back to article Turns out humans are leading AI systems astray because we can't agree on labeling

Top datasets used to train AI models and benchmark how the technology has progressed over time are riddled with labeling errors, a study shows. Data is a vital resource in teaching machines how to complete specific tasks, whether that's identifying different species of plants or automatically generating captions. Most neural …

  1. Paul Crawford Silver badge

    What would happen if a self-driving car is trained on a dataset with frequent label errors that mislabel a three-way intersection as a four-way intersection? The answer: it might learn to drive off the road when it encounters three-way intersections.

    Clearly that is not intelligence at all. You have faulty software because you did not have a complete grasp of the programming of it. Some might even say a negligent approach as you assumed the Mechanical Turks provided valid data, and you did not verify it yourself.

    1. Version 1.0 Silver badge

      Mistaking a a three-way intersection for a four-way intersection sounds like drunk driving - perhaps we need to ban the coders writing the AI software, and those selecting the learning images, from drinking?

      1. EarthDog

        No, it has to be the companies.

        Programmers aren't professionals so there is no license to pull. Unlike engineers, doctors, nurses, etc. As such they have no liability. But the companies do.

        1. Anonymous Coward
          Anonymous Coward

          Re: No, it has to be the companies.

          Are you an Uber driver? Drivers are not workers so no health care insurance is needed or vacation pay, and no Uber liability when accidents happen.

      2. katrinab Silver badge
        Flame

        The people doing the training are the people doing the annoying recaptcha challenges.

        1. ortunk

          I try to sneak in two wrong amswers to every challenge

  2. Pascal Monett Silver badge
    FAIL

    An enlightened explanation of how we got ourselves into this mess

    A remuneration scheme where, if you don't choose "bucket", you don't get paid.

    And people think that "AI" is actually worth something.

    1. Chris G

      Re: An enlightened explanation of how we got ourselves into this mess

      Based on the examples given in the article, there are no labeling errors, the fault lies with the simplistic software and it's lack of ability to encompass enough parameters. A human would have little problem in making sens of a a 'bucket of baseballs' encompassing both a number of baseballs contained in a bucket or a bucket containing a number of baseballs.

      The errors in these cases at least lie with the programming.

      1. Richard 12 Silver badge

        Re: An enlightened explanation of how we got ourselves into this mess

        Well, the specification.

        Expecting every image to have one, unambiguous label is just plain stupid.

        1. sketharaman

          An enlightened explanation of how we got ourselves into this mess

          +1. Precisely.

        2. John Brown (no body) Silver badge

          Re: An enlightened explanation of how we got ourselves into this mess

          "Expecting every image to have one, unambiguous label is just plain stupid."

          I came here to say exactly the same thing!

        3. Warm Braw

          Re: An enlightened explanation of how we got ourselves into this mess

          There was a recent article in which the 'AI' model recognised the label as qualitatively the same as the thing being labelled.

          Perhaps these researchers could do with the odd philosopher on their teams.

          1. Tom 7

            Re: An enlightened explanation of how we got ourselves into this mess

            Or someone a bit older and more cynical than them, someone with large complicated systems experience? I think AI is going to be fantastic, once greedy 'entrepreneurs' and technically blind managers are removed from the equation. You cant see the wood for the trees with a cash flow obscuring your vision.

            We're no where near the tipping point in AI becoming really useful but we may get there if the above mentioned dont fuck it up with greed and exploitation. It can do some fun things and I have no doubt given time to develop properly will be really useful for humanity but too many greedy bastards or premature cockups may see the end of it in commerce but you can bet someone worse will keep plugging a way at it.

            1. Ken Moorhouse Silver badge

              Re: once greedy 'entrepreneurs' and technically blind managers are removed from the equation

              Actually a lot of technological progress has been made on the back of mistakes (e.g., stainless steel), or "nerdy" types challenging themselves to do something that initially has little or no commercial value. The greedy entrepreneurial greed element emerges later on.

              The inherent problem with AI is that its "brains" have to be programmed in by a human, and therefore have the same limitations of "intelligence" that a human possesses. The only way I can see AI progressing is to have a homogenous fabric of genetic-like mechanisms that fight each other for survival, in a similar way to that used by the body to protect against unknown diseases (T-cells, antibodies, etc.). The problem still is that rules need to be imbued into the fabric to make anything happen. Unless the designers faithfully mimic genetic interactions completely and faithfully there will always be a human bias to the outcome of the model. The thing about genetics is that there is a great deal of latency in its mode of operation to enable evolution to take place. Any temptation to strip out those imperfections is to be resisted - because that constitutes a human generated rule.

              If this conundrum can be solved the "programmer" then has more of an "animal husbandry" role to play in the proceedings, throwing in parameters (such as the starting point for Conway's "Life") and see what emerges at the far end.

