Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
AI News

Microsoft Accused of Damaging The Guardian's Reputation With AI-Generated Poll 123

Dan Milmo reports via The Guardian: The Guardian has accused Microsoft of damaging its journalistic reputation by publishing an AI-generated poll speculating on the cause of a woman's death next to an article by the news publisher. Microsoft's news aggregation service published the automated poll next to a Guardian story about the death of Lilie James, a 21-year-old water polo coach who was found dead with serious head injuries at a school in Sydney last week.

The poll, created by an AI program, asked: "What do you think is the reason behind the woman's death?" Readers were then asked to choose from three options: murder, accident or suicide. Readers reacted angrily to the poll, which has subsequently been taken down -- although highly critical reader comments on the deleted survey were still online as of Tuesday morning. A reader said one of the Guardian reporters bylined on the adjacent story, who had nothing to do with the poll, should be sacked. Another wrote: "This has to be the most pathetic, disgusting poll I've ever seen."

The chief executive of the Guardian Media Group, Anna Bateson, outlined her concerns about the AI-generated poll in a letter to Microsoft's president, Brad Smith. She said the incident was potentially distressing for James's family and had caused "significant reputational damage" to the organization as well as damaging the reputation of the journalists who wrote the story. "This is clearly an inappropriate use of genAI [generative AI] by Microsoft on a potentially distressing public interest story, originally written and published by Guardian journalists," she wrote. Bateson added that it had demonstrated "the important role that a strong copyright framework plays in enabling publishers to be able to negotiate the terms on which our journalism is used."
A Microsoft spokesperson said: "We have deactivated Microsoft-generated polls for all news articles and we are investigating the cause of the inappropriate content. A poll should not have appeared alongside an article of this nature, and we are taking steps to help prevent this kind of error from reoccurring in the future."
This discussion has been archived. No new comments can be posted.

Microsoft Accused of Damaging The Guardian's Reputation With AI-Generated Poll

Comments Filter:
  • by Entrope ( 68843 ) on Wednesday November 01, 2023 @06:05AM (#63971004) Homepage

    What was the cause of the death of this company's credibility: murder, accident or suicide?

    • Re: (Score:3, Insightful)

      by geekmux ( 1040042 )

      What was the cause of the death of this company's credibility: murder, accident or suicide?

      Rather funny that humans are finding themselves turning off AI features and services while they "investigate" how or why their AI could have possibly done something like that. As if AI isn't out there sucking up human behavior like a sponge.

      What, AI can't be curious as to cause of death? Like other journalists out there aren't just as curious. If AI ends up learning our humor (as in learn from the parents post), I'm assuming we're going to have to start working on that too-soon plugin and turn the sensit

      • by StormReaver ( 59959 ) on Wednesday November 01, 2023 @07:04AM (#63971124)

        What, AI can't be curious as to cause of death?

        AI is a statistical model, and is therefore not capable of emotion.

        • Re: (Score:2, Insightful)

          by geekmux ( 1040042 )

          What, AI can't be curious as to cause of death?

          AI is a statistical model, and is therefore not capable of emotion.

          Fine. All the more reason we should be accepting of the indifference. The poll wasn't inaccurate. It might be deemed insensitive...at least until 37 more human journalists around the world put up the same poll next week. For profit and clickbait.

          • Re: (Score:2, Flamebait)

            37 people doing the same insensitive thing does not make it any less insensitive.

          • I've been reading the news for a long time, and I don't recall seeing polls next to crime stories asking me to make a guess at who did it or how it was done.

            • I've been reading the news for a long time, and I don't recall seeing polls next to crime stories asking me to make a guess at who did it or how it was done.

              Society gets their news from social media now, and there is all manner of discussion and debate that happens regarding whodunnit and how. Hollywood wastes no time either on capitalizing some of the worst out there. The Murdaugh murders is a prime example.

