Google Stock Tumbles 8% After Its Bard AI Ad Shows Inaccurate Answer

This content has been archived.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

From Investor’s Business Daily:

Alphabet (GOOGL) tumbled Wednesday after Google’s parent company published a new ad for its Bard artificial intelligence chatbot that offered an incorrect answer. Google stock fell more than 8% after the ad fluke.

Google posted a video on Twitter demonstrating the “experimental conversational AI service powered by LaMDA,” the company wrote. LaMDA is Google’s Language Model for Dialogue Applications, which applies machine learning to chatbots and allows them to engage in “free-flowing” conversations, the company says.

In the advertisement, Bard is prompted with the question, “What new discoveries from the James Webb Space Telescope can I tell my 9-year old about?”

Bard quickly rattles off two correct answers. But its final response was inaccurate. Bard wrote that the telescope took the very first pictures of a planet outside our solar system. In fact, the first pictures of these “exoplanets” were taken by the European Southern Observatory’s Very Large Telescope, according to NASA records.

Link to the rest at Investor’s Business Daily

It seems that this is not a good day for Google’s image at all. A quick look at Twitter for trending Google mentions disclosed the following as a prominent item:

14 thoughts on “Google Stock Tumbles 8% After Its Bard AI Ad Shows Inaccurate Answer”

  1. That last guy – the Engineer that got caught up in the layoffs?
    He had it right – you are not your job. And, if you fall down, rest a while, then GET UP.
    Just another reminder:
    Don’t spend all your money. Live frugally, as you WILL be out of a job at some point (or have other life crises).

    • Try this:

      https://www.businessinsider.com/chatgpt-ai-compared-to-new-bing-search-chatbot-answers-2023-2

      Business Insider did a 20 Questions comparison between ChatGPT and new BING. They are very interesting questions and the answers different enough to see they are based on comparable tech but one (BING) is more focused on the real world of ’23. Of note:
      It accesses current (realtime?) data.
      It knows copyright and refuses “in the style of” requests.

      “Insider asked Bing to write an article in the style of Business Insider. Bing
      What I liked: The clear answer explains why it can’t write an article.

      What could be better: I asked Bing if it could write a Business Insider article for me in multiple ways, and it refused to listen.

      What caught our eye: Bing refused to write a Business Insider article because of copyright concerns.”

      They threw some pretty sophisticated questions and it handled them pretty well.

      There’s a lot of hype going around this week but this article suggests there’s some meet to New BING as a *search* tool. I can’t wait to get my hands on it and give it a try.

      • I don’t have the time to play with ChatGPT right now, but I suspect that there are specific questions I can ask that can only be answered by my posts. That means ChatGPT is only echoing back our own information, paraphrasing if you will.

        That did not end well for Narcissus.

        • Of course it only returns what already exists.
          It’s just an information tool, a new front end to the “global library” that is the internet. All ChatGPT does is remix what it saw during training.

          Doesn’t mean it isn’t useful.
          It went from zero to 100M active users in two months and it’s still growing. It’s not all just curious testing, either. An unspecified fraction of the users is getting good use out of it. Remixes have their uses.

          But Google’s panic isn’t from ChatGPT itself. They just rushed out a copy of it, after all. What has Google spinning around like Curly Howard is how MS quietly integrated a newer GPT model with their new search engine, PROMETHEUS, to ground it and provide current data. Where ChatGPT 3.5 was trained on 750Billion parameters, the Bing version was trained on 100Trillion.

          https://m.youtube.com/watch?v=T2iZPRif2i4

          Two different creatures, as Business Insider’s test proves.
          Me, I have no interest in ChatGPT but New Bing looks useful. That one I signed up for.

          Hype aside, MS isn’t billing it as a creation tool (they’re working on a different product for that: microsoft CREATE) but as a search assistant. As I said in the other thread, it’s most useful features are its citations (so you can verify where it came up with the answer) and its “stickiness” that asks and lets you ask followup relevant questions.

          Its not an Oracle.
          Just a new class of software.
          And like all other software, what you get will depend on how you use it.

  2. If your intelligence is echoing what the internet says, then you will be as right as the internet is – which is not necessarily perfect.

    • Which is something both MS and Google have factored in.
      In diferent ways.

      In fact, Google was bragging in their demo about their N.O.R.A. subsystem to address stuff with “No One Right Answer”. Which is why the demo error was so costly; the right answer is all over the internet so to surface such a wrong answer the bot had to “hallucinate” it, something all “AI” developers are actively addressing.

      Microsoft, on the other hand, demo’ed a bad answer on purpose to highlight their “challenge” button. They don’t pretend their tech is 100% yet. It is similar to the way they’ve successfully deployed their game streaming tech despite its limitations. Google’s own game streaming failed for a variety of reasons, one of which was overpromising and rushing it to market half-baked.

      As a result, the BARD failing was seen as same old, same old.

  3. Here’s a video contrasting Microsoft’s intro of Bing and Edge with GPT tech against Google’s intro of their Bard chatbot:

    https://m.youtube.com/watch?v=5X1O5AS4nTc

    Their summary is particularly on the spot: in this product “war” Microsoft has little to lose (a few billion bucks, say a week of *net* profit) while Google has their entire business model (and 60% of their net profit) at risk. Their first step was a pratfall that cost them $100B in one day.

    The game is just starting but it started with a knockdown, in boxing terms.

    • Thinking on it, Google’s BARD is following a long tradition of Google betaware failures:

      https://killedbygoogle.com/

      281 and counting: 51 apps, 195 services, and 21 pieces of hardware.
      They’re great at selling online ads but internally developed software and platforms? Not even close.

      Their biggest problem is that most of their products–hardware and software–are just Me2 copies of somebody else’s product. Even google search falls into that category only winning because Yahoo failed to update ALTAVISTA, the first keyword-driven search engine.

      The trick work once but not since.

  4. What in the world is the “Civics and Election space”?

    I have some theories, but am too busy/lazy to try to verify…

  5. Instead of LaMDA they should have called it LaMBadDA. 😀

    (And this is the chatbot the idiot thought was sentient?)

    Still, as Meatloaf sang “Two out of three ain’t bad.”

    Yesterday, demonstrating the language model for their web search service, MS asked it for a short poem about Google.

    “ChatGPT answered:

    “There once was a search engine named Bing

    Which was struggling, not doing much thing

    But with ChatGPT on its side

    It finally got its pride

    And left Google searching for its next fling”

    https://www.yahoo.com/entertainment/microsoft-report-chatgpt-asked-write-223051627.html

    Not exactly literary but I’ve seen worse.

Comments are closed.