        4. David Glasgow

          Re: An enlightened explanation of how we got ourselves into this mess

          Expecting every image to have one, unambiguous label is just plain stupid.?

          It depends entirely on what your wife sent you to look for in the basement.

    2. Michael Wojcik Silver badge

      Re: An enlightened explanation of how we got ourselves into this mess

      So twenty people here (so far) think machine learning is defined by a labeling procedure with perverse incentives.

      I think we can discount the possibility of human intelligence, too. Certainly it's evidence for a paucity in critical thinking.

  3. Anonymous Coward
    Anonymous Coward

    Google Captcha

    I was asked to click on squares with Bicycle but show a woman with a shopping trolley. It wouldn't let me continue until I'd said the shopping trolley was a bicycle.

    1. Yet Another Anonymous coward Silver badge

      Re: Google Captcha

      Captcha - guessing what other users outside America think a parking meter looks like, multiplied by what a bunch of Uzbecks were told to click on for $1/day

      1. TDog

        Re: Google Captcha

        And then re-enforcing their model by having failures leading to you being classified as an AI:

        "Look - we are so good we caught X^X thousands of AI's yesterday alone".

        Or it might just be Chinese cultural imperialism - by reprogramming Uzbeks (and every one else) to meet the Real Middle Kingdom Perception of Reality (tm) as promulgated by the Real Middle Kingdom (please ignore those imposters in Taiwan and understand that Tibet was always part of the Real Middle Kingdom according to the Real Middle Kingdom protocols), and thus for a small loss in unnecessary population that chose voluntarily to not meet Real Middle Kingdom standards of integrity and mental coherence mediated by their own choices the PRC (interestingly that was also a set of USA walkie talkies which were generally described as pricks) is subtlety and quietly changing the perceptions of their future greatest allies. Carries on like this till the next Peoples Congress changes the rules (and that will be just after the Greek kalends)

        1. Anonymous Coward
          Anonymous Coward

          @TDog - Re: Google Captcha

          You should get some fresh air from time to time.

          1. EarthDog

            Re: @TDog - Google Captcha

            it was perfectly coherent.

            1. John Brown (no body) Silver badge

              Re: @TDog - Google Captcha

              Well, I guess all you dogs stick together. I thought it was AmanFromMars1 and had to check the poster name.

            2. jake Silver badge

              Re: @TDog - Google Captcha

              Nobody said anything about coherency.

            3. Michael Wojcik Silver badge

              Re: @TDog - Google Captcha

              "Coherent" is debatable, but it certainly wasn't well-written. Some authors (Faulkner, say) can get away with stringing one clause after another together into a breathless mega-sentence. TDog does not appear to rise to this level of prose skill.

              1. amanfromMars 1 Silver badge

                Re: Horses for Courses

                "Coherent" is debatable, but it certainly wasn't well-written. Some authors (Faulkner, say) can get away with stringing one clause after another together into a breathless mega-sentence. TDog does not appear to rise to this level of prose skill. ..... Michael Wojcik

                Would you find it strange to discover, Michael Wojcik, that whereas you shared that it certainly wasn't well-written, a few others who may indeed number many, find it to be excellently crafted with a level of prose skill it is a joy to read exists elsewhere too to highlight and share the views they witness there, here with us on El Reg.

                I ponder and wonder on whether such warrants a red flag warning and be classified as COSMIC Top Secret Pornographic Steganography given how easily it can be intelligently designed to totally deprave and sublimely corrupt. The perversion in doing that though, would be that it would make it even more attractive and universally popular making its pictures and products impossible to unilaterally command and exclusively control.

                1. jake Silver badge

                  Re: Horses for Courses

                  There is no debate about it at all! Coherent was extremely well written!

                  But then I have a thing for assembly language ...

    2. Anonymous Coward
      Anonymous Coward

      Re: Google Captcha

      And? By doing so, you just helped us crash SkyNet.

      On the other hand, we just need more lonly, bored, honest people...like I used to be. I used to do Captcha's for fun from 9pm until 3am every day(many years ago) . The more I did, the harder they became.

    3. DS999 Silver badge
      Mushroom

      This is all my fault

      I hate Google captcha with a passion, so when I'm asked to do it I deliberately spend a minute getting it wrong. If I REALLY want to continue on with the site that is requiring it I might then reluctantly decide to get it right, but usually I'll just abandon it after that.

      If enough people would deliberately pollute Google's results like that they would drop it, and stop trying to force people into teaching their AI for free. If their AI ends up not being able to tell a car from a bicycle my job is done!

      1. Cederic Silver badge

        Re: This is all my fault

        With you on that.

        Maybe they're not testing image recognition but instead training AI to identify when people are intentionally trying to trick AI. Seems about the only actual use they'll get from it.