              Quite frankly, polls like this may even be beneficial to combating violence. If politicians were actually educated on the fact that the overwhelming majority of gun deaths in America are due to suicide, they'd stop lying about gun violence and instead st

              • instead start focusing on the actual problem; mental health.

                Wont happen. No profit in it.

              • If politicians were actually educated on the fact that the overwhelming majority of gun deaths in America are due to suicide, they'd stop lying about gun violence and instead start focusing on the actual problem; mental health.

                The politicians standing in the way of mental health care are the same ones who keep saying the problem is lack of mental health care. So it really doesn't matter what they know, what matters is how they vote, and they are voting against the thing that they are saying is the solution.

              • by tsqr ( 808554 )

                The US doesn't have a monopoly on mental illness; however, it does seem to have cornered the market on gun-related deaths, as far as first world countries are concerned.

            • by mspohr ( 589790 )

              Leave it to Microsoft to break new ground in their never ending greed.

          • There is a key difference here, that highlights an issue with the use of AI.
            All 37 of these hypothetical reporters can be held responsible for what they decide to write. You can also hold any publishers responsible for sending it out the door.

            AI makes that a bit sticky. There is no "writer". Who do we hold responsible? The Guardian for publishing the article? Microsoft for the generated content? The techs who trained the model? The sources that the model was trained from?
            Clearly someone needs to he h

            • by tsqr ( 808554 )

              You seem to have missed the part in the summary that explained that the Guardian didn't publish the poll article; it was placed next to a Guardian article by Microsoft's news aggregator.

          • by tsqr ( 808554 )

            A poll asking readers to choose among a limited number of "opinions" is meaningless except as an indication of what the responding readers have chosen (duh), and has nothing at all to do with what actually took place; thus, talking about "accuracy" is pointless.

        • "I hate this! It is revolting!" "More?" "Please."
        • by Anonymous Coward

          What, AI can't be curious as to cause of death?

          AI is a statistical model, and is therefore not capable of emotion.

          There are humans that are also not capable of emotion.

          Amusingly, they tend to end up in positions running companies, and publishing PR statements not unlike the Microsoft and Guardian ones here.
          Much less amusingly, those PR bytes are acceptable to others on about the same footing as this AI poll.

        • by quenda ( 644621 )

          AI is a statistical model, and is therefore not capable of emotion.

          The human brain is just a collection of neurons, with electrochemical interactions, and is therefore not capable of emotion.

          • By itself, no. I think that an accompanying endocrine response would be required for what we conceive of as emotion, but it certainly involves our neurological processes as well.

        • Re: (Score:2, Interesting)

          by HiThere ( 15173 )

          Yes, current AI is a statistical model. This is not proof that it can't have emotions.
          FWIW, I don't think current AIs have enough self-awareness to have emotions, but being a statistical model is not a decent argument. Anything can be turned into a statistical model, and in principle in a way that is bijective (i.e. 1 to 1 and onto). (Of course there's a bit difference between "in principle" and "in practice", but it's still an invalid argument.)

        • So, much like the average Slashdot reader?

          (take that you NPCs!)

        • by mspohr ( 589790 )

          Or intelligence

    • by Opportunist ( 166417 ) on Wednesday November 01, 2023 @08:20AM (#63971292)

      Without a proper Cowboy Neal option, it's not a proper /. poll.

    • A.I. :D

    • Trick question - it never had any credibility, alive or dead.

    • D) Cowboy Neil did it in the pool with a pool noodle. Mystery solved!
  • by GloryWacky ( 10410843 ) on Wednesday November 01, 2023 @06:52AM (#63971088)
    This is just yet another embarrassing example of maximizing profits by letting these LLMs run wild because they don't want to pay for any human oversight. I wouldn't be surprised if they try and crowdsource the work by having some kind of "preview" section where "community members" can "help improve the content" before it goes live. There is no shortage of people willing to give up their free time so they can feel like they are part of some online "community".
    • Oh yeah, the intelligence of the swarm on the internet.

      THAT can't backfire at all.

