The Gaps Between the Law and Artificial Intelligence

The burgeoning AI industry is raising questions for which federal law has few answers.

By
16
A visitor watches an animated screen at the Mobile World Congress, the telecom industry's biggest annual gathering, in Barcelona. (Photo by JOSEP LAGO/AFP via Getty Images)

Artificial intelligence has rocketed to the fore of American consciousness in the past few months, thanks to the launches of generative AI programs like ChatGPT and Midjourney. But as AI use has grown, so have the regulatory questions around it.

The Senate Homeland Security and Governmental Affairs Committee is holding a hearing Wednesday to address some of these questions. But the lack of AI-specific policy has already started to play out in the legal arena.

Last month, for example, the U.S. Copyright Office rejected an attempt to copyright an AI-generated piece of art because it lacked the “human authorship” necessary for protection. The copyright request was filed on behalf of an AI program, and while a human seeking copyright protection for a work generated through AI might have a different result, the review board noted that copyright law does not outline rules for non-human authorship. But AI will almost certainly play a bigger role in content production in the future, which will spur more such requests.

Similar questions abound in other areas, including self-driving cars and liability laws, as well as AI voice cloning and deepfake software that raise questions about what rights individuals have to their likenesses. In foreign policy, lawmakers will have to determine how best to navigate the minefield of American businesses providing AI research and tech to states unfriendly to the American interests.

Keep reading with a free account
Create a free Dispatch account to keep reading Get Started ALREADY HAVE AN ACCOUNT? SIGN IN
16
By
Comments (16)
Join The Dispatch to participate in the comments.
 
  1. BenP

    We are going to need to update our laws to deal with AI, but our current post-ChatGPT moment has induced techno-fabulist hallucinations that we'd be wise to ride out. The comedown will be gentle and in a year we probably won't even think to notice that almost nothing we care about has been affected by ChatGPT or any other sort "revoluationary" AI.

    ChatGPT is a fun toy that's good at parroting text it's seen before. But a machine that parrots text ain't as useful as the techno-fabulists want you to believe. It screws up basic math, it can't count, it fails at logic tests that toddlers can handle just fine, it has no concept of truth and so you can't trust anything it says, it contradicts itself without realizing it, it is just as confident when making up BS as when reporting facts, if you tell it that it got something wrong when it was actually right it agrees with you and apologizes for the "mistake", yadda yadda ya. Oh and if you ask it to explain something it always writes a closing paragraph that starts "in summary", like it's designed to write 9th grade essays.

    If you ask it a question and you care about getting a correct answer, at best it'll give you a reasonable guess that you should go and confirm elsewhere. It isn't remotely as dependable as Wikipedia, which everyone seems to think isn't dependable. And no, it's not going to "learn" it's way out of all this. It'll stop making the specific mistakes that it's told are mistakes. It can never learn how to avoid the mistakes it hasn't been taught already. It's the extreme version of a student who can memorize everything but never comprehends.

    But go search "ChatGPT" on YouTube and you'll find hundreds of videos telling you that our world will never be the same now that ChatGPT is here. Next year there'll be some other new fun "AI" toy that does neato things we hadn't seen before and is also dumb as a brick on the stuff that matters, and the prognosticators will go wild again and pretty much nothing will change again.

    Collapse
    1. MathewA

      AI won't change everything next year, but between AI and automation there's going to be a whole lot of jobs that are going to go away in the next 10-20 years. And not just blue collar jobs.

      AI will automate away accounting and legal departments. Customer support too. Even a lot of medical jobs including billing etc.

      Collapse
  2. Curt Crowley

    The lack of AI-specific policy is a growing concern as AI technology evolves and finds its way into different areas such as art, self-driving cars, and even foreign policy. I'm curious, do you have any thoughts on what kind of AI-specific policies should be in place to protect people's rights and ensure responsible use of this technology?

    Collapse
  3. Avatar photo
    Show Jay Janney the 🚪!

    Your source is a UD graduate! Matthew is a great person! I wish him the best in his career.

    Collapse
  4. Avatar photo
    George.Skinner

    IP protection for AI-generated work could exacerbate the problem of patent trolls. I could imagine somebody using a parametric approach to flood a space with derivative designs or ideas, applying IP protection, then waiting for another entity to infringe on something.