      2. Tom 7

        Re: This is all my fault

        I've found quite a few that if I hadn't spent a couple of years in the US I would have failed due to them using US english that most of the UK over the age of 30 would probably not have a clue what they meant.

    4. Michael Wojcik Silver badge

      Re: Google Captcha

      Oh, ReCaptcha is complete crap. It exists only because Google's found a way to get lazy people who run sites like StackOverflow to get them free labor. It's one of the reasons I quit contributing to SO. (Another is the fact that the site is no longer usable unless you enable scripting, thanks to their use of crap from cookielaw.)

      1. Anonymous Coward
        Anonymous Coward

        Re: Google Captcha

        Also the irony that the crap from cookielaw crap "solution" to a privacy problem actually results in StackExchange telling yet another third-party site exactly which websites you are visiting. Duh!

        I have half a mind to whinge about it on the meta site to see if they'll do anything about it...

    5. Stuart Castle Silver badge

      Re: Google Captcha

      When responding to these captcha things, I regularly get asked to identify all the “Crosswalks”. Trouble is, being English, I don’t call them “Crosswalks”. I call them “Crossings”, sometimes adding the word “Zebra” before.

      This applies to a lot of sites, even uk ones.

      1. Graham Dawson Silver badge

        Re: Google Captcha

        And then it forces you to identify on-road warnings (like SLOW) as a cross-walk and fails if you don't comply. It's a weirdly demoralising thing to experience.

      2. wegie

        Re: Google Captcha

        @Stuart Castle "...Trouble is, being English, I don’t call them “Crosswalks”

        Not just "crosswalks". Now identify all the "fire hydrants"...um, well, there aren't any. There's a load of streetside hose connections of some kind, but no rectangular iron plates in the road or pavement labelled "FH". The complete cultural imperialism and cross-culture blindness is breathtaking. Just imagine a car trained on that on UK roads.

        1. Ken Moorhouse Silver badge

          Re: but no rectangular iron plates in the road or pavement labelled "FH"

          I don't think it registers with us Brits as much as in the States. Yes, we do have them and I remember having it explained to me as a kid what the two numbers on the yellow H sign stand for, but they are just not part of our everyday awareness, as they presumably are Stateside.

          https://commons.wikimedia.org/wiki/Category:Fire_hydrants_in_the_United_Kingdom

          1. Terry 6 Silver badge

            Re: but no rectangular iron plates in the road or pavement labelled "FH"

            Aren't there laws (state or federal) about parking in front of those big kerbside hydrants. They do seem to feature in films ("movies" as they call them there) quite a lot. People crashing in to them, making fountains with them and so on.

            In UK and other places they're just tucked away below ground.

        2. Blofeld's Cat
          Devil

          Re: Google Captcha

          "... Just imagine a car trained on that on UK roads ..."

          "On the left? What do you mean 'they drive on the left'?"

        3. Tom 7

          Re: Google Captcha

          When you get one that asks you to click all buses and there's not a circuit board or chassis shown!

          1. Rob Daglish

            Re: Google Captcha

            The one I particularly dislike is "Click on all squares containing traffic lights" - In the UK, these tend to be a pole with a cluster of lights at the top, but most of the google images are of the same light clusters cantilevered across roads or hanging from cables strung over the road... OK, it's easy enough to recognise the light cluster, but which squares do I click - only the ones with the light cluster, or do all of the ones with the cantilever/support poles/wires get counted as part of the traffic light?

    6. xyz Silver badge

      Re: Google Captcha

      Makes me want to say, stop the world, I want to get off

  4. jake Silver badge

    And of course people are bloody minded ...

    ... I know people who intentionally mislabel pictures that might be slurped, just to bugger up the data. Some because they are horrified by ihe concept of AI, some because they hate big business, some just for the lulz, and some for various other reasons, or combinations of reasons.

    Computers sometimes make mistakes. To really bugger it up requires a human.

    1. handle handle

      Re: And of course people are bloody minded ...

      Aye, but computers can make them a thousands times faster.

      1. EarthDog

        Re: And of course people are bloody minded ...

        And pollute a huge number of data streams

        1. jake Silver badge

          Re: And of course people are bloody minded ...

          Yes, computers are a force multiplier.

          (For clarity, please note that that word is not "multiplayer".)

          1. John Brown (no body) Silver badge

            Re: And of course people are bloody minded ...

            "Yes, computers are a force multiplier."

            Give me a lever long enough and a fulcrum on which to place it, and I shall move the world. - Archimedes

            I don't think anyone could disagree that computers have "moved the world" :-)

          2. Coen Dijkgraaf

            Re: And of course people are bloody minded ...

            > Yes, computers are a force multiplier.

            “To err is human but to really foul things up requires a computer.” —Paul Ehrlich , American scientist (1926- ).

            1. jake Silver badge
              Pint

              Re: And of course people are bloody minded ...