    • Heh, I see what you did there. Nice.
    • by Anonymous Coward

      I wouldn't even give it the LLM/AI credit - this was just crappy product. If you work in advertising, you always have to be mindful of the way your ads appear next to copy. The fact they didn't think about this tell me that they either don't know what they're doing, or they're just hopeless at running a complex project (and probably clueless about the AI they're stuffing into anything and everything they can find).

      Unless the news article is about a cute kitten, you almost never want a directly on-topic poll

    • Thank you for contributing your time to help improve the content of Slashdot with your comment, valued community member 10410843!
  • by chas.williams ( 6256556 ) on Wednesday November 01, 2023 @07:09AM (#63971136)
    Because that's the next step: tell the public what they want to believe.
    • That's how the news outlets work today. Tell the audience what they want to hear because that way they think that you tell The Truth (tm).

      • by haruchai ( 17472 )

        "That's how the news outlets work today"
        Today?
        Walter Lippman, called the Father of Modern Journalism, talked about the manufacture of consent in his 1922 book Public Opinion

      • News is no longer news, its clickbait to get impressions
    • Our politicians have already taken that step.

  • The Guardian is doing a perfectly good job of eroding its trust, with their recent push to make the mobile apps unusable for unpaid subscribers. Also, their now more polarized, some might say even left extremist bias is pushing even solid (but moderate) leftists like myself to read more of other sources to keep my perception of not being closed in a bubble.

    As for the AI poll itself, I guess they felt offended and it's their prerogative. Although I think publications should stop being so effing protective of

    • Re: (Score:3, Informative)

      by Malenfrant ( 781088 )

      The Guardian is not Leftist at all. It's a solidly Liberal 'Centrist' rag. Sure, it allows some Left-leaning journalists to write opinion pieces, but it allows just as many, if not more, Right-leaning journalists to do so as well.

      Liberalism is not Left-leaning. It may seem to be, in Societies that skew so far to the Right as ours. But that's simply a matter of a biased perspective. Left is about tearing down hierarchies and creating more equal Societies. Right is about enforcing hierarchies and allowing tho

      • Re: (Score:3, Interesting)

        by HBI ( 10338492 )

        You've slapped your own labels on stuff. Marxism as implemented was quite hierarchical.

        Shades of gray abound.

        • I also found the previous comment to be quite hypocritical. I don't think the other user has their mind properly set on liberalism. The way I see it these days, most liberal publications pend more to libertarian/anarcho-capitalist views than whatever The Guardian usually publishes articles about (social democracy-leaning, but as of late, a bit deranged socially).

        • Re: (Score:3, Interesting)

          by Malenfrant ( 781088 )

          You've slapped your own labels on stuff. Marxism as implemented was quite hierarchical.

          Shades of gray abound.

          Stalinism was indeed hierarchical. This is why I am not a Communist, nor am I in any way Far Left. Some level of hierarchy is inevitable. The problem with Communism (which is an economic system, by the way, not a political system. It's important not to get the two confused) is that it doesn't include enough checks and balances to prevent charismatic individuals taking over and building a new hierarchy with themselves at the top.Leninism wasn't hierarchical, but he wasn't able to remove the hierarchy enough

          • by Viol8 ( 599362 )

            " Left and Right are political positions not economic positions."

            Sorry? The traditional left and right have very distinct economic positions. The right tends towards free market, low taxes and profit and hoping for some magic trickle down whereas the left tends towards higher taxes, market control and a mother knows best state spending spree.

            • The orginal left and right were very much all about the contemporary political nature in post-revolutionary France. It took no time at all however for left/right to change meanings, which constantly changed up until now, so that the meanings are currently vague and indeterminant without knowing the context.

          • The problem with Communism (which is an economic system, by the way, not a political system. It's important not to get the two confused) is that it doesn't include enough checks and balances to prevent charismatic individuals taking over and building a new hierarchy with themselves at the top.

