I am torn, as many are, I believe, between AI skepticism and fear. On the one hand, I’m unimpressed, and on the other, I think there’s a real existential risk in the development of these technologies. I suppose one way to resolve this tension is to articulate it like this: I don’t think the literal destruction of humanity—our biological species—is likely to occur, but I do think a kind of mass dehumanization is possible and even desired by its creators, although they may not fully realize it or admit it as such. And I don’t think it’s a coincidence that the AI crisis comes in the midst of a broader crisis of the humanities. Whatever else they may involve, the humanities explicitly require and cultivate judgment and experience rather than technique and process.
In Book VI of Nicomachean Ethics, Aristotle distinguishes between the two types of “intellectual virtue” or mental powers that deal with doing and making: techne and phronesis. Techne is often translated as “art,” but for our purposes its sense is better preserved in its derivatives like “technique,” “technical,” and “technology”; in essence, it refers to a set of procedures or fixed rules that can be followed in order to produce things. Phronesis is translated either as “practical wisdom” or “prudence”; it deals with knowing the right thing to do in a particular situation. It’s close to what we might call good sense, judgment, or even insight. Aristotle says that while the young might be talented in mathematics or geometry, they can’t possess practical wisdom because it requires not knowing just general notions but with many particulars, familiarity with which can only come through long experience. Phronesis is involved in deliberation and is the virtue proper to politics and statesmanship.
AI might possess techne, or really it is reified techne, but it can’t really possess phronesis, and therefore it can’t deliberate and know how to accomplish any human goods. It can’t really act as such; it can only behave, or be “agentic” in the ugly terminology of Silicon Valley. The tech world is aware of this on some level, as well. Although many tasks of programming can now be automated, the primary talent leaders in technology now talk about is “taste.” As 20th century philosophers like Hannah Arendt and Hans-Georg Gadamer have elucidated, taste is phronetic: it judges particulars but without a determinate rule that can be given in advance. There are rules of thumb, general notions, but none of these can really be called concepts. For instance, elegance and simplicity are often thought of as hallmarks of good taste. But obviously the application of simplicity alone cannot generate elegance. It can just as easily make something barren, boring, and bland, like a lot of contemporary design. The tendency to apply these aesthetic ideas as if they were simple procedures that can be followed by anybody is why there is so much of what Emily Segal has cleverly called “tasteslop", which I think is something like the idea of good taste reduced to algorithmic regularity.
My friend Jay Caspian Kang is writing a series on the concept of shame and how it has declined in our culture for The New Yorker. The first example of shamelessness is how NBA teams “tank” in order to receive more favorable draft picks. The second installment looks at how AI might further degrade our collective sense of shame. He points out that the AI boosters have a very different process of moral reasoning than our usual assessment of what is praiseworthy or blameworthy: they apply some kind of clearly delineated, abstract, mathematical rule. In both cases, the types of goods internal to the practice—being as excellent as possible at basketball and being a good person—are subjected to a technical process in order to accomplish an abstractly and quantitatively defined outcome. As Gadamer put it, this is part of a modern tendency that “degrades practical reason to technical control.”
What unites the entire “culture” of Silicon Valley is that type of degradation: the reduction of everything to an algorithmic sequence of steps.
What is ironic or maybe merely unfortunate of the contemporary situation is that precisely when robots might be able to take over the merely technical or deductive parts of thinking is that the importance of judgment should become even clearer. We could enter a golden age of the study and contemplation of human accomplishments. Instead, we seem to be mulching it with all the rest. One thing I’ve noticed with young people is that they seem to know—and want to know fewer—particular things: far fewer names of famous politicians, events, artists, athletes, or works of art that are remembered for their excellent qualities. The memory and preservation of those things is not a small thing; It is culture itself.
Democratic, publicly debated decisions on how and when to apply and develop artificial intelligence are important because they preserve our judgment of what’s proper to do. This is exactly what communities should deliberate and decide together. And the simple reason “it’s creating ugly things” is not as thin or meager as it might seem; it’s pretty much all we have if we want to remain in a human world.
Aristotle's phronesis is similar to how Bernard Stiegler uses "savoir vivre," or knowledge of how to live. Savoir vivre isn't knowledge in the sense of facts and information; it's an organic faculty of judgment that can't be reduced to a rule or pattern. Stiegler argues that technologies that exteriorize memory (what he calls "mnemotechnologies") degrade savoir vivre, and that we're undergoing something like a savoir vivre extinction event right now, which is part of a broader process of mass proletarianization via tech.
Some academics are writing about this. Sun-Ha Hong's essay "Prediction as Extraction of Discretion" is good: "prediction is not primarily a technological means for knowing future outcomes, but a social model for extracting and concentrating discretionary power. Prediction is a ‘relational grammar’ that governs this allocation of discretion: the everyday ability to define one's situation. This extractive dynamic extends a long historical pattern, in which new methods for producing knowledge entail a redistribution of decision-making power." https://journals.sagepub.com/doi/10.1177/20539517231171053
There are a lot of useful concepts for this faculty of judgment (phronesis, savoir vivre, discretion, etc). However we choose to frame it, it does seem to be one of the biggest casualties of AI and compulsive internet use, but also one of the most difficult to describe and vouch for -- especially to people who've already lost theirs, or never had much of it to begin with.
The problem of AI as a cultural force is insoluble, not because it is algorithmic but because its fundamental structure is based on the sum of all accessible human writings. The structure (billions of weighted connections between words) that the AI employs does not favor Aristotle, Confucius, et al, but a lowest common denominator. Familiarity with the shelves of a comprehensive library or the internet reveals this to be in substantial part amoral, gullible, and above all tasteless -- lacking in judgment. "Training" and "alignment" are supposed to rectify this But, even aside from the fact that these are in the control of the AI owners, some of the worst people on the planet, it is proving impossible to overcome the gravitational pull of the total mass of human writings.