              When you have to explain the joke ... Ah, fageddaboudit.

              Have a beer :-)

      2. jake Silver badge

        Re: And of course people are bloody minded ...

        "Aye, but computers can make them a thousands times faster."

        But only if told to do so by a human. Computers can't think for themselves. They are not intelligent (Sorry, amfM ... but deep down even you know that this is true.)

        1. Tom 7

          Re: And of course people are bloody minded ...

          Most humans aren't! I think computers can be Intelligent, its only humans stopping them at the moment, much as they stop other humans.

    2. DS999 Silver badge

      Re: And of course people are bloody minded ...

      As I just confessed above before seeing this post lol!

      1. jake Silver badge
        Pint

        Re: And of course people are bloody minded ...

        :-)

        It's Friday. Have a beer.

    3. Tom 7

      Re: And of course people are bloody minded ...

      Asimov had this covered - his computers took this into account. Given left pondians still have trouble with UK sarc I guess he was a bit ahead of his time or perhaps we need a new Intelligence definition.

  5. ecofeco Silver badge

    I think I see the problem

    This work is often outsourced work to services like Amazon Mechanical Turk, where workers are paid the square root of sod all to sift through the data piece by piece,

    You get what you pay for. GIGO!

    1. doublelayer Silver badge

      Re: I think I see the problem

      Yes. Mechanical Turk pays the participants so little that they have a lot of data pollution problems. Also, they have the problem that the people doing the work are either bored people who will give up in fifteen minutes or people who really don't have better ways to get money, so you can't expect consistency or strict attention to detail.

      It's along the lines of all those studies they do at universities where students are paid to participate in research with an amount of money which could be used in a vending machine in 1995. Especially the economics studies which effectively boil down to "Would you take this action if we cut your meaningless money to even more meaningless money?". I participated in a few research programs while studying, but always because I was bored and didn't mind wasting a few minutes. I did actual work to earn money.

      1. Terry 6 Silver badge

        Re: I think I see the problem

        Also, they have the problem that the people doing the work are either bored people who will give up in fifteen minutes or people who really don't have better ways to get money, so you can't expect consistency or strict attention to detail.

        Long before the computer age I had a job with a mail order company.Our work was filng little slips of paper - coupons or something- with people's name and address they'd sent in. These had to be packed really tightly into stiff plastic wallets in alphabetical order. It was tedious and painful, the plastic would cut into your fingers and forcing the paper apart to insert the new slip was really difficult. None of us had been there long. And it was obvious why. The wallets were a mess. Sections would be in order, then they'd be random. As a new bunch of underlings were recruited, got bored and were fired. As we were.

        The crap pay and the mean fisted decision to force as many slips as possible into the envelopes as tightly as possible meant that the filing system was close to useless..

        The spirit of those days (early 80s?- the Woolworth's fire happened across the road while I was there) apparently lives on. And since there is a phrase "Spoiling the ship for a happ'orth of tar" it seems to go back a long way before that too.

      2. Tom 7

        Re: I think I see the problem

        Economic studies in universities are generally there to confirm a certain view of economics and only that view is correct. Its a view that wont pay for the truth.

        1. doublelayer Silver badge

          Re: I think I see the problem

          If that's the case, they're not very good at their job. Most of the economics studies involving meaningless money contradict many other theories. Behavioral economics really likes these. Whether that's because the previous theories were wrong (probably), because people act different when they actually care about incentives (probably) or because the researcher is deliberately messing with the results (probably not), they don't tend to be blatantly confirmatory.

    2. H in The Hague

      Re: I think I see the problem

      "You get what you pay for. GIGO!"

      Yup.

      There's currently a project commissioning human translators to translate a fairly large amount of text, for use as input for a Machine Translation system. Not a bad idea, but there are some issues.

      Offering 1/2 to 1/3 of the going rate probably doesn't help attracting competent translators. Asking the translators to provide two alternative translations isn't a bad idea. However, that fails when you have to translate the Dutch 'Het gras is groen' into English, as 'The grass is green' is essentially the only sensible translation. An alternative translation would simply add misleading input. Finally, all the sentences to be translated for this project are completely unconnected from each other. That's a major issue, as anyone who's done more than a day or two of translating will tell you that 'context is everything'.

  6. Blofeld's Cat
    Coat

    So ...

    Everything is fine until your self-driving car spots someone carrying a basket of balls across a five-way intersection ...

    1. jake Silver badge

      Re: So ...

      ... and has to pick between running over them, or the crocodile in the next lane, doing a fair approximation of a lightbulb.

  7. CrackedNoggin Bronze badge

    "One picture is worth 1000 words" - not just ONE WORD dummies!

    FAIL

  8. T. F. M. Reader

    Does anyone know...

    ...what happens to a British-trained AI-driven car when it crosses the Channel? And what happens when it crosses back a month later?

