The big internal-drama spectacle at OpenAI has once again brought into sharp relief the debate about artificial intelligence, “effective accelerationism vs. super ‘alignment’.” I also posted two comments on Weibo, which I record here:
I support neither side. Accelerationism believes that whenever new technology disrupts society, the result is always good. But the problem is that in history, good outcomes are also the result of positive social change; if society merely passively accepts the push of new technology, there may not necessarily be good results. The problem with alignmentism is that it is fanciful: human values have never been aligned, so how could machines be? I support checks and balances on technology, but not by transforming technology; rather, by actively transforming society.
Reply 1: I am not trying to block technology, but to accelerate social change, rather than waiting around for technology to change society automatically. Also, if you really believe that technology cannot be blocked, then there is no need to care about those who try to block technology, because their efforts are futile. If, on the contrary, you hope to block the efforts of those who are trying to block technological innovation, then that means you still believe blocking can be effective.
Reply 2: The problem with alignment is this: if there are 100 AI projects in the world, and 99 of them are aligned while the remaining one is not, then it may still go out of control. But as for the direction of social transformation: if there are 100 autonomous communities in the world, and 99 fail to make a revolutionary breakthrough, if one of them has its proactive revolutionary direction correctly set, then it will ultimately also be able to adapt to the AI era.
Let me add a little more explanation here:
The “double revolution” theory that I have mentioned many times before expresses a similar idea: the reason we see technological change ultimately bringing good results is that every technological transformation is accompanied by a social transformation that ultimately succeeds. But these social transformations are the result of human effort, the result of humans and technology “rushing toward each other” from both directions, rather than something people simply sit around waiting to fall from the sky.
I do not support deliberately obstructing or slowing down technological development, but I also do not oppose those who make such efforts. On the one hand, for the reasons I mentioned above: if you really believe technology cannot be resisted, then there is no need at all to resist the resisters. On the other hand, it is also a matter of maintaining a humble “anti-God’s-eye-view” stance. Because we have no right to prescribe values for others or for future generations. Our resistance to the values embedded in technology may only be a small current in the long river of history; many others who embrace technology do not support us, to say nothing of future generations, who may not even be able to understand our thinking. So I have the right to resist it (I myself do not use new technology), but not the right to obstruct it (to prevent others from using new technology as well).
I do not support blithely and completely letting technological development run free; rather, I advocate keeping technology in check, or in other words, balancing and steering it. We know that when riding a horse or driving a carriage, carefully maintaining balance, deliberately controlling direction, and even occasionally braking and slowing down is not meant to slow down or obstruct forward movement in the overall sense; on the contrary, checks, balances, and control are there so that one can accelerate more comfortably and for longer. Letting a runaway horse gallop freely may produce a noticeable sense of acceleration in the short term, but in the long run it may well end in wreckage and death, or it may simply spin in place; it may not truly go farther.
But at this point my view actually seems to coincide with “effective accelerationism”: so-called effective acceleration also emphasizes mastering technology and reforming society to adapt to technology. The difference is that they (taking the more mainstream formulation as the standard) usually view technology as a neutral tool, and therefore believe that whether technology is good or bad depends entirely on how humans use it; whereas I think technology is some kind of “half-tool—half-living-being,” similar to a mule or horse, or a “company serf”: it may be used as a tool, but it also has a will of its own. On the other hand, what they call reforming society to adapt to technology ultimately still focuses on how to make society more favorable to the proliferation and iteration of technology, rather than on society as a force that can check and balance technology.
In addition, neither accelerationism nor alignmentism seems to take seriously the complexity of human society itself. Accelerationism believes that technology can be controlled by human beings, but the problem is: whose will exactly is supposed to do the controlling? Alignmentism hopes to align machines with human values, but the problem is: whose values exactly are to serve as the standard? In fact, within human society, different individuals and different communities have never formed a unified will or a unified set of ideas, so in the final analysis accelerationism and alignmentism are still just two camps each insisting that it should represent humanity—and disputes of this sort have occurred countless times in human history, without ever producing a once-and-for-all solution.
My position first of all requires abandoning a top-down perspective. Any attitude or strategy I discuss toward technological development is bottom-up: starting from myself, starting from small autonomous associations. I do not seek to bridge the gap between accelerationism and alignmentism, nor do I try to sabotage either side’s efforts; rather, I advocate another dimension of effort, one that begins from the grassroots of society. The Web3 and DAO that I favor point in this direction.
If misunderstandings could be dispelled, I would be willing to call my position “socialism,” where “society” refers to bottom-up, autonomously associated, culturally plural communities in the plural sense. In order to adapt to and steer the development of technology, we need to take the initiative to transform society from the bottom up.
Translated from the Chinese original with AI assistance. The original text is authoritative.
Leave a Reply