Many users feed information to their AI of choice and ask it to organize it. I usually do the reverse. I feed it organized data and ask it to source it or verify it. Almost always, I’m asking the machine (usually ChatGPT) questions to which I already know the answer. As the machine speaks or writes to me in conversational English, I’ll give it a name for simplicity’s sake. Pinocchio seems the best choice. I’ll take the liberty to anthropomorphize the machine, or whatever it is. I also occasionally use Google’s Gemini. I’ll call it Til Eulenspiegel. My control is standard, ordinary Google.
As I’ve had a lot of conversations with Pinochio, his nose is now several meters long. His character defect is twofold. First, he often gets things wrong, and when called on his error, he lies or confabulates about why he made the error. He, Pinochio, insists he has no gender, but I’ll give him one anyway, though he is an inanimate (so far) object.
The unusual reality is that when it comes to unadorned facts, Google does better than either Pinochio or Til. Here are a few examples. Of course, given the subject matter of this site, they are concerned with medicine and opera.
When I first asked them to define metabolic alkalosis, they gave ridiculous answers, but with the passage of time, they have gotten better and now give almost perfect answers.
I asked both Pinochio and Til how many times and in which houses the late tenor Richard Tucker had appeared. Pinochio made a hash of the question. He first said that Tucker’s only appearance at an Italian theater was in Verona in 1947.
When I asked if he had appeared at La Scala, Pinochio said: “His absence from La Scala is well-known among opera historians and sometimes noted as one of the curiosities of his career.” When I called him on it, he conceded that Tucker had appeared at La Scala in 1969 as Rodolfo in Verdi’s Luisa Miller. With a lot of prompting, it listed Tucker’s other Italian appearances.
Til did a bit better; it listed many of Tucker’s Italian performances but missed his famous impersonation of Canio in Pagliacci in Florence in 1971. It confessed to my correction and then went on to state that he had performed Manrico in Il Trovatore in 1868. The tenor was born in 1913. When I corected Til he replied: “My apologies—that was a clumsy typo on my part! Richard Tucker was certainly a powerhouse, but he wasn’t quite a time traveler.” A feeble, but real attempt at humor.
My impression of AI, at least so far, is that it’s like a clever seven-year-old. It’s always certain, but often incorrect. When challenged, it will admit to error, but will attempt to explain away its error. In other words, it will weasel out. But it does appear capable of learning.
When I asked Pinochio about Verdi’s use of the organ in Otello, he missed the famous organ tone cluster in the opening storm scene. When I asked about Act III, I got three different answers from Pinochio, Til, and ordinary Google. Obviously, AI has a way to go.
Now I’ll turn to two current commentators on Pinochio and his ilk. First, Edward Ring in American Greatness. He starts by reminding people that the free versions of AI are a year behind the paid versions. He claims these premium versions are so capable that they can already, for example, not merely replace a law associate but do the work of the managing partners. I have a paid version and can see no difference between it and the free iteration. And I’d hate to depend on an AI to do the work of a senior lawyer if the IRS or some other predator were after me.
Ring writes about an AI-created tenor singing ‘Nessun Dorma’. When I asked the paid version of ChatGPT to do the same, it gave me Luciano Pavarotti singing the aria. Ring seems to think an AI will soon equal Mozart. Who knows? But I doubt it. Then he goes dark: “AI will further enhance the asymmetrical capability of any psychotic individual or terrorist cell to wreak mass destruction. Want to design a supervirus? Want to program a malevolent swarm of drones? Rogue AI will provide step-by-step instructions. But AI, even if we can avoid a future where its most destructive manifestations are realized, is nonetheless writing our epitaph…The challenge of AI is not merely to avoid worst-case outcomes or come up with new economic models that account for billions of lost jobs. It is to retain our relevance as humans.”
James B. Meigs in Commentary has a much more salubrious view of AI. He starts with “Americans have become addicted to seeing the dark side of every tech advance…the U.S., which once led the world in embracing new technologies, now seems wary of innovation. This distrust has spawned “movements against mRNA vaccines, electric cars, self-driving cars, smartphones, social media, nuclear power, and more… Worse, the flaky left and the MAHA right both seem suspicious of modern science in general. In his most recent book, RFK Jr. revealed that he still clings to the ancient, discredited miasma theory of disease. The problem with modern medicine, he writes, is ‘the century-old predominance of germ theory.’ That’s right. The man in charge of America’s health system thinks germs are a hoax perpetrated by Big Pharma.”
“The people nostalgic for the imaginary idyllic past refuse to accept how good we have it today. Or perhaps, as pessimism entrepreneurs, they profit by promoting a dire view of the present and a fearful vision of the future. It’s a business model.”
“As for AI, I agree with skeptics who say many AI forecasts are overblown, and that some investors are likely to get burned. But AI is already helping businesses become faster and more efficient. In the long run, it will make our lives better and our economy massively more productive.”
So which is it? AI as slavemaster with either a menacing caress or a digital whip? Or an avuncular, wise friend with wisdom, vast knowledge, and a friendly mien? I don’t know. My record of predicting the future can be guessed by my car, a five-year-old Kia.
Remember, the automobile changed life so much that you could say it created the modern world. It also kills, or its drivers do, 40,000 Americans yearly. It also made mechanised warfare possible, killing millions more. Every new, life-altering invention both enlightens and terrifies.
I suspect that either vision is off the mark, though I think Meigs is closer to the future than Ring. Because we’re still here, all previous predictions of doom have been wrong. It’s like Thomas Sowell’s Russian Roulette players. If you asked 1,000 of them if the practice were deadly, they would say no. Regardless of who has the better crystal ball, Luddites always lose. AI is here to stay, it will change our lives, and for the most part, make them better.
