that works like a human brain, it is going to have all the problems a human brain has (years of teaching, distraction, mental illness, and a propensity for error)."

    Is the single truest statement about AI I have read in a long time.

    I would also add in addition to list (teaching, distraction...)

    logical fallacies, including blind loyalty, confirmation bias, etc.
    quarrelsome (among themselves)
    greed

    and my personal favorite:

    lazyness

    • by gweihir ( 88907 )

      That one is older, but nonetheless quite true. It is conveniently ignored by the AI fanatics, in particular those that think simulating a human brain would create AI. If feasible, it would have exactly all these issues and be basically useless.

  • by XSportSeeker ( 4641865 ) on Sunday September 24, 2017 @02:59PM (#55255147)

    This particular battle of semantics has been going on for a while now, and much like previous battles (hoverboard, drones, HDR in 4K), it'll be won by advertisers who don't know better.

    The point is building interest in a generic marketing term even if it comes at the cost of the original meaning of the word. Scientific or technical terms (and in some cases, terms made up by sci-fi authors) have always been appropriated, it'll keep happening.

    But is AI being overhyped? Definitely. Because behind all the AI craze, the real interest for several companies is in user data collection which is becoming the new coin of the day. It is a very convenient way for tech companies to imply that there are some vague gains to be had using their products while not mentioning that they are harvesting your data or saying that they need to do it "because the AI needs it to work better".
    Notice how it's also super convenient for companies and services to use vague terms like that because they not only "fancy up" their products, it also serves as a convenient scapegoat when things go south (see how "algorithms" is losely employed by social media networks to put the blame on for mishaps).

    For those who didn't see the dimention of this overhyping just yet, here's a comprehensive list of a whole ton of products and services where the term is used, most of which have zero AI in it:
    https://medium.com/imlyra/a-li... [medium.com]
    Some of them barely have any intelligence on them at all.

  • Lets begin with the state of the art. The voice and face recognition technology is the same as what defeated human players in Go. While it is not yet the same kind of general intelligence as humans it proves that you don't need the same number of neurons and connections as a human to be very very intelligent in at least a narrow domain. They are true intelligence by any measure, just not as general as human intelligence. Furthermore human level intelligence is not required for machine intelligence to be a p

  • Despite appearances, the computers are not thinking. You might argue that neural networks could become big enough to emulate a brain. Maybe, but keep in mind that the brain has about 100 billion neurons and almost 10 to the 15th power interconnections. Worse still, there isn't a clear consensus that the neural net made up of the cells in your brain is actually what is responsible for conscious thought.

    This is very much correct. Much of what we call artificial intelligence today we could instead call functions of best fit. We emulate aspects of biology in systems and these aspects allow a pseudo-intelligent matching to occur. The matching function might be able to identify your face to unlock your phone or identify lingual patterns to generate language that a human speaker will feel is somewhat natural or identify what animal an image is of or even to beat human players at Jeopardy.

    This isn't consc

  • Its not really general purpose AI. However, deep learning sometimes performs better than the procedural or statistical algorithms it replaces. That aspect isnt hyping.

C++ is the best example of second-system effect since OS/360.

Working...