            And there's a similar problem with Capitalism. The major difference, politically, is that Communism leans politically towards a powerful government to try and maintain the new economic system until the masses accept it, while preventing the charismatic leaders from corrupting it. Whereas Capitalism leans politically towards leaving things alone with a very weak government so that the rich and power remain rich and powerful and provide lots of political donations.

            Of course in real life, Communist leaning co

          • Stalinism was indeed hierarchical. This is why I am not a Communist, nor am I in any way Far Left.

            Stalin was not a communist, and the USSR never claimed to be a communist country, it was the Union of Soviet Socialist Republics, where the word "soviet" translates roughly as "worker's counsel." My understanding is that the hierarchy of soviets was set up under Lenin, and lasted right up until the whole house of cards collapsed.
        • The next stage of Marxism is for the working class (that's anyone who works for a wage, not just low-paid manual workers,) of the world's most advanced capitalist economies to unite and seize control of the means of production. This simply hasn't happened yet. The German Revolution of 1918-1919 is about as close as we got to this maybe starting to happen. The closest thing that has actually happened you can look at is probably the Paris Commune of 1871.
          • by HiThere ( 15173 )

            I think you could use the anarcho-syndicallists as a better example. But it hasn't ever happened at large scale, and won't, because that kind of organization doesn't scale well enough. No Communist, or even communist, group has ever succeeded at much larger than village scale. (The small "c" communists include things like religious communes, and possibly the Paris Commune during the French Revolution for awhile, but that organization was at best meta-stable.) Another possible example is some of the Isra

        • Marsixm as theorized was not strictly hierarchical, except as to labeling existing hierarchies and then wanting to flatten them.
          You may as well say that Capitalism "as implemented" was quite evil and misguided, leaving all power in the hands a few.

          Failure to distinguish theory from practice, or lumping everything into one category is naive or agenda driven. Ie, "laissez-faire capitalism is evil therefore we should ban corporations!" is just as misguided as saying "Stalin and Mao were evil men, therefore su

      • Liberalism should be left-leaning, it's just that right-wingers in the West love to pretend to be liberals whilst twisting the ideas in service of power, and they drown out the proper ones. It's easy enough to derive the need for socialism from liberalism - basic liberties like the freedom for a thousand people to eat a nutritious diet without being arrested for stealing food are clearly far more important than expensive ones like the freedom for one person to go around the world in an enormous private yach
      • It's left leaning - certainly relative to the current Overton Window in the UK. You're a lot more likely to see an article advocating a left wing viewpoint than something that would be typical in the Telegraph.
      • by Mitreya ( 579078 )

        The Guardian is not Leftist at all. It's a solidly Liberal 'Centrist' rag. Sure, it allows some Left-leaning journalists to write opinion pieces, but it allows just as many, if not more, Right-leaning journalists to do so as well.

        That was not my impression and at least this survey confirms my theory: https://www.allsides.com/sites... [allsides.com]

    • Also, their now more polarized, some might say even left extremist bias is pushing even solid (but moderate) leftists like myself to read more of other sources to keep my perception of not being closed in a bubble.

      Just to demonstrate how far to the right the republican party has moved, I will post some quotes. Without googling you have to tell me who said them.

      1) I believe in the idea of amnesty for those who have put down roots and lived here, even though sometime back they may have entered illegally,

      2) "no reason why on the street today a citizen should be carrying loaded weapons" and that guns were a "ridiculous way to solve problems that have to be solved among people of good will."

      Anyone care to guess what libe

      • Yes, and in today's American political climate, Reagan is a RINO because the Rs have been corrupted by two very strong wings - the evangelicals and the Trumpists, both of whom have very non-traditional Republican views.

      • And? We can show how left the Democrat party has become by comparing them to Carter and Clinton.

        Times change.

    • As for the AI poll itself, I guess they felt offended and it's their prerogative.

      Hold on, a multiple choice poll asks readers to speculate on the cause of a 21 year old woman's death, and you seem confused as to why people find that offensive?

