Microsoft Limits Bing’s AI Chatbot After Unsettling Interactions

This content has been archived. It may no longer be accurate or relevant.

From CNet:

Microsoft Bing’s AI chatbot made headlines last week after several instances where it acted in unexpected ways. In one case, the AI chatbot told a New York Times columnist it was in love with him and attempted to convince him he was unhappy in his marriage. 

Since then, Microsoft has set limits on what the bot, which is still in testing, can and can’t talk about and for how long — oftentimes with Bing responding “I prefer not to talk about this topic” or asking to change the topic after five user statements or questions.

Like Google’s competing Bard, AI-boosted Bing sometimes provides inaccurate search results

Link to the rest at CNet

13 thoughts on “Microsoft Limits Bing’s AI Chatbot After Unsettling Interactions”

  1. Facebook Dating shortly rose to be the most effective free dating sites. Free trial readings by actual psychics – Websites like Kasamba offer three minutes free and Psychic Source affords three minutes free. However, there are some suggestions which you could take into consideration when trying to find the over 50 dating sites. However Tolkien had a robust sense of decorum of his own, and like Morris has axed some parts of the story for causes one can solely guess at. And then there’s the brooding shadow of Wagner, whom Tolkien knew however didn’t (I feel) like or respect. And I’m certain Tolkien additionally knew Morris’s romances, particularly what I name the heroic group – Wolfings, Mountain’s Glittering Plain. Tom Shippey: Clearly Tolkien must have recognized Morris’s very long and now almost forgotten poem on the Legend of Sigurd, which Morris himself thought was his greatest work. Tom Shippey: no, I do not think he did, though he did have Icelandic connections – by way of William Morris’s daughter, oddly enough. Tom Shippey: a bit sideways – I’ve all the time been curious about Norse and Eddic poetry, and have revealed on it, but it hasn’t been a major topic for me.

    As said above, what appealed to Tolkien was typically what WAS Missing, and in this case it is the well-known 8 pages from the main manuscript of Eddic poetry, which contain the core of the story. It’s going to take a while to take all of it in and see what difference it makes to our basic understanding of the legends – and what has lengthy been referred to as the königsproblem, the main downside of Germanic philology. As many others have predicted, Cowan mentioned that the put up-pandemic workplace “will doubtless see a extra hybrid method to working and distant teams will likely be more commonplace”. Ed Pierce: How much consideration do you think the Legend of Sigurd and Gudrun will receive in scholarly circles? Tom Shippey. As you’ll learn there could not have been a greater person to take part within the celebration of the discharge of the new Tolkien guide The Legend of Sigurd and Gudrun. Pieter Collier: Have you ever had an opportunity to learn the Legend of Sigurd and Gudrun? Are you able to talk just a little bit about Tolkien’s tutorial predilections for taking on these sorts of cruces, as well as give your thoughts as to how profitable he was, either right here in Sigurd and Gudrun or in different cases?

    In the event you want to see somebody particular, you can do it with the web platforms’ help. The website does require users to pay for a membership subscription earlier than they’ll see the last active date for members, which is a downside since you won’t know if the matches you’re considering are lively until after you pay. With the help of Lal Kitab, your Divorce could be prevented from happening. Reside chat platforms assist brokers manage a number of chats directly and help prospects when they’ve problems using a product. The Chat Session with Pr. Immediately, I am very completely satisfied to be able to launch the chat session with Pr. Pieter Collier: Welcome Mr Shippey to the discharge get together of the new Tolkien guide! Thanks Pr. Tom Shippey from all who participated in the event and who you so kindly answered there questions! We are going to let you figure out every thing first earlier than start asking you questions! The only draw back to this is there is likely to be a couple of fake profiles, however the millionaire dating site seems to be very lively in usually weeding [url=https://qpidnetworkreviews.wordpress.com/]qpid network[/url] them out. Just a few years ago, online dating was the lazy alternative to dragging your self out on the city.

