AI is humans. Various aspects of machine learning applications, for example, model selection based on criteria such as robustness and complexity, and feature engineering that leverages data to create new variables that are not in the training set, must be carefully configured by human experts. AI needs humans to function.

Only humans can mitigate some of the limitations of AI systems. Notwithstanding the staggering efforts by many on the path to the realization of human-level intelligence in machines, we have managed to build only weak or narrow AI, which is limited and ineffectual without humans-in-the-loop.

AI remains largely ideological in the sense that its achievements and its aura of possible futures obscure and conceal the social relations that AI assemblage is both produced by and which its uses impact. The optimization of every AI project is external to the system.

AI assemblages emanate from either the reveries of human coders, data scientists, engineers, and data architects who are working within defined project teams, or from such work done by the rest of the world as can be appropriated, and used by project teams elsewhere.

Machines have outperformed humans in image and speech recognition. It is undebatable that in complex games of strategy, machines have defeated human geniuses. But we are quite some distance from applications of machine learning that are fully automated and improve habitually with experience.

In the age of digital capitalism, AI assemblages are rendered with an aura of magic and enchantment. They attract attention as influential, mysterious, and futuristic technologies that are highly desirable. But AI assemblages are also tools that satisfy human needs and are circulated and exchanged on markets for money.

In this sense, they are commodities (Marx, Capital: A Critique of Political Economy, 1906, p. 41, p. 43). In the commodity form, AI has both “use value” and “exchange value”, and becomes an object of commodity fetishism.

Commodities in capitalism are “mysterious”. Commodities are raw materials, the inputs, used to manufacture other products and services that consumers desire. Commodities are those shadowy inputs in production, rather than finished goods that are sold to consumers.

Lithium for example is a specialized commodity. It is produced in limited quantities. This makes supply disruption a significant factor in price fluctuations. Additionally, changes in consumer trends, particularly in consumer electronics and EVs, can disrupt demand and supply.

Investors and traders buy and sell commodities directly or via derivatives such as futures and options. Hence, investors allocate commodities in a portfolio as a hedge against inflation. Marx observes that “at first sight” commodities appear to be “very trivial things” and “easily understood” (Marx, 1906, pp. 81-82). But analysis uncovers that they are in fact, “a very queer thing”.

As a value in use, there is nothing mysterious. But as “soon as it steps forth as a commodity, it is changed into something transcendent”, something ideological by hiding and concealing social relations which humans work to produce.

Lifting the veil that conceals AI, we find values, datafication, dreams, reification, and commodity fetishism.  Reification spawns a particular form of alienation whereby products or relationships are conceived as existing independently of humans.

Fetishism is a term borrowed from anthropologists who apply it to the belief that an object — including a human-made object — can have ghostlike powers or can confer such powers on its owner.

Marx adopted the term fetishism and drew an analogy with the realm of religion, where inputs ignited by the slow fuse of the imagination become autonomous figures endowed with a life of their own, and enter into relationships with each other and with humanity.

As a form of reification, commodity fetishism does not present economic value as arising from the human relations that produced the commodity, the goods, and the services. Marxian thinkers associate reification with an attitude of detachment and nonparticipation.

Reification has to do with the layering of formal rationality in the social world. It is therefore the social process by which something abstract like AI and its entanglements in politico-economics is made into something real or material. It is the process by which the needs of the system obscure the social origins of production.

AI assemblages are therefore bestowed phantom objectivity. They are given an autonomy that seems so rational and pervasive as to conceal their fundamental nature – their relationship with humans.

It is projected that human-level machine intelligence may be achieved by 2062. But even then, this will be just human-level—nothing beyond.  No one can say with confidence that the probability is zero this century. But all things around AI remain broadly uncertain.

Algorithms can solve many things. But there are limits to computation even when one considers the hype around Generative AI. There are claims that such systems have increased abilities to learn and understand at the same level as humans.

But there will always be undecipherable problems that test the limits of computers. Conundrums that have a degree of undecidability. Tasks that call upon human creativity, intuition, and common sense.