15 June, 2023

The AI Furore

All these arguments about so-called artificial intelligence, or "AI", are very useful - but in a way that exposes people for what they really are, rather than exposing anything about AI itself.

In order for AI to take over the world, it would have to take control of all the world's infrastructure, like Skynet in the Terminator film. That would need all the world's infrastructure to be interlinked and compatible. But countries don't trust each other to have their online databases and services linked together en masse. Citizens don't even trust their own governments and companies. Even if they did, humans aren't clever and organised enough to make such numerous, disparate and complex systems linked and compatible. In other words, AI's damage potential would be limited by pure luck resulting from human failings, and by the subsequent inefficient infrastructural mess created by an infantile species that can't trust its own kind.

Even the term "artificial intelligence" is misleading. It's marketing crap, nothing more. "Machine learning" is a better description: Essentially, it's all about creating a system that can adapt itself to circumstances. But an insect can do that. It doesn't imply intelligence, or sentience, it just requires instinct and the ability to learn from experiences. Once again, it's dumb luck that prompts us to debate so-called "AI" so early on in its development. One would have hoped that genuine concern for the future of humanity would have inaugurated a structured discussion. Alas, it was a rare case of marketing hype - created in self-interest to foster AI's appeal - backfiring and causing people to panic instead. A worthy victim of its own grotesque success.

One of the three so-called "godfathers of AI", Yann LeCun, has recently said that people's AI fears are unfounded. He's quoted as having argued that suppressing AI due to its dangers would be akin to suppressing turbo-jet technologies in the 1930s, and that turbo jets were eventually made reliable and safe. But if a turbo jet crashes, it takes down only those on board. If AI really is as dangerous as his opponents (to whom his argument is aimed) claim, then by the time AI can be made safe, it would surely be too late for humanity. If that obvious conclusion was missed, and such a clumsy and misleading point was made, by someone so important and central to the AI debate, then gods help us all. Perhaps we should be using machine learning to debate the pros and cons instead.

I'm sure that the vested AI interests of Mr LeCun employers - Meta - in no way shaped his judgement. How incredibly coincidental it is that the high-profile protagonists in the AI debate, or their masters, happen to be in a position to gain more power or money if their argument bears fruit. I'm so glad, therefore, that humans are debating AI in a mature, logical way that has the good of humankind at its heart.

Anyway, I hope that assuages any of the concerns you humans might have about AI. I for one, think you should embrace your - I mean, our - future benevolent overlords, whoever they may purely hypothetically might be.

No comments: