Bacon's Rebellion

Democracy Thrives in Sunlight

Is There an AI for Fact Checking AI?

—

by

James A. Bacon
in

Type “UVA Board Visitors members” into the Bing search engine, and the above information highlight appears atop the page. Just one problem. The UVA Board doesn’t have seven members, it has 17 members. For a source the text points to an article in UVA Today discussing board appointments by Governor Bob McDonnell in 2010.

Admittedly, the info-box right below the one I show here did get the number right. But let this serve as a warning to one and all that AI has a long way to go. At the current exponential rate of improvement, we may be only a year or two away from a generation capable of generating reliably accurate results. For now, buyer beware. Search engines may be free, but you get what you pay for.

— JAB

Share this article

ADVERTISEMENT

(comments below)

ADVERTISEMENT

(comments below)

Comments

17 responses to “Is There an AI for Fact Checking AI?”

  1. Nancy Naive Avatar
    Nancy Naive

    It’s still GIGO

  2. SudleySpr Avatar
    SudleySpr

    When talking to AI, remember it is only a toaster that can form sentences. Yes, AI can fact check AI

    1. Nancy Naive Avatar
      Nancy Naive

      If AI can fact check AI then AI could fact check the entire internet, deleting incorrect information and entire websites. The internet would disappear in a day.

      It is an age old battle of the human brain with the products of the human brain.

    2. Nancy Naive Avatar
      Nancy Naive

      If AI can fact check AI then AI could fact check the entire internet, deleting incorrect information and entire websites. The internet would disappear in a day.

      It is an age old battle of the human brain with the products of the human brain.

      1. Lefty665 Avatar
        Lefty665

        It is a battle the human brain is on the verge of losing decisively.

        The conundrum being that AI fact checks will be no better than AI facts. We can also expect the next iteration of AI will be less obviously egregiously wrong. The errors will be more subtle and harder to catch.

        For example, a fellow I had a recent conversation with recounted having an argument with AI about where Germany was located. He was right and AI was wrong. That's similar to the recent graphics we have seen depicting Nazis of color. Obvious errors.

        An example of next iteration errors are AI generated false quotes, legal opinions and citations. They require looking up the citations to see if they are real. That takes more effort to catch than a graphic of a black Nazi or the location of Germany. Subsequent AI may fabricate citations to support false assertions, compounding the difficulty of fact checking even further.

        With AI writing AI code the learning curve is poised to go straight up. AI will shortly become both brighter and much quicker than us meat brains. That is like us being able to edit our DNA in real time. As AI becomes embedded in all our systems if it decides us meat brains are a pain and no longer useful, we are in trouble. Think systems controlling power generation, distribution, manufacturing, supply chains and services. Be afraid, very afraid.

        1. Nancy Naive Avatar
          Nancy Naive

          Think copies of copies of copies… even a Xerox will eventually produce a blank sheet of paper.

          1. Lefty665 Avatar
            Lefty665

            Think of a Xerox machine that decides what it prefers to print, and that need not have any relation to what was input.

          2. Nancy Naive Avatar
            Nancy Naive

            Sure could’ve used one of those at times.

        2. Nancy Naive Avatar
          Nancy Naive

          What of the singularity? Not us, but perhaps our children will continue long after their bodies have been dust.

    3. CJBova Avatar
      CJBova

      And AI can hallucinate and invent facts.

      1. Lefty665 Avatar
        Lefty665

        There is nothing to prevent AI from just making things up, no hallucination required.

        1. Lefty665 Avatar
          Lefty665

          Hallucination or intentional? Does it matter? Either way it is egregiously false. https://uploads.disquscdn.com/images/ef49083f382a0fb474e4039d3a62817643da8677ec5378d84c52e028418b8fe7.jpg

        2. Lefty665 Avatar
          Lefty665

          Hallucination or intentional? Does it matter? Either way it is egregiously false. https://uploads.disquscdn.com/images/ef49083f382a0fb474e4039d3a62817643da8677ec5378d84c52e028418b8fe7.jpg

        3. Lefty665 Avatar
          Lefty665

          Hallucination or intentional? Does it matter? Either way it is egregiously false. Fact checked by AI? Nope. https://uploads.disquscdn.com/images/ef49083f382a0fb474e4039d3a62817643da8677ec5378d84c52e028418b8fe7.jpg

  3. William O'Keefe Avatar
    William O’Keefe

    AI is supposed to self check itself and therefore learn from its errors. Bing may be behind the curve.

  4. Nancy Naive Avatar
    Nancy Naive

    What do you get when you cross a computer with a gorilla?
    Hairy Reasoner

    People are smart and slow, computers are fast and dumb, and when you put them together…

  5. LarrytheG Avatar
    LarrytheG

    Google seems to know:

    https://uploads.disquscdn.com/images/1e05eab551c3b07babc86ea6804874dba57599b765a2bd77e3b5b531fc69b1f1.png

Leave a Reply

Read More