BY MATTHEW HERBERT
In case you haven’t noticed, there has been a flurry of news reporting recently on the astonishing writing capabilities of new artificial intelligence (AI) applications. I haven’t tested one yet, but you can give one of these apps a writing prompt and some basic parameters, and it will produce a stylish and effective text that meets your specifications to a T.
A trick that no journalist seems able to resist these days is to insert a passage into their article written by the AI they are writing about. And the insertion is seamless; it reads as if the author wrote it herself.
In this fairly typical article from the Atlantic Monthly, we learn that lawyers are already using a leading AI app, ChatGPT, to produce legal briefs. And why wouldn’t they?–“ChatGPT passes the torts and evidence sections of the Multistate Bar Examination,” we also learn.
Robo-lawyers, anyone?
But there is much more to consider. If AI can craft an effective legal brief, AI can understand and adjudicate one too. So then we have: Robo-judges, Robo-juries.
Today, I would like to resist my usual urge to plod through a topic and tell you at distressing length how I think Orwell would judge it an offense on the human spirit. Instead I simply ask you: How big a deal do you think this is?
The ability to generate narratives around which groups of humans can be organized and rallied (to do absolutely anything, from underwriting home loans to pursuing happiness to killing millions other people) is the core skill of being human.
The historian Yuval Noah Harari has been arguing for the last several years that knowledge workers will soon follow the path of blue collar workers in being displaced by machines. The literate class has long been able to avoid thinking about the implications of AI-level automation because there was such a clear difference between what we do–manipulate abstract symbols–and what blue collar workers do–manipulate material things. We may have thought, fleetingly, it was a pity that factory workers were having it tough keeping up with machines, but I doubt many of us actually cared all that much. It wasn’t our problem.
But now it is.
Harari believes we are almost all of us destined to join what he calls the useless class, something completely alien in our history. Humans struggle, strive, create, accomplish. What will we do when machines can (and do) concoct the narratives that goad, instruct and inspire us in our highest ambitions? You might not need to lift boxes or sew zippers into jeans to feel human, but by god you need to do something that serves a higher purpose. What will it be? We can’t all create newer, better AIs–they will do that for themselves soon enough.
So I stop uncharacteristically short and ask for your thoughts. What does it mean that AI can now, instantaneously, write better than 99 percent of us? Is the useless class being created right in front of our noses? Are we in it yet?
Wiser, Braver, More Optimistic 
I’m experimented myself with a variety of algorithmic art creation things since The Sixties. The ability to “fool” an audience is not all that difficult. Here’s a few quick observations I drew from that. These are not firm conclusions, only ways of assessing this:
Audiences can be fooled, after all this is one way we can explain why artist A or B that we think is worthless, is popular. Most people do not read or look etc all that deeply.
Those who look deeply will see patterns, creating links to those patterns. They will complete the comprehension task that “faults” in algorithmic art slips up on, and they may even enjoy the leaps of forming imagination in that process.
If you are fooling an unaware audience who’s response is enjoyment, is that a violation of trust? Many enjoy watching a stage magician. The same tricks used by a charlatan will outrage many.
Actual human creativity is more assemblage and collage that we admit. Some artists will admit this, but point out that the essence of their art is the curation of what it combined. But this curation can also happen post-facto, selecting from algorithmic creations.
What I see in some current AI text examples appears to be culling from search engine tools we may already use, and weighing some selection from their results and putting them together is a way thar fits some rules or order or composition. In the “old days” initial research searches took shoe leather and lots of page turning, and though something is being ordered and selected by an algorithm, I feel empowered by modern tools there. The second step of combing the graphics or discovered text I think I do better, but it’s not unlikely that AI can do this as well as the average human who doesn’t do research. I just worry that no one will sanity check the results for anything that has real consequences. Does the machine check the machine? Quis custodiet ipsos custodes.
LikeLike