Dario Amodei’s dreamy liberalism

This piece is a twin to another which looked at the failure of various modern liberalisms and was a plea for an alternative: conservative pragmatism.

Here, I discuss what I call Techno-liberalism. I hadn’t noticed it properly until I read Dario Amodei’s 10,000 word essay, “Machines of Loving Grace”.  Its author is the CEO of a leading AI innovator, Anthropic and thus both the right person to listen to about AI, but also suspect when he opines on huge related human matters.

Firstly, we are lucky to have him. He has bothered to articulate his underlying thinking and to sketch out where he has confidence and where he has only aspirations for the AI he is inventing.

Dario’s thinking is only firm and realistic on the nuts-and-bolts advances that AI can accelerate, but is far more speculative on the really big stuff about AI and the future of human society.

I identify four main weaknesses in Dario’s thesis.

One is that in one hop he leaps from AI’s technical advances in, say, biological science and on to vast social and political advances he sees as potentially available to our future if AI goes well. This leap soars in silence over the existing and potential crisis intensified by AI. He seems not to care that whilst it may have energised invention, AI has also energised plagiarism, deepfake, fake news and the creation of pseudo-identities. Its impacts on personal psyches and thus on humanity as a whole may well be very great. These developments will, I suggest, reinforce the need for a revival of authentic, characterful, reliable, real-world adulthood in leaders and citizens alike. 

Dario Amodei’s second failing seems to be that he is far more nuanced about AI and the future of humanity than he is about AI’s ability to transform scientific research. And yet he keeps insisting that if only enough kindly bright people aim to keep a grip on AI, it can be tamed, harnessed and rendered humane on the grandest scale. Contrariwise, I struggle to see how regulation can keep up with a technology so fundamentally transformative. Indeed, some of Dario’s thinking seems to be an invitation to abrogate human responsibility to the very computer code we know to have energised illiberal and debilitating forces.

The third main weakness I detect in Dario’s thesis is that he doesn’t notice that liberal democracies are going to have to learn how to be sufficiently authoritarian to censor the worst effects his AI will wreak in human minds. Even when they become intrusive, liberal states will probably still fail to curb AI’s worst incursions on the best of our centuries’ old normalities.

The fourth weakness I detect matters the most, probably. Dario strongly hopes that AI can buttress existing liberal democracies, and also be an engine of propagation for the model. But he mostly describes AI’s potential as a matter of applying computer rationality to achieve human consensus. I fear that Dario is that variety of Bossy Liberal whose habits have led to the current crisis seen in many liberal democracies. Liberal democracies exist, not to contrive settled consensus, but to reconcile deeply competitive views within a perpetual and shifting dynamic tension. So I fear he has picked both the wrong version of liberalism, and the wrong technology to make it a reality.

If you don’t believe me on this fourth issue, try looking at JS Mill’s view on truth-finding. He was a Bossy Liberal, and a believer in a new religion of Humanity, but he was undreamily sure that argument was the way to understanding. In argument, he says, participants gain “the clearer perception and livelier impression of truth produced by its collision with error”.

My conclusion is that Dario (“upholder of the good”, in Italian) Amodei (“lover of God”, in Italian”) may be little more than a born-again 1960s West Coast techno-hippy in the mode of the poet whose title he takes as his own. In the degree to which this is so, then he is hoping that the hippy dream of a loving anarchy, perhaps machine-led, can come to the rescue of humanity.[ Footnote: Lest I sound sneery, I should say that I rather admire the career of Stewart Brand, who has been a pretty serious bridge between the back-to-the-land ethic and up-up-and-away technical ambition. See his Wiki entry and a Stewart Brand essay here.]

I would prefer to put my money on hustings, argument, evidence, and elections. I don’t imagine that computers can do much better than our existing liberal democracies, though they may try to be dictatorial. It is a dreadful truth that in Italy (1922-45), Germany (1933-45) and Spain (1939-75) varieties of demagogic fascisms were replaced by a return to the very liberal democracy the dictators had sought to extinguish, not revitalise.

Politics is a competitive business which achieves about as much co-operation as humans can bear. (Arguably, that is what democratic capitalism, or social democracy, does in the political economy sphere: it holds the ring for the fight between capitalistic competition and the world of kindlier co-operation.)

Any intellectual weakness we see in Dario Amodei doesn’t matter much. We do not expect to see a serious predictive capacity in the historic thinkers and inventors who turn out to have made our world. The cave persons who discovered fire, the Renaissance humanists who tore into superstitious religion, the astronomers and navigators who forced Europe onto Australia and the Americas, the Chinese who gave gunpowder to the world’s warriors, the WW1 armourers who developed the machine gun’s mass slaughter, the physicists who gave us the atomic bomb, even the socialists who gave us political and intellectual revolutions, are all alike. They none of them could say what would happen next.

Nor do inventors have to be philosophers, let alone that vital subset – political philosophers.

I imagine, however, that it will be interesting and informative to see where Dario Amodei travels as a special kind of public intellectual. Will his (qualified) big-picture optimism last? Will this Techno-liberalism prove satisfying?

My own feeling is that when the universe produced consciousness out of carbon – when it produced human consciousness – it produced a risk-taking species which might fairly cheerfully contemplate its own extinction if that’s what the advance of consciousness required. AI might be the universe’s Next Big Thing in the way of consciousness, but even that would tell us little about any human balance sheet it might enhance or disrupt.

I don’t really believe Dario is likely to be successful in his ambition to see AI regulated so as to do no harm. I imagine we will try to tame it, and that we should certainly try. Whatever. The whole enterprise will remain a huge risk, but as a species we never turn away from this sort of excitement. History is one damned thought after another and they can’t be unthought.

My more detailed comments on the Dario Amodei piece are in the identical downloadable Word and PDF documents I append here. I have taken some striking quotes from the original and added what I hope are pertinent comments.

Leave a comment

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Publication date

11 January 2025