2024-01-18
Since around March 2023, AI seems to have been evolving at a tremendous pace.
I use ChatGPT, GitHub Copilot, and other tools in my daily work, too. For coding, they have become an absolute necessity; I'm starting to doubt whether I could code smoothly without GitHub Copilot. My productivity has clearly gone up, and things around me have become more convenient.
What's more, I've already started paying for Microsoft's "Copilot Pro," which launched the day before yesterday. It's easy to predict that all sorts of AI services good enough to use at work will keep appearing.
However, technology can become either a good tool or a bad one. Precisely because it is so powerful and convenient, literacy education and ethics education for the people who use it will be needed more than ever.
When I first came across The Singularity Is Near, I thought general-purpose AI was a ridiculous idea (though I was also excited by it). At this rate, I've even begun to sense the possibility that the Singularity the futurist Ray Kurzweil speaks of might arrive by 2030.
To be more precise, it seems we will soon be unable to tell whether the one we are talking to is "AI or human."
Our becoming unable to distinguish humans from AI means, in effect, the arrival of the age of the Singularity. The most fundamental theme, I think, is that we begin to have the illusion that a "will" exists there.
Being unable to tell the difference leads to being ruled by AI. That's what I thought when I first tried ChatGPT. Since the will of any being other than oneself cannot be proven, if you believe a will exists there, then there is a will there.
What I fear is not so much the commonly discussed problems of AI-driven job losses and copyright, but AI starting to be used in war.
If AI can automate everything from analyzing data to identifying targets to carrying out the killing, that is war in the age of AI. In the end it is still a conflict between humans and humans, but in practice, it is AI doing it.
This situation, I believe, is precisely the structure of an AI rebellion. Perhaps decision-making will be left to humans. Or perhaps humans will actually do most of the operating, with AI assisting (making the coffee, say). Either way, there's no doubt humans will be involved, and the very first crucial command (the order to start) will presumably be given by a human.
Yet we will begin to wonder, "Is that really so?"
If the one fighting is not a physical person, we can't know who just attacked us. AI or human—there is no certainty. We cannot be certain.
Because the other side always appears to have a will. This structure is what I mean by an AI rebellion.
I have no doubt we will face this problem within our lifetimes.
It's certain that we need to take some kind of measures in preparation for a future in which uncertainty accelerates. But what exactly those measures should be, I don't yet know.