    Just curious because it's not always trivial for humans.

    1. sad_loser

      Re: Does anyone know...

      GPS means it is location-aware, so just switches to using the horn instead of indicating. Problem solved.

      1. jake Silver badge

        Re: Does anyone know...

        Would it think it was tomorrow, moving East of Greenwich?

        (I've seen worse programming errors ... )

        1. Primus Secundus Tertius

          Re: Does anyone know...

          I once saw an aeroplane system 'leap' from Cologne to Leipzig. A bit embarrassing for an RAF plane to suddenly be over East Germany.

      2. katrinab Silver badge
        Unhappy

        Re: Does anyone know...

        But would the British-trained car have microphones to hear the horn blasts emitted by others, and know how to interpret them?

        1. Rich 11

          Re: Does anyone know...

          And would it have two fingers to stick up in response?

    2. katrinab Silver badge
      Meh

      Re: Does anyone know...

      Or even if it crosses the border into Scotland?

      In Scotland you sometimes see a (70) speed limit sign in situations where you would see a ( / ) sign in England. But ( / ) signs do exist in Scotland.

      1. Dave559 Silver badge

        Re: Does anyone know...

        Different Road Traffic Acts in England than in Scotland. As far as I know, "Special Roads" in Scotland (namely, motorways and some others under similar special legislation (with restrictions on what sort of vehicles can use them) such as the Edinburgh city bypass) have to show the "70" sign instead of the "national speed limit applies" sign. Exactly why this bit of signage hyper-precision exists, I don't know, but the folks at SABRE and the like probably do…

  9. Gomez Adams

    I am gob-smacked that labellers are allowed to know what other labellers are identifying objects as while doing their job! :o

    1. Doctor Syntax Silver badge

      Maybe they don't. Maybe they just click on something until its accepted.

    2. jake Silver badge

      Why can't people ...

      ... stop putting labels on things and just accept them as they are?

      1. Terry 6 Silver badge

        Re: Why can't people ...

        Because the labels are how we a) communicate shared concepts and b) store them for recollection.

        Think compression alogorithms

  10. analyzer

    AI in everything is the problem

    Current AI is just pattern recognition on steroids, there is still no real intelligence in any AI system. The single word tagging of pictures has been known to be suboptimal for a long time now and yet it is still done. If there was any of the I in AI then multiple word tags should not be an issue as it should be capable of recognition of multiple items in one picture.

    Additionally, and I can't find the fine article that was in this exalted place, you can carry a tag around with you and be misidentified as the tag because the AI 'learns' the tag as well as the picture.

    Certainly artificial, definitely not yet intelligent.

    1. John Brown (no body) Silver badge
      Thumb Up

      Re: AI in everything is the problem

      I think you might need an apple iPad

    2. veti Silver badge

      Re: AI in everything is the problem

      Well, if only we could agree on a (non-circular) definition of "intelligence", that would be a step in the right direction. Until then, I don't see how we can hope to get anywhere.

      Notice how no-one talks about the Turing test any more? That's because it was passed, and so everyone promptly decided "oh no, that's not intelligence after all". As long as we're allowed to keep moving the goalposts like that, they're not going to make it.

      1. Ken Hagan Gold badge

        Re: AI in everything is the problem

        "Notice how no-one talks about the Turing test any more? That's because it was passed, ..."

        It was? When? How? Who? I've never seen any computer system get anywhere close. Come to that, I've seen *people* fail it (usually working in institutions as customer-facing staff and following rule sets, to be fair). Or did they re-define Turing test, so that I'm not allowed to ask questions that any 5-year-old could answer but that fall outside the computer's domain of expertise? (For example, does this picture show baseballs or a bucket? Obviously a 5-year-old could answer that, but apparently an AI researcher can't.)

        1. doublelayer Silver badge

          Re: AI in everything is the problem

          It's unclear. People like to hold Turing test challenges, and programs in those challenges have been ruled human before. That might not be a great basis to declare the test passed, but it is what the test specifies. It also depends a lot on what we want them to do. The original Turing test didn't include sending images to the other party, therefore not requiring the AIs to see. Also, a program trying to pass the Turing test, because it sends back text, has the ability to say "both" where the programs here which are just identifying things have to pick a single one.

          In my opinion, the Turing test is a rough test that is likely to have too much uncertainty to prove intelligence. There were probably a lot of people who liked to talk about it back when it seemed impossible, but now they're pointing out defficiencies in the concept. The only problem is that a lot of people attack anything termed AI without even trying to define what they think AI is, or provide an unrealistic explanation which means something is only AI if it acts entirely human and attained sapience itself without ever being programmed.

    3. C.Carr

      Re: AI in everything is the problem

      To state the obvious, the labels need to consist in more than their complex relationships with other mere labels. The actual things need to be represented by sensory data, and the AI system needs to be located, discretely, in 3D space --- that is, if we want a system to actually *know* what things are.

