In a new video essay published yesterday for the New York Times, Ezra Klein points out that many of the people currently building AI—the people who are now telling us that they believe their own creation might kill all humanity within the next decade—also believed that AI posed an existential threat to humankind even before they started building it.
Indeed, they started building it precisely because they thought that someone would eventually build it, and they thought they needed to be the ones to build it in order to ensure it was safe and would serve human interests, because they thought that otherwise it certainly would lead to human extinction.
Now, they are the ones telling us that their own creation might destroy us all, and that they have already in some critical ways lost control of their own Promethean invention.
As Klein points out, there are almost Greek-drama levels of tragic irony in the situation. He quotes a line form a novel to suggest that the current generation of AI leaders tried to fight against their own doomsday "prophecy" of dangerous "rogue" AI, and in doing so they only managed to bring about its fulfillment.
When they started out in their careers, AI operating at any sophisticated level did not exist. They were convinced AI would go "rogue" and endanger humanity long before it had anything close to the power to do so. Now, they have given it the very powers that they feared.
I have said before that many of the AI tech leaders belong to what amounts to a "doomsday cult" that was convinced of the imminent destruction of humanity at the hands of a godlike AI "singularity" before large language models even existed.
Like many a cultist before them, it seems they actually yearned for the apocalypse they were forever prophesying. As a character in William Gaddis's Carpenter's Gothic says of the millenarian fundamentalist right of the 1980s, with their survivalist camps and "prepper" kits: these guys "are hell bent on a self fulfilling prophecy."
I therefore almost agree with Klein's point—except that I would edit it just a bit. Instead of fulfilling their own prophecy in the very act of fighting against it, I think these guys were never all that serious about trying to fight against or prevent it.
I suspect (though, of course, cannot prove) they always felt the messianic allure of their own apocalyptic visions; and in acting now to create the very danger they warned against, they are responding to a subconscious urge.
These are guys who were fascinated by a science fiction scenario involving rogue AI before such technology was anywhere close to existing. They convinced themselves that they needed to build such a machine in order to preempt other people from doing so. In the very act, they made their own dystopian science fiction warnings a reality.
I'd say that this was a rather bizarre situation if it had not happened before in the history of dangerous and potentially world-ending technologies. It is said that some of the Manhattan Project scientists got the idea for building a nuclear bomb from reading H.G. Wells's prophetic novel on the subject, The World Set Free.
In both cases, then, we find a tragic illustration of Oscar Wilde's famous observation: "Life imitates art far more than Art imitates life."
To quote Wilde's great essay ("The Decay of Lying") further on this subject: "Scientifically speaking, the basis of life [...] is simply the desire for expression, and Art is always presenting various forms through which this expression can be attained. Life seizes on them and uses them, even if they be to her own hurt."
The people building our current AI models—which they already admit they effectively cannot control on their own—started building these things because they were obsessed with what was at the time only a hypothetical—even fanciful—literary possibility. "What happens if we made a superintelligent machine and then it escaped our control?"
"Life" has apparently made full use of this artistic premise, in Wilde's sense—"even if it be to her own hurt."
No comments:
Post a Comment