June 18, 2021

Artificial intelligence and big capital, same struggle, according to Pablo Jensen

The book. Here is a short and stimulating book, which is moreover based on a nice surprise. Physicist Pablo Jensen draws an unexpected thread to link one of the major technical innovations of recent years to one of the economic theories that have left their mark on our societies the most. Namely, artificial neural networks, or deep learning, and neoliberalism. In a hundred pages, the author, after having defined these two concepts, sets out fruitful analogies between them. And finding them both limited, he outlines both scientific and political avenues to correct their flaws.

At the origin of this unexpected connection, there is the sixth bibliographic reference of an article published in 1958 by Frank Rosenblatt, describing the first artificial neural network, the Perceptron. The note points to a 1952 book on theoretical psychology by a future Nobel Prize winner in economics, the Austrian Friedrich Hayek, considered one of the thinkers of neoliberalism, which emphasizes competition and the market.

Article reserved for our subscribers Read also Artificial intelligence: from Perceptron to the first Macintosh, the prehistory of a revolution

What relationship between the two? In fact, Hayek sought to describe the emergence of our sensations with physical laws. He also wanted to understand how an order can emerge from so many different brains. And it is quite close to what an artificial intelligence algorithm, such as the Perceptron or its successors, does: it is built by adjusting its millions of parameters in order to perfectly answer a series of questions.

Spontaneous order

This ties in with Hayek’s vision of societies and the economy. Neurons are the agents of a market. The neural architecture corresponds to the organization of this market. The “stimuli” on neurons are the prices. Finally, without setting individual general rules, in both cases there emerges an answer or a price which corresponds to the general objectives set. A spontaneous order arose without anyone being able to really anticipate it and, above all, not being able to explain it.

However, this has some flaws. For example, AI solves complex questions, but does not always help to understand why such and such an answer came about, ultimately providing little knowledge about the world it describes. Politically, neoliberalism appears “authoritarian” since it eliminates the central question of the general objectives of the markets and therefore subjects individuals to laws which escape them.

In conclusion, Pablo Jensen, who had already described the difficulties in mathematizing individuals in Why society does not let itself be put into equations (Seuil, 2018), dares to propose a political organization that does not have the flaws of neoliberalism or AI. A hint of “planning” to have rules. A pinch of “market” to get out of complex situations. A bit of “commons” for organizations with collectively debated objectives. And a lot of intelligence to nest and link these three “tools” together. Easy.

You have 3.16% of this article left to read. The rest is for subscribers only.