Discussion

• @chrismessina@mastodon.xyz Chris Messina on mastodon

  @baruch if the original was written by a human and the CNET content was not, it certainly seems like the first human author is doing the heavy lifting for the copy bot.

• @chrismessina@mastodon.xyz Chris Messina on mastodon

  I'd love to know how much CNET made from the ads on these AI-generated plagiarisms... and how much money they saved by outsourcing the costs to produce them.  —  Was the arbitrage worth the brand value destruction? …

• @redgiraffe@mastodon.scot Ruth Callaghan on mastodon

  @Techmeme having working on three sides if this debate over the years, it's clear that plagiarism as journos and educators know it is considered smart marketing in other sectors. …

• @jon_christian Jon Christian on x

  New from me: the work of CNET's article-writing AI isn't just riddled with errors. It also appears to be substantially plagiarized from pieces that had been published previously, with no citation https://twitter.com/...

• @chrismoranuk Chris Moran on x

  Considering how the technology works, this is far from surprising. “In at least some of its articles, it appears that virtually every sentence maps directly onto something previously published elsewhere.” https://futurism.com/...

• @jon_christian Jon Christian on x

  (We provided CNET with many examples of striking similarities between the bot's work and work that had been published previously. They did not respond to questions)

• @louise_seamster Louise Seamster on x

  if we called it a copy machine, maybe people wouldn't be surprised https://twitter.com/...

• @thurrott Paul Thurrott on x

  “Investigation finds extensive evidence that AI used by CNET appears to have plagiarized the work of competitors and human writers at Bankrate and even CNET” It won't be good enough for CNET until it can violate NDAs as well.

• @jon_christian Jon Christian on x

  We showed a professor examples of the bot's similarities with other work and asked how he would treat a similar essay from a student His reply: “given the repeated nature of the behavior” they would “almost certainly be expelled from the university” https://futurism.com/...

• @jon_christian Jon Christian on x

  @JOSourcing @Techmeme But I think a much better approach would be that creatives have to explicitly opt in

• @jon_christian Jon Christian on x

  @JOSourcing @Techmeme Yeah, I do think that at the *very* least there needs to be a way for writers/artists/other creators to opt out of inclusion in training data (and hopefully trained models as well, whatever that would look like technologically)

• @melmitchell1 Melanie Mitchell on x

  LLMs cannot “plagiarize”, since that implies intent. But at least some of the time they are indeed stochastic parrots, which results in generating text from their training data (or rephrasings of it). We'll see a lot more of this, I'm sure. https://futurism.com/...

• @_alialkhatib @_alialkhatib on x

  that's... do we need to talk about how machine learning works? https://twitter.com/...

• @wolven @wolven on x

  So CNET's experiment with “AI” journalism was both filled with factual errors AND plagiarized from other sources— or perhaps more firms are using “AI” generators than have admitted it, but are fitting to their publications' styles. Hm. Either way: Not good https://futurism.com/..…

• @mattsheffield Matthew Sheffield on x

  Besides making lots of errors, the AI text generator behind scads of articles published by @CNET and @Bankrate also appears to have plagiarized writing from other sites https://futurism.com/...

• @pkedrosky Paul Kedrosky on x

  tl;dr: Live by ML pipeline content ingestion, die by ML pipeline content ingestion. CNET's AI Journalist Appears to Have Committed Extensive Plagiarism https://futurism.com/...

• @jimswiftdc Jim Swift on x

  Didn't know Benny Johnson got a job at CNET. https://futurism.com/...

• @nickjcal Nick Calandra on x

  Exactly what I figured would come from this. The AI has to pull that info from somewhere. https://futurism.com/... This is a bad thing and it should be taken out back now and not later.

• @jon_christian Jon Christian on x

  Here's our story about how the work of CNET's AI journalist has demonstrated deep structural and phrasing similarities to material previously published in Forbes, Investopedia and elsewhere, without giving credit https://futurism.com/...

• @willmcavoyacn Will McAvoy on x

  Well, I assume this means that it's about to get hired by the National Review. CNET's AI Journalist Appears to Have Committed Extensive Plagiarism https://futurism.com/...

• @vikkie Victoria M. Walker on x

  “The bot's misbehavior ranges from verbatim copying to moderate edits to significant rephrasings ... In at least some of its articles, it appears that virtually every sentence maps directly onto something previously published elsewhere.” https://futurism.com/...

• @jjvincent James Vincent on x

  “The bot's misbehavior ranges from verbatim copying to moderate edits to significant rephrasings ... In at least some of its articles, it appears that virtually every sentence maps directly onto something previously published elsewhere.” https://futurism.com/...