  11. Ken Moorhouse Silver badge

    Reminds me of...

    Google bombing.

  12. steelpillow Silver badge
    Coat

    Simples

    All we need is an AI that translates between AI data labelling languages.

  13. iron Silver badge

    I do hope that somewhere in ImageNet thereis a picture of a banana that has been labelled "female aardvark."

    1. Mark192

      My local Tesco occasionally has the bananas mislabelled as "small, off-duty Czechoslovakian traffic wardens".

      Serves them right for leaving the label maker out ;)

      1. jake Silver badge

        The Petaluma (California) Whole Foods sometimes has ...

        ... lychee labeled as Hedgehog Eggs. Down the street and around the corner, the local Lola Market has occasionally had their Spiked Choyote (Sechium edule) relabeled as Porcupine Eggs. Given their proximity, I suspect the joker is the same person at both stores. No, it's not me.

        1. Ken Moorhouse Silver badge

          Re: Hedgehog Eggs

          In the UK we had Hedgehog Crisps which caused the Advertising Standards Authority some headaches.

          1. keith_w

            Re: Hedgehog Eggs

            Are they related to Spring Surprise Chocolates?

      2. Ken Moorhouse Silver badge

        Re: "small, off-duty Czechoslovakian traffic wardens".

        I learn something new every day from this site - much of which has nothing to do with technology.

        ===

        Warhol would have run rings around AI with his placement of everyday objects as art. Bananas reminds me of Nico's distinctive vocals...

        https://www.youtube.com/watch?v=AkDJcUCyjCU

      3. Eclectic Man Silver badge
        Joke

        Small off duty ...

        Mark192: bananas mislabelled as "small, off-duty Czechoslovakian traffic wardens".

        That is ridiculous, don't they know that the Czech republic and Slovakia split apart in 1993?

        https://kafkadesk.org/2018/10/30/why-did-czechoslovakia-break-up/

        "On January 1, 1993, Czechoslovakia split into two independent states, the Czech Republic and Slovakia, in what is now known as the “Velvet divorce” (in a reference to the Velvet revolution) due to its peaceful and negotiated nature."

        1. StuartMcL

          Re: Small off duty ...

          https://www.youtube.com/watch?v=oB-NnVpvQ78

        2. Dave559 Silver badge

          Re: Small off duty ...

          "And while Slovak nationalism sentiment strived for more autonomy, Czech nationalism embraced Czechoslovakism, mainly due to their privileged position within the federation."

          -- https://kafkadesk.org/2018/10/30/why-did-czechoslovakia-break-up/

          As the article itself alludes to, I'm sure we can all think of certain other multi-ethnic countries where similar feelings apply…

    2. Anonymous Coward
      Anonymous Coward

      It'd be wrong though. Clearly it's the Bolivian navy on manoeuvres in the South Pacific.

  14. Anonymous Coward
    Anonymous Coward

    Bias is endemic to society, right down to the very roots of the evolution of the languages we use. Cultural biases of thousands of years are embedded in every culture on the planet, and it results in biases in the data sets, even if they are "correctly" labelled. Prejudices in society that lead to fewer members of a minority being successful in particular fields leads the machines to conclude that people of such background are *unlikely* to be qualified, rather than that they are under-represented.

    It is a thorny issue, and one that isn't easily solved, because it begins to tread on concepts of morality and justice in society, and the one thing we can all agree on is that no one agrees on those subjects.

    1. EarthDog

      No one mentioned it but how are idioms handled? They are probably the hardest things to translate.

      1. doublelayer Silver badge

        With automatic translation software, it either just looks at the words and tries to do them literally or it finds an idiom in a big dictionary. When I did some translation, I would always replace them with factual statements. It had less flavor, but at least I knew the reader would understand it without taking the risk that the closest idiom I could come up with was regional. Then again, I was not a professional translator, just a person who spoke multiple languages and hadfriends who didn't.

    2. stiine Silver badge

      I have to disagree, just to prove your point. Please take no offence, or take offence, I don't care.

  15. sketharaman

    GIGO

    In short, AI suffers from GIGO. zzzzz.

  16. EarthDog

    GIGO

    Garbage In Garbage Out. A saying as old as computing. When I was spending a large amount of time merging data from other sources into the databases of a company I was working for I made sure my juniors and the SMEs spent a bit of effort vetting those data. All our data had to be defensible in court. It is the cavalier attitude to data and assumption all data are perfect which caused to avoid those areas. There are no standards of quality though they wouldn't be hard to to develop.

    1. John Brown (no body) Silver badge

      Re: GIGO

      "Garbage In Garbage Out. A saying as old as computing."