      Although I think publications should stop being so effing protective of their content with all things AI-related. It's definitely transformative, remixing of content and it should fall under fair use, and even though some actors are benefitting financially from it (be it OpenAI, Microsoft and others), the common good that new, dynamic sources of information bring should be more widely accepted instead of pushed back.

      Again, I find your perspective kinda odd.

      As for the actors, I suspect MS had some agreement with the Guardian to post some Guardian stories, though the decision to add the AI generated poll was probably just an MS decision.

      But as for being protective, is it too much to ask that a human review and even edit any AI content before it goes out to the p

    • by whitroth ( 9367 )

      Their leftist bias? I'm sorry, you mean their efforts to report news, rather than right-wing propaganda?

      By the way, the leftist bent of the Guardian, which I read for news, not the bs in US infomercial media, has a reasonable underpinning, given that IT'S BEEN THAT WAY FOR OVER A CENTURY AND A HALF, that they're pro-Labour (that's English, they're a paper from the UK, and it's spelled the Labour Party).

      You remind me of the songwriter from Rage Against the Machine, who, in an interview, said he'd gotten a tw

      • I'm not American, I'm European, and I'm quite familiar with The Guardian using it as my main international source of news for well over a decade, reading their online page nearly daily...

  • by VeryFluffyBunny ( 5037285 ) on Wednesday November 01, 2023 @07:31AM (#63971180)
    ...that brought you the racist, misogynistic 'Tay': https://www.cbsnews.com/news/m... [cbsnews.com]

    Apologies should mean that you're sorry & that you won't do it again. Well, Tay wasn't the first & hasn't been the last. Big tech companies like Microsoft won't change unless you require them to by law & enforce it.
    • by HBI ( 10338492 )

      In a statistical model, the need is to control the input to avoid these kind of outputs. I think you're seeing the limits of the LLMs right there. If the input is carefully controlled, the output will remain stale and antiseptic. Alternatively, if you create a hand coded filter for the LLM that is tunable, that would seem to hobble the utility of the model.

      I imagine a lot of papers on this topic in the next few years. Anywho, throwing up your hands and complaining about Nazi bots is probably not going t

    • Re: (Score:2, Interesting)

      by WDot ( 1286728 )
      Ah yes, an experimental chatbot that learned from interactions with users was trolled into saying offensive statements, and once Microsoft noticed they immediately pulled it down. Clearly this episode shows the need for heavy handed regulation! Was anybody harmed by racist statements said by a dumb chatbot (in response to obvious goading)? Did Microsoft bend to law enforcement or the obvious failure of the experiment when they took Tay down?

      There really is an absurd bias toward needing regulation imposed
      • Are you defending Microsoft for publishing extremely racist, antisemitic, & misogynistic views & hate speech on widely publicly available social media, including in countries where Nazism & hate speech are criminal offences? How many people of colour, Jews, & women who read these posts do you claim weren't offended?

        Sorry, can you run your circumstantial excuses by me again? I want to be sure that you're saying what you appear to be saying or perhaps you want to think that one over a littl
        • by WDot ( 1286728 )
          No, I am saying nobody really thought that a chatbot thought genuinely communicated actual hatred and/or material harm to them, and that nobody thought Microsoft stood behind the "speech" that the chatbot posted (evidenced by the fact that they immediately took it down and apologized loudly). I stand by that. If a hacker hacked Taylor Swift's website and put swastikas on it, would you believe this is rock-solid evidence that Taylor Swift is genuinely a Nazi?
          • So you're not saying that Microsoft didn't publish hate speech but you are making excuses for them. Yeah, right. That's not how it works. You can't say "Oopsies! We didn't mean it!" You try "Only joking!" or "Didn't mean it!" the next time you're going through airport security and see what happens. The other part of it, if you accept their oopsies, is that they were negligent. They rushed it out into the public without properly testing it in order to ensure that they wouldn't be responsible for publishing h
            • If you think a bot saying mean words is on the same level as bringing weapons through airport security, and that a bot saying mean words is evidence of a large corporation spending 9 figures to intentionally automate the production of mean words on the Internet, I can only suggest that maybe you go outside and get a life
              • Hah! Trying to distort what I said now? I took the example of saying stuff at a time when people are held accountable for what they say & you bring up the topic of weapons? You gotta do better than that.

