In the stunning mural by Polish multidisciplinary artist Goshka Macuga, “Make Tofu Not War” (2018), exhibited currently at the Prada Foundation in Milan, one sees one set of images if one looks with the naked eye and something else if one looks with 3-D glasses. I took this selfie in 3-D glasses. Wearing these ironic tributes to old technology (audiences first wore 3-D glasses to screen the 1920s silent film The Power of Love), the mural behind me becomes more political and pointed, with the resulting text reading: “Democracy Must Be More Than Two Wolves and a Sheep Voting on What To Have For Dinner.” That’s how the text reads in my selfie. However, when I tried to re-size the selfie using ChatGPT, the text reverted to what I saw without the funky 3-D glasses on, then when I reproduced the newly proportioned image, it seemed to flip back again. Now, I’m not sure what I’m seeing or not seeing as Macuga’s powerful admonition (sometimes, like so many things, attributed to Benjamin Franklin, among others) deteriorates into a confusing mish-mash of unresolved words.
Think of that as a metaphor. What do we see when we look at the world through AI glasses? Is it want we expect to see? What we want to see? What we hope to see?
I’m especially attuned to this problem today because the New York Times just published an article on the renowned “deep fake” detective and authenticator Hany Farid who acknowledges that he–with all his knowledge and all of his equipment–can no longer tell the difference between a “real” image and an AI-enhanced, manipulated, altered, or generated image. If Farid can’t, what about the rest of us?
The problem (and I know I sound like a broken record) isn’t that AI has some magical superhuman powers. It is that we humans have abdicated our responsibility for safeguarding, regulating, and in other ways controlling the impact and effects of the GenAI technology currently being developed and released into the world at breakneck speed… and without any guard rails. Those developing AI have not included in its training ways to safe guard or authenticate the information the AI tool produces. We are living in an age of extreme oligarchy and kleptocracy when the wealthiest AI developers and some of the most corrupt political leaders in the world are working hand-in-glove not to protect or safeguard the rest of us but, seemingly quite intentionally, to exploit all of us in all the ways imaginable and unimaginable. “Two wolves and a sheep voting on what to have for dinner”. . . and guess who’s dinner?
Needless to say–but it’s important to say it over and over again–it is long overdue that policy makers insist on safeguards and guardrails. If Hany Farid cannot tell the difference between real and fake (AI generated) information anymore, than why isn’t it illegal for any information to be disseminated without authenticity records. And please don’t tell me that someone could then fake out the authenticity certification. If the AI is trained well enough to produce indistinguishable misinformation, it can also be required to be trained to authenticate its information in ways that cannot be manipulated.
We’ve heard a lot from the tech entrepreneurs about how AI is “superhuman.” But AI cannot be both “superhuman” in the dangerous content it generates but then incompetent, weak, incapable and “subhuman” when it comes to being able to restrain itself, to work within limits set either by the corporations that train the technology or the governments and policy makers who require restraints.
It’s hypocritical, a deception, that the technology is too powerful to be tamed. It’s equivalent of saying that there is no way to control the fastest car today (let’s say the Koenigsegg Jesko Absolut which can reach a speed, theoretically, of over 300 mph). Of course the KJA is incredibly powerful. But that does not mean we have no power to prevent the KJA from zipping down the highway at full speed. We all know that such would be insane. Just because a car can reach those speeds, does not mean they should be allowed to reach them–and especially not on highways that were not designed for them, racing between other cars not built for such speeds. We create safety regulations, speed limits, that require cars to travel under, say, 65 miles per hour. All cars. Even ones powerful enough to go 300 mph. That’s common sense. Why then do we allow AI these boundary-less, unregulated, un-policed privileges?
AI is no different than any other technology. It can be used as a powerful tool for good or a powerful tool for evil. And it is up to us to determine when, where, and how it is used, who uses it, for what purpose–and what the punishments will be for those who violate these rules. That’s what “speed limits” are for.
According to the Times article, Hany Farid, one of the most popular professors at UC Berkeley, can’t stand living that close to Silicon Valley any more. He and his wife Emily Cooper, a professor of vision scientist, have moved to Vermont and are now teaching at Dartmouth. “I can’t stand this place anymore,” Farid confessed to the Times reporter. “These major tech giants will burn everything to the ground as long as they’re making a profit. They’re not interested in anything that’s going to slow them down.”
“It makes me anxious for our students,” Cooper added. “It’s starting to scare me.”
It scares me too. All of it. And so does Wall Street crowing over Elon Musk now becoming, officially, the world’s first trillionaire, the richest human being to ever walk the planet–and an avowed white supremacist who claims responsibility for helping bad actors become important leaders all over the world.
As activist and artist Goshka Macuga reminds us: “Democracy Must Be More Than Two Wolves and a Sheep Voting on What To Have For Dinner.” Yes, yes. It must be. But who, exactly, decides?