      Try saying that to someone in the bulk recycling business :-)

  17. Skiron

    So why isn't there an AI program somewhere that can actually label the photograph itself? 'Cos AI isn't AI really, it's what the programmers tell it to do. As stated above, GIGO and it always will be until a 'machine' is actually cognitive.

    1. doublelayer Silver badge

      I think that approach is too limited.

      "So why isn't there an AI program somewhere that can actually label the photograph itself? 'Cos AI isn't AI really, it's what the programmers tell it to do."

      Whatever our definition of intelligence ends up being, it's got to do what the programmers tell it to do at a low level otherwise it's no longer artificial. We could probably argue about what intelligence is all day, but if AI is possible at all, it has to be implemented with code and machinery which means its instructions are created by another intelligence.

      "As stated above, GIGO and it always will be until a 'machine' is actually cognitive."

      Now this doesn't sound fair. GIGO applies to anything. If you take a human child and prime them with a bunch of false data without giving them the ability to learn that you're lying, they'll believe you. If everybody the child meets insists that the tall wood things that grow outside are called squirrels, they will believe that those are squirrels until they meet some other people who correct the misconception. If you're going to define intelligence as "the ability to figure out that everything people tell you is wrong even though you have no other source of information", that's a high bar. Computers don't get to meet random people and learn from them or even experiment with actions to see consequences. In life, we have a lot more input than any of these programs ever have, and yet there are a lot of humans who get incorrect concepts of how the world is.

      1. Ken Moorhouse Silver badge

        Re: it's got to do what the programmers tell it to do at a low level

        I did tinker with "learning" games some time ago. Connect4, was one such attempt. Very easy to program the rules of the game in. Easy to get the program to "learn" by making moves which result in a win or a lose. But there is no realisation that a mirror-image of that move would have the same win-lose effect... unless it is programmed in. Then there are the counters that are on the periphery of the action... their existence or absence in every possible combination, and in every possible chronology takes up a huge space in the learning dictionary, and in the amount of time taken to crunch games involving those moves. Unless the programmer uses his/her intelligence to collapse down those combinations. The only learning involved is by the programmer in coming across those combinations and adding code in to deal with them. Learning to me would involve the program recognising that a part of this game is a mirror of that game and to insert the code in itself to do that. I suspect that would involve grids full of esoteric data which comprehensively weights every single play, in every possible chronology. Such a program would make sense to nobody because it would be the epitome of abstraction. It would be completely unmaintainable beyond the individual team or individual writing it.

        So what am I saying? I'm saying that the law of diminishing returns applies to AI and that the best compromise is for the programmers maintaining the code for the AI program to be considered as the AI behind the program. This is nothing new. This is the way programs have always been written.

  18. Michael H.F. Wilkinson Silver badge

    I don't like saying i told you so, but ...

    I frequently rail against claims made in AI papers that if the ground truth contains a percentage of errors, any AI system trained on them is likely to end up with a similar actual error rate. I have seen people claim an increase in performance from 97.6% to 98.1% (error bars not included) on data sets where there are two ground truths, drawn up by to medics, which are at odds with each other. In our own earlier work, we managed to get a sort of pareto optimum of 92.5 ± 0.6% on both ground truths, but were in places penalised for finding blood vessels the doctors had missed. It turns out, somehow ground truth 1 has been elevated to The Ground Truth, and the other demoted to "a human observer". And now AIs are better than the poor "human observer" simply because they have been taught to copy all the mistakes the other human has made.

    If ImageNet contains up to 6% error, I will continue to take all claims of 99% or better performance with a considerable pinch of salt. Furthermore, if error bars ar not included, how can they claim to be better than an earlier method if the differences are sub 1%.

    I am not saying deep learning and CNNs are useless, it is just that sloppy science does them a disservice.

  19. theblackhand

    Simples...

    We need another standard....

    And an AI to create the standard of course.

  20. Sandstone
    Devil

    Evil Human

    When some friends of ours had their first child and he was learning to talk, I would point to random objects and say, "This is an aardvark."

    1. Eclectic Man Silver badge

      Re: Evil Human

      In the Dorling Kindersley 'ABC' book for toddlers, each letter of the (Roman) alphabet has several pictures. A is for Apple, Ant and Avocado, but I didn't spot an aardvark.

      However, I applaud your teaching the little one that adults are inveterate liars and can be naughty too. That lesson should serve him well in the coming post-apocalypse dystopia, however it is caused.

  21. Anonymous Coward
    Anonymous Coward

    compounding errors

    It always irritates me that when the "I am not a robot" captcha asks to highlight all motorcycles, the user has to also highlight scooters or else the result will not be accepted.

    1. DeathSquid
      Stop

      One weird old trick to deal with captchas...

      Look near the top of your browser window. There's a back button. Press it, and the captcha disappears.

