How to use AI to ensure a more human society

A century-old article in The Annals of the American Academy of Political and Social Science outlines a central aspect of human society with this beautiful paragraph:

...the most fundamental contribution of aesthetic training to citizenship and democracy is the common and intelligent love of the beautiful which makes possible the finer forms of social intercourse and is essential to the most manysided enjoyment of individual leisure. Curiously enough it is in a free system of public education rather than in prohibitive material and social conditions, that aesthetic enjoyment finds its real limit. The only obstacle which still stands in its way is a lack of that good taste and manysided interest which education alone can develop.

This is precisely also what AI cannot do. It can pretend to appreciate a work, it can regurgitate judgement from a smattering of criticisms humans have penned over centuries, yet it can not demonstrate a satisfaction developed solely through nuanced appreciation for the so-called finer things in life.

AI is simply a statistical reordering of existing knowledge that relies on one specific, age-old outlook: that everything new can only ever be a permutation of everything old; that novelty is pretence. That the appreciation AI demonstrates for things is similarly pretence is evinced by its supposed appreciation for the writings of Paul Krugman, cucumber sandwiches, the paintings of Monet, and Cabernet Sauvignon—at least half of which it has almost certainly not experienced but somehow appreciates nevertheless.

So if what life is really all about—like the Utopia of Star Trek—is the ability for individuals to socialise, follow their independent interests unconstrained by necessities for living, appreciate and produce art, food and wine, and debate on philosophical sophistications that could make life better for us all (while also better guide AI itself), then we would have to look to making AI do work that we can truly benefit from neutrally.

In recent news there are two polar opposite examples of this which suggest we still really have no idea what we are doing. First the good: AI was recently used to divise new antibiotics based on a large dataset comprising almost every single known microbe in existence. This is an example of the correct and efficient use of large-scale computation (and also begs the question as to whether the term ‘AI’ is, in fact, simply corporate-speak).

As a counter-example there is the absolutely outrageous suggestion that AI can be used to send our “AI avatars to meetings for us”. Unlike the example with antibiotics where, like a wrench or a lever, new technology was used to ease work that humans would have been inefficient at performing, the AI avatar suggestions treds on the very appreciation for finer aspects that human society should retain in human hands. It attempts to replace humanity. Of course coporations understand this just as much as anyone else which is why Zoom CEO Eric Yuan attempted to sell this as a means to “free up time to spend with family”. In reality, the company will simply give you some other work to do during this ‘saved’ time.

But if freeing up time for family is what we really want, we have better ways to do that than by having AI train on individual preferences and mannerisms—inching closer to human likeness. We could use AI to replace executives at the very top of a company, like Eric Yuan for example, so that for every situation a number of possible decisions and consequences can be detailed for consideration by a board of human supervisors who do not have financial interests dictating their choices. Companies can be a force for good while daily tasks and entire meetings can take place within an AI network until human intervention is actually needed.

Not only do we now have models to make AI replace CEOs and others in the C-suite, we have had such ideas since the 1980s. The only roadblock to this development are CEOs who would obviously not want to prioritise replacing themselves. (This should remind us of tobacco companies in the ’60s and fossil fuel companies today, both of whom campaigned against better solutions in the interests of self-preservation.) Meanwhile with prudent AI employment humans can spend time doing more human things. And that is precisely why legislation at this point is a critical necessity: AI has the potential to be a force for good, for constructing the most-promising gateway yet to a human society, just not as long as it is in the hands of corporations.