    Collapse
    1. Avatar photo
      Show Jay Janney the 🚪!

      Blech! That would be very easy to do.

      Collapse
      1. Ingenero.lux

        So would AI attempts to get around those patents by finding and collating everything that might infringe, or in figuring out flaws in the applications that make the patents unenforceable, or more. Plus there will come a point where the patent office won’t have a choice but to crack down on trollery, if only to ensure the patent backlog doesn’t become utterly insane.

        Collapse
  5. Pishposh

    Just wait until someone actually figures out how to upload a real person into a computer. I'm willing to bet it eventually happens. Information is information. If the uploaded person's corporeal body then dies does the computer retain ownership of that former human's assets? What if you could then download that person into another corporeal body? The sheer number of fights I can imagine on the subject are mind blowing.

    Collapse
    1. BenP

      This is one of those things that I doubt we'll ever have to worry about. Just because we can get computers to do human-like things, doesn't mean a computer is doing something similar enough to what our brains are doing that it could become equivalent.

      Collapse
      1. Avatar photo
        Will Bates

        Our neurons work remarkably similarly to a CPU. The rest of our bodies is just the battery and coolant.

        I’d give it good odds.

        Collapse
    2. Ingenero.lux

      I have a feeling that would end with a Butlerian Jihad and the question being mooted by people who got so tired of having all of Twitter electronically arguing forever that they just smashed the machines. That and the younger generation deciding that waiting forever to run or inherit things was ridiculous, so Grandpa Bucket O’ Bolts was going to find out what degaussing feels like.

      Collapse
  6. Avatar photo
    Ben Connelly

    A thorny issue. I’m torn between the “just wait and see what people do and then right down the rules to codify that” impulse and the “better make it all illegal or AI is going to screw up everything” impulse.

    The most important bit to me is the intellectual property protections. Also it should be a crime to make a deepfake porn video of somebody. The Butlerian Jihadist in me wants to make laws abolishing AI, but man has to learn from experience, so we’re going to have to go with a wait and see approach (for now). Maybe AI will be a nothingburger.

    Collapse
    1. BenP

      Sometimes I wonder what our reaction to these computer programs would have been if we hadn't decided to describe them using words that invoke humanity. "Artificial intelligence", "machine learning", "deep neural nets"... these are all pieces of software for prediction and classification that work by detecting patterns in data. If, instead of "artificial intelligence", we called chatbots and the like "prediction and classification algorithms" or "pattern detection algorithms", would our reactions be the same?

      I think you're mostly right about the nothingburger. It isn't that "AI" tools aren't useful for many things - they are. It's pretty amazing that I can write a paragraph using an electronic pen and my computer can convert it to text with high accuracy. Or that I can run my finger over a picture of a keyboard on my cell phone and the phone turns that into the text I wanted to write. Or that I can dictate text and the computer converts it. That stuff is really cool, and also not worthy of the word "intelligence". And, in my view, our decision to call it "intelligence" has everyone falsely imagining that these pieces of software are slowly becoming human. This is like imagining a drone is slowly becoming a hummingbird because it can stay still in the air.

      Collapse
      1. Avatar photo
        Ben Connelly

        But then we wouldn’t be able to make great science fiction jokes about them…

        No, in all seriousness, that is an excellent point! I keep trying to remind people I know that these are just tools, but some people seem to take the word “smart” a little too seriously (as you say, falsely imagining that these things are slowly becoming more human). In my experience, it’s people who have not a lot of technological/scientific expertise/background. So they still treat technology as being sort of like magic.

        “ This is like imagining a drone is slowly becoming a hummingbird because it can stay still in the air.”

        Haha, that’s a great line.

        Collapse
        1. BenP

          Thanks! It was between that and "maybe the Model T Ford will eventually become a horse".

          I think you're right that people with limited background in this stuff tend to see it as more magical, which is understandable. But there are still people who should know better making this mistake, so I blame them. After all, it's computer scientists and statisticians who decided that "deep neural network" would be a cool name for their new class of pattern recognition algorithm.

          Collapse
          1. Avatar photo
            Ben Connelly

            Some AI researchers really do believe their own hype. They called it “deep neural network,” because they believe they are trying to pattern it after the human brain (or what we think the human brain is - there’s a good chance “the brain is a neural network” is like saying an airplane is a bunch of metal).

            Collapse