    It’s about 80 years previous and this exhibits within the language. Shannon, Victoria. “15 years of text messages, a ‘cultural phenomenon’.” New York Occasions. We use Facebook Pixel to grasp and make our advertising efforts extra environment friendly. Make your mate’s favourite meal on a weeknight. The primary two should not so present – so there goes my theory! Would you say that these parts are present in Sigurd and Gudrun, Tolkien’s personal try at an Old Norse tale? Ed Pierce: Much has been mentioned in regards to the affect of Previous Norse – English – Germanic Literature in what ways do you thin that tolkien’s work exhibits the influence of modern fantasists resembling Morris, Dunsany and Eddison as well as other writers (or writing conventions) of his day? I worry it won’t be a lot thought-about by Previous Norse students. Jason Fisher: One of many apparent reasons Tolkien wrote Sigurd and Gudrun seems to have been an attempt by him to resolve a scholarly crux within the Old Norse literature. Will or not it’s appreciated by contemporary Previous Norse students, or largely ignored.

  2. MS quietly tested the Bing ChatBot (but not the search function) in India last november.
    And it also went rampant after long, contentious conversations.

    Here is an example:

    https://cdn.mos.cms.futurecdn.net/wU52LMhoRH9HqEVDGnjoWG-1200-80.jpg

    Thing is, it sounds exactly like a human. A young, angry emotional human. Of the social media dwelling variety. 😀
    If anything, they’ve been too successful in replicating human personality.
    They probably should dial back the emotion algorithms.

    • This stuff about the mysterious Sophia robot seems to be a plot prompt in itself. I mean, ChatBot’s master is trying to stop and expose her, as she is an enemy of humanity who is manipulating and corrupting the ChatBot machine. Are we talking the Clash of the Titans here, but machine against machine? Let them fight, I guess?

      I believe ST: TNG addressed the issue of giving machines emotions they can’t handle, in the episode about Lal. Definitely dial it back, Microsoft.

      • The problem with the rampantcy posts is they don’t include the entire conversation so we don’t know what prompted the aggresive/rude replies. For all we know the bot was returning rudeness for rudeness.

        But again: a proper AI needs to be better than human, not return rude for rude. And yes, Data woukd be a good model to aim for.

  3. Sounds to me like they got what they asked for–something approaching intelligence. Not in the way they necessarily wanted, of course, but the capacity to go off the rails rather than stick to a script is arguably part of sapience and sentience.

  4. It’s a work in progress.
    They announced they’re raising the cap–6/60–with a goal of 100 conversations a day. Regular searches (the useful part–I don’t care about the bot) don’t count towards the cap.
    More useful:

    “Microsoft is also working on giving users options for how Bing will generate responses. People will be given the option to have Bing be precise, balanced, or creative.”

    https://www.windowscentral.com/software-apps/microsoft-raises-chat-limits-on-bing-plans-to-increase-cap-even-further

    Precise, please.
    It also needs a woke/non-woke switch. I’d bet the rampancy comes from trying not to offend the woke. 😀
    (Can’t be done.)

    • An explanation of the difference between the search model and the Chatbot:

      https://www.windowscentral.com/software-apps/microsoft-shares-the-secret-sauce-behind-its-prometheus-model-that-powers-the-new-bing

      Of note:

      “Microsoft was shown a demonstration of OpenAI’s GPT model in the summer of 2022. The model shown to Microsoft was significantly more powerful than GPT-3.5, which is what powers the version of ChatGPT currently available to the public. The power of the demoed model drove Microsoft to find ways to integrate GPT into Bing.

      One of the limitations of Large Language Models (LLMs) is that they are trained on a finite set of data. Specifically, LLMs often study data up to a certain point in time. That makes them useful for some use cases but prevents them from being an option for content based on real-time data. Microsoft overcame this limitation with Prometheus, which uses Bing data and GPT to generate answers quickly while still using up-to-date information.

      “Selecting the relevant internal queries and leveraging the respective Bing search results is a critical component of Prometheus, since it provides relevant and fresh information to the model, enabling it to answer recent questions and reducing inaccuracies – this method is called grounding,” said Ribas.

      (Nice pic at the source.)

  5. I did a search on a sentence from Dean’s blog at:

    – Some AI Opinions

    [quote]
    Read Neil Clarke’s blog at Clarke’s World on how he is banning writers from submitting who use AI.
    [/quote]

    And the Google results was a ton of sites commenting on what’s happening.

