Transcending AI

Discussion of genAI in my circles, which are largely academia, tends to fall into three camps:

1. Deflating #AIHype
2. Considering ethical and regulatory guardrails
3. Finding positive uses

Now, I find 3 very puzzling. It is a new technology, it may be useful, it may not be. Why put all this effort into trying to find good uses?

Academia has taken on board and found many good uses, over the years, of other new technologies, including email, the web and social media. In those cases the good uses were pretty obvious and largely very easy to achieve. The misuses, bad consequences, and general problems arising from adoption of the new technologies were far from obvious and took years to emerge.

With genAI, the reverse is true: the misuses and problems it can create in research and teaching are obvious. The good uses far from obvious. Those that have been suggested so far amount to: this is a fun and zeitgeisty way to do things you have rather better ways of doing already, but it has risks they don't have.

On virtue ethics

One of the problems with this way the debate has been structured is that ethics comes in as a side-constraint, alongside regulation. Now I would be the last to deny that there is hugely important work to be done thinking about and articulating the ethical pitfalls in various uses of genAI. That is a big chunk of my day job, after all.

However, this is a narrowing of ethics to a domain of 'thou shalt not', of restrictions on natural, or at least socially acceptable and culturally normative, human behaviours, such as making money, having fun, and generally being part of a consumer society. It is important to do this, because we don't want society to 'sleepwalk' into a very bad place, but ethicists can offer so much more.

Virtue ethics approaches try to do this by grounding ethics in a conception of human flourishing. But the structure of virtue ethics has been - at least since Aristotle - to think of human flourishing as something already determined, a state of being we may need to discover and find our way to, but the nature of which does not need to be created. On this view, each of us needs to find our own way to a condition of flourishing, and that may differ for different people, but there isn't a project of collectively creating a conception of what it is to be human and to flourish as such.

Transcending AI¹

One thing we have already learned from genAI is that a certain type of linguistic skill, producing texts which are well-structured, grammatically precise and often lucid, is not constitutive of intelligence and understanding. It was for centuries a good indicator of a certain level of formal education, and success in formal education generally requires intelligence and produces understanding.² The recent boom in genAI has produced a situation where this is no longer such a good proxy for those valuable traits.

This creates a situation where we need to think long and hard about what it is we should value most in our limited, finite human nature. We need to think why we are so impressed with AI's problem-solving abilities, or the speed of its responses. If we value too highly those characteristics in humans, then we run the risk of reducing our humanity to being fleshy machines.

We need an ethical conversation about what it is to be human, about what are the features of humanity that we should value and cultivate. This conversation needs to be careful and thoughtful, to take into account the wide diversity of human lives and abilities. Historically humans have tended to valorise inherently exclusive traits such as problem-solving and 'quick thinking', but also neurotypicality, diligence, ambition, cis gender and the (straight) 'nuclear family'. We value things which create an ideal of humanity which is essentially non-inclusive, which will by its very structure place some lives as better than others.

The rapid emergence of Ai in our society gives us the opportunity to start again, to observe that so many human achievements we had prided ourselves on are actually, from the cosmic perspective of advanced technology, rather mundane. We need to think about how to identify, cultivate and learn to value those features of being human which won't suddenly seem less important when the next technology replicates them.

We need to learn to transcend AI.