    2. Henry Wertz 1 Gold badge

      Re: compounding errors

      I do what it asks -- I will not mark scooters as motorcycles since they aren't; I will not mark SUVs as cars since an SUV is not a car. Go ahead and claim I'm wrong, I know I'm right.

      The other one that it seems to have problems with are marking the traffic lights, where I'll mark off the traffic lights and it'll claim I'm wrong (I assume some people are probably marking off the whole pole and everything, not the lights? I don't even know.)

      For some reason, some people don't seem to know what a crosswalk is either, given being able to exactly mark the crosswalks and having it fail.

      1. Ken Moorhouse Silver badge

        Re: For some reason, some people don't seem to know what a crosswalk is

        I think it has already been mentioned, but there is no such thing in UK, so YMMV with how us Brits will answer.

        "The other one that it seems to have problems with are marking the traffic lights, "

        I read the question, where it says click on those that *contain* a traffic light, I'm afraid that I will click on all the images containing the pole, because to my mind the traffic light cannot exist without the supporting pole. Google's devious reaction probably would be to show a traffic light fixed to brickwork with a white line painted to the ground (looking like a pole - the resolution of the images is so awful it is guesswork at the best of times), just to be awkward.

        Thinking about it Google could easily profile people by their answers. Show pictures of Armstrong on the Moon, for example and ask users to tick those images showing the moon. Conspiracy theorists could be segregated by their refusal to select those images.

  22. a_yank_lurker

    Languages

    Human languages are not exactly the most precise in many ways. Anyone who has ever done any translation knows the imprecision in the source makes an accurate translation difficult at time. Also, context is very important as to how something should be interpreted. This nuance makes Artificial Idiocy not Artificial Intelligence the likely outcome of these systems.

    I would like know how these idiot systems would handle the phrase 'While you've a lucifer to light your fag' from the WWI song 'Pack Up Troubles in Your Kit Bag'.

  23. Anonymous Coward
    Anonymous Coward

    A very small casserole.

    A bucket of baseballs should be labelled as a bucket of baseballs.

    God the Renaissance was something that just happened to other people for you wasn't it Baldrick.

  24. Henry Wertz 1 Gold badge

    Reason for errors

    So, I think there's 2 main reasons for these errors:

    1) People (both in the article and forum) have discussed the structural problem, if you have to give each image "a" tag, a bucket of balls is not accurately tagged.

    2) The other issue -- who tagged these things? I bet when these were tagged, you either had someone getting paid minimum wage to go through 1000s of images; some Amazon mechanical turk type thing where they're getting like 1 cent an image (which might make it even less accurate since they'd then prefer to just tag them as fast as possible and probably still not get minimum wage); or student interns (whether paid or not) being asked to tag piles of images. I don't have a suggestion of a better way of doing it, but a) I'm guessing most people would do this as quickly as possible rather than as accurately. b) Even if the person doing it was going for accuracy, after like 1000 images how many people will be paying full attention to what they're doing still?

  25. Fruit and Nutcase Silver badge
  26. amanfromMars 1 Silver badge

    Thank your lucky stars such is so far as light years from the truth

    Turns out humans are leading AI systems astray because we can't agree on labeling

    Oh please, failing to follow leading AI systems is the all too apparent default astray human condition/endemic systemic weakness and exploitable vulnerability resulting in a vast catalogue of problems and conflicts presenting in mayhem and madness.

    However/Nevertheless ......

    At some point soon, through virtual intervention and systems collapses aka program and project hacks, will civilisation be transformed, and that is the stock it would be rank foolish not to be investing in. Such is an inevitability and foregone conclusion held by many but still only practically known to a relative few.

    Such is the current present state of virtually adept Great Games Play. I Kid U Not.

    Forewarned usually affords one the opportunity and pleasure to be forearmed. Good luck with finding the suitable weapons to wield and to yield in defence and attack against such as all of that.

    ........ is certainly infinitely better than the punitive alternatives headlined for consideration and possible activation they be in reply to ...... Hedge Fund CIO: "At Some Point, Through Inflation, War Or Confiscation, The System Will Restart"

    Why is it that so many tend to congregate to try repeating past failed methodologies expecting them to produce something different and new? It is surely illogical and may even be an indicator of a weakness supporting one's flights of crazy fancy into a personal hell and private madness?

  27. wjake

    Terrible example

    "What would happen if a self-driving car is trained on a dataset with frequent label errors that mislabel a three-way intersection as a four-way intersection?"

    What a crap example.

    The self-driving car should have accurate map information, so it won't have to rely on recognizing what type of intersection is coming up.

  28. JamesTGrant
    Happy

    Title…

    Was hoping to read about AI Systems ashtray.

    Am disappoint.

    1. Ken Moorhouse Silver badge

      Re: AI Systems ashtray

      No need to be disappointed. It would probably be full of Stubs anyway.

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