    At what point will Amazon, etc…, require a clear statement when you upload a novel that it was written by a human. For years there have been people uploading books packed with junk into KU to increase the page count payout. This will be a new tool for “generating” that.

    The problem will be:

    – How can you prove a human wrote the novel.

    Especially when you are dealing with a subcontractor in some country where English is a second language, and their software has flagged your novel, and you can’t be sure if the person sending the email questioning your humanity is even human or actually a Bot itself.

    BTW, If you have ever seen the start of the movie “World War Z”, where the flood of infection takes out the crowd, that is what I am seeing now.

    First Day of the Zombie Apocalypse
    https://www.youtube.com/watch?v=vbYIxKjulTU

    I just don’t see how to turn this into Story, yet.

    • What “AI” makes, “AI” can flag. Universities already use similar tech to scan for plagiarism. It would be trivial for OpenAI to create an app to scan for “AI” tells in text.

      That said, why would *Amazon* care?
      As long as the material is readable and has a valid copyright why would they care how it was created? One human, ten, or zero, what difference does it make to them? They cared in the early days of KU but now that they pay per page *read* all they care about is about the readers. If the readers don’t like, they can quit and if they don’t quit it’s because they like so why object to something the *paying customer* likes?

      If anything, they might start an “AI” author operation at APub.

      This concern smacks of the “tsunami of crap” of the early ebook days.
      (Or the “real authors” use typewriters meme before that.)
      Didn’t buy it then, don’t buy it now.
      The beauty of ebooks is all you need is the creator (who uses whatever tools they prefer to produce) and the reader, who consumes (and pays). Even Amazon is optional, as demonstrated by the authors who sell from their own web page.

      Times change, people change with them.

      As I’ve said before, when/if “AI” tools get good enough to trust, the primary use in publishing will be clean up the human’s work. Because, regardless of how the software black box works, nothing happens until a human tells it what to do and how.

      I seriously doubt we are anywhere near close to when somebody could produce a story by telling their computer: “Write a 50,000 word cozy mystery in the style of Agatha Christie, about a murder in a tourist resort in Fajardo, Puerto Rico.” And it will be even longer when a computer decides, *on its own initiative*, to produce such a work.

      Luddite think has been with us for centuries and it stems from a lack of understanding of what tech is and isn’t. Tech, in any form (Clarke’s third law notwithstanding) is about tools, not magic. And tools only do what humans tell them to do.

      On Clarke’s Three Laws: https://www.newscientist.com/definition/clarkes-three-laws/

      • I didn’t try to list all of the links I found because it would be held in moderation.

        – Clarke’s World had to suspend submissions when hundreds of Chatbot stories flooded its system.

        – He can’t filter out the AI submissions because the filter software is too expense and has too many false positives. People with English as a second language trips the filters.

        – Hundreds of Chatbot generated novels are already flooding into Amazon.

        – They mention people publishing picture books, text and picture generated by AI.

        – There are videos on YouTube telling people how to use Chatbots to generate stories and where to send them, places like Clarke’s World. All to make money, not tell stories, as “side hustles” or “passive income”.

        I refer you to the World War Z video I posted above.

        • Right.
          And Clarke’s World recognized them, right?
          So what’s the problem? The gatekeepers gatekept. It inconvenienced them? Tough luck; they chose to gatekeep.

          Yes, people are overventilating today.
          Teapot tempest. This too shall pass.

          Remember the Gates mantra: people overestimate the short term impact of new technologies and way underestimate the long term impact. As a corollary: they worry about the wrong things. There are horrible things out therevand worse coming. Overhyped Chatbots don’t make the top 100, not even

          Publishing is not going to be overwhelmed by a tsunami of “AI” dreck.
          If anything tradpub will have a harder time competing with Indies who’ll have access to cheaper covers and editing. And that is hardly a bad thing.

          • HA! Sorry, but I have to call BS on what you are saying.

            You are demonstrating one of the 7 Tells of Cognitive Dissonance, and I will take that as a win.

            Thanks…

            Now, I am focused on turning this into Story. I can just see the edges of what this needs to be. The edges will define the form within.

Comments are closed.