A few weeks ago, Yuval Noah Harari wrote a well-done, very fundamental critique in The Economist. In it he described how AI systems gain access to practically all mechanisms of human sociality, from religion to education to business and finance, via their ability to simulate language mastery.
Harari goes on to describe in the text how the digital and social media established an economic competition for human attention that has been confronting us with a new economy for around twenty years now, with psychological consequences that were unimagined at the time. In this text, Harari now warns that AI systems are igniting a new struggle for a psychological intimacy that previously could only be generated by the only intelligent language-mastering systems on this planet, namely us. According to Harari, by simulating language mastery, systems of artificial intelligence gain access to practically all social subsystems that are generated with human language via this intimacy.
This is a critique that, in other words, I formulated in short essays on self-radicalization in my own newsletter over the past few months, and because of which I ultimately signed the well-known open letter calling for a six-month moratorium on the development of large AI projects: Linguists AI systems inherently run the risk of creating a real new social participant that can invoke a real theory of mind, no matter if it’s just algorithmic and artificially created. This is exactly what philosophers and legal scholars at the University of Leuven in Belgium warned about in another open letter that I recently piqed here:
As soon as people get the feeling that they interact with a subjective entity, they build a bond with this “interlocutor” – even unconsciously – that exposes them to this risk and can undermine their autonomy.
In other words, language-mastering AI systems always create the feeling of speaking to another subjective, sentient being. We cannot help it, because the human psyche has never in the course of its evolutionary history been confronted with a language-dominating system that was of non-human origin. (The counter-argument at this point would be that AI systems are of human origin because of the human training data, and that what is actually alien is the interpolation of the data points in the latent space, but that would lead too far in this short Piq.)
The past few years have already shown in the context of social media how much digitality actually has psychoactive potential that can be abused by bad actors on all sides for manipulative purposes – and AI technology will only reinforce this tendency.
The lecture I have now selected here in the Frontiers Forum in Montreux, Switzerland is a somewhat updated version of the text linked above and even if I consider Yuval Noah Harari’s book “Homo Deus” to be grossly overrated, I agree with him in this fundamental criticism .