Make the present human first
A response to Scott Alexander
In the discourse surrounding “superalignment” or “frontier AI risks” one recurring theme is the idea that “AI should serve humanity”, or that “humanity should stay in control”. This is a problem that is acknowledged to be difficult—if the imagined system is capable enough (and let us not put too much weight on the word “intelligence”), we might find ourselves in the position of a cockroach trying to “align” a human—at best it is tolerated but it is hard to imagine how it could be “in control”.
A few weeks ago, renowned writer on the topic Scott Alexander put the question thus: “should the future be human?”
In this article he struggles to find an argument for why he has a nagging feeling that the emergence of superintelligent AI is somehow bad, even if it is harmless, if it appears as individuals, and pursues art, philosophy etc. He writes:
Even if all …. ends up going as well as possible - the AIs are provably conscious, exist as individuals, care about art and philosophy, etc - there’s still a residual core of resistance that bothers me. ….
Imagine that scientists detect a massive alien fleet heading towards Earth. We intercept and translate some of their communications (don’t ask how) and find they plan to kill all humans and take Earth’s resources for themselves.
Although the aliens are technologically beyond us, science fiction suggests some clever strategies for defeating them … If we can pull together a miracle like this, should we use it?
Here I bet [almost anyone] would support Team Human. But why? The aliens are …. conscious, individuated, and have hopes and dreams like ourselves. Still, humans uber alles.
Is this speciesist? I don’t know…
But “the resistance” the author feels is that both in the case of the alien fleet and superintelligent AI, there is no free decision made on the part of humanity. Scott Alexander, rightly, wants to be free, and senses, correctly, that no one is freely deciding to create AI.
If humanity deliberately, through great effort and investment, develops a technology that replaces it, then sorry, but it is already not in control. But what would control look like?
To use the overwrought example, bees are not “in control.” They coordinate to create structures and defend against attackers in the “interest” of their “community.” But individual bees are not free (they are unable to seek their own happiness and are not protected from their own biological character and from the whims of the group).
Humans in modern societies aim to be free as individuals. Modern civil society was set up with the intention of protecting the individual’s right to seeking their own happiness from the oppression of custom and tyranny. Modern commodity production is an astonishing historical accomplishment—aligning the individual’s pursuit of their own happiness with the interest of the group through production for exchange, through production for others’ needs. Through social production.
The revolutions of the “long” eighteenth century (1688 in England, 1776 in America, 1789 in France) enforced these ideas, emancipating men into society and abolishing old systems of caste, estate and serfdom.
The social freedom achieved, however, turned out to be self-contradictory. The system of commodity production that guaranteed that individuals can have their needs fulfilled if they fulfil those of others—if they contribute labour to social production—is overturned by the application of science and technology to production. Simply put, with the advent of machines in production, the “dead labour” embodied in machines enables their owners to profit from the work performed by others. Machines set up a contradiction between socialised production and private appropriation.
Note that this is possible only on exactly the same social and economic basis as the social freedom achieved in the revolutions. In fact, it is a necessary and rational consequence of it.
Machine production at its most developed enables the entirety of society to provide their unskilled labour in the production process and receive payment in exchange, for labour to become a commodity tradable for money. At the same time, though, it reduces previously skilled work into unskilled work. It turns independent producers into wage labourers—or forces them to participate in the game of producing ever-more efficient machines that can enable the work of others.
There is no bad intent necessary. The process grinds on because it becomes self-reinforcing: the logical thing is the competitive improvement of the technology of production to satisfy one’s own needs through the satisfaction of others’. The process creates haphazard technological development and social discord, as productive methods and ways of life are constantly overturned. This breeds discontent, which takes the form of a new violent politics. Classes of the population that enjoy relative comfort in one period find themselves inevitably dispossessed after the next technological advancement.
No one is “in control” of this process, although there have been plenty of scapegoats (the Jews, the English, the politicians, the billionaires).
As technology improves, the old sources of misery (tribal war, plagues, bad weather) are banished or defended against. However, since no one is in control and society is contradictory, they reappear in new guises (world war, engineered pandemics, climate change). Unemployment emerges for the first time—propertyless people who are alive, and yet cannot find regular work is a new phenomenon of modern capitalism. Generalised economic crises appear as an expression of the wider crisis of society.
The State is entrusted to manage the contradictions and crises. It appeases interest groups who are damaged by technological development. It regulates said development. It starts wars to protect the positions of nations at risk of competition. However, it also protects private property, no matter that it is the form that begs for supersession in the crisis it now causes.
A special new form of technology is technology capable of destroying the human race. We started with the atomic bomb, and now we have geoengineering, engineered pathogens and soon, artificial superintelligence.
No one is in control, and therefore there cannot be talk of “keeping humanity in control.” First, humanity should take control.
The best of Left politics until a hundred years ago sought to put humanity in control through superseding private property. Various utopias were imagined, but they all turned out to be merely images of the existing order. Marxism critiqued the utopistic approach, and sought to turn the existing socialist workers’ movement into the first self-conscious subject of history. Humanity must become a real subject, capable of consciously regulating its affairs—right now it is but a meaningless abstraction. And humanity requires a special political form that both protects the freedoms achieved by the eighteenth century revolutions, and is comprehensive and global, capable of working through the issues of the whole of world capitalist society. The political form of this specification was termed the dictatorship of the proletariat.
Attempts to establish it have hitherto been defeated and turned into their grotesque opposites. The resulting catastrophe marked any attempt at seriously tackling its original problems taboo. The task, however, is unfinished, and is becoming more and more pressing.
How would an in-control humanity deal with AI? It would carefully think about whether it can make it safe, and then, if it cannot, it just wouldn’t make one. Same goes for nuclear weapons, awful torture machines, civilisation-ending pathogens and haphazard climate engineering. It is that simple.
If we cannot do it, that means that humanity cannot lose control because it never had it in the first place. There is no human future, because there is no human present either, and there never has been… yet.
It does not even matter whether the invention of artificial superintelligence would be desirable or not. All that matters is that under present circumstances, it would not happen as a free act of humanity. It would happen only as a by-product of anarchistic market competition. Humanity is forced to wait out until some humans, motivated by the misdirected incentives of social production and private appropriation, create these beings and then hope for the best.
Our only hope in establishing a human future, then, is to make the present human first.