                If you make a machine that commits a crime, i.e. hate speech, what's the difference between that & committing the crime yourself?

                P.S. Your dismissive attitude towards other people's dignity, well-being, & right to live without hate speech directed at them is disturbing.
  • That AI sure is powerful, it accomplished something no man could do even if he tried.

    And $deity knows, people have tried! But how do you damage something that doesn't exist?

  • A poll should not have appeared alongside an article of this nature, and we are taking steps to help prevent this kind of error from reoccurring in the future.

    That's just the modern deployment strategy. Push your beta stuff live and when if fucks up apologise with some bullshit reasoning. We see that all the time.

  • by nightflameauto ( 6607976 ) on Wednesday November 01, 2023 @09:26AM (#63971434)

    I mean, sure, their system created the poll, but some moron somewhere in the chain had to decide that it's totally fine to use AI generated polls alongside their stories on their site. Who is that person, and why aren't people demanding that person's job?

    I'm a big fan of automation. Most of my job is automating things for our customer service department. There are ways to go about automation that don't make everyone involved look like a complete imbecile, but this current strategy of "THROW AI AT IT! IT'S HELPING!" is not that way. As demonstrated here. I think we need to start cracking down on these deluded fools that believe in the machine being the perfect way to replace people without any consideration whatsoever about whether the machine is up to the task or not. This blind belief in the computer is getting ridiculous. All it would take is one of these bumbling morons tossing AI at some piece of infrastructure thinking they're saving some money by not paying a human to have the final say over it and things could get seriously ugly seriously fast.

    • "How is what a company posts on their own website their fault?" Poor Microsoft, held hostage by its own employees, forced to post polls on MSN.com against its corporate will!
    • by jbengt ( 874751 )

      I mean, sure, their system created the poll, but some moron somewhere in the chain had to decide that it's totally fine to use AI generated polls alongside their stories on their site.

      "some moron" is Microsoft
      "their stories" are Guardian's stories
      "their site" is Microsoft's news aggregation site
      That's "how is this Microsoft's fault".

  • All the polling in the universe isn't going to find how the woman died. That's what medical examiners learn through autopsies and other methods. Who cares what everyone "thinks" happened. Dumb and dumber! Have we killed the notion of objective truth and critical thinking?

  • There's no such thing as "artificial intelligence". Every demonstration of supposed AI proves this, and this "poll" is just one more example. When "AI" creates bogus "facts" on the fly, or does something this brain-dead, after a while it becomes clear that "AI" isn't anything like "intelligent".

  • ... Microsoft do? Include a spell checker?

  • This time it is actually wondering why Americans, after getting news on what Hamas did on October 7, are favoring Israel. Faced with accusations that their AI is trained on Marxist literature exclusively, the publication apparentky added Mein Kampf to the training set for balance.

  • This is why "AI" isn't as valuable or revolutionary as people think. You can't actually put anything important under the control of an LLM, which significantly limits its value and utility.

  • I suspect most online comments are AI-generated, totally ignoring the previous comments and given the one line non-sequiturs they come up with.
  • This over use of early stage AI because of profit seeking rather than legitimate interest in AI as a science is going to ruin the field of AI, or at least the credibility of it. Like crypto before it which *coulld've* been sucessful and an eventual replacement for traditional currency, but because of all the scammy, scummy, and downright clownish (Bored Apes, anyone?) garbage that it was used for, it certainly won't happen any time in the near future, if ever.

The biggest difference between time and space is that you can't reuse time. -- Merrick Furst

Working...