e're getting another blizzard of articles trying to scare us
into allowing the government to restrict AI in case anyone ever
invents it.
What's remarkable is that the articles, some signed by luminaries in the field like Geoffrey Hinton, (a) use almost identical wording and (b) are mostly content-free, as if they were written by a chatbot.
An example from Science magazine:
This unchecked AI advancement could culminate in a large-scale loss of life and the biosphere, and the marginalization or extinction of humanity.
Notice the writing style: does he mean large-scale loss of the biosphere? If so, what would constitute a small scale loss of the biosphere? And isn't humanity part of the biosphere, making the second clause redundant? Bad logic with perfect syntax is a classic sign of a chatbot.
Another article says failure to defend against AI could have “dire consequences.” What are they all babbling about? Is there a real threat or is this more scare-boasting? Telling us there's an “evolving threat” and we're “on the precipice of a future” is meaningless. We're always on the precipice of a future. If they bother to give a specific, it is that AI will kill us all or put us all out of a job or maybe create a deadly virus that will kill millions of people.
For the millionth time: there is no such a thing as AI. The only thing that's new is that people have redefined ‘AI’ to be what we already have. It's like 5G phones. Few people seem to be aware that they're all working on the midband, not the higher band which was supposed to give us those super-fast downloads. Any increase in speed is mostly an illusion from more towers, more channels, and better modulation schemes.
But no. Go into a phone store and the first thing they do is to tell you to throw your old rickety 4G LTE phone away. Even commentators stopped wringing their hands about falling behind the 5G revolution. The ‘revolution’ came and went like a gentle spring rain.
Yeah, send those ideas to Hollywood. That claim that AI would find millions of new drugs is an example. Oh, it found millions all right. Then it was up to the humans to figure out which ones could actually be synthesized. And it would take billions of mice to test them all. As with 5G, the one thing that would have made it useful, it couldn't do.
Another clue is in this statement:
We urgently need national institutions and international governance to enforce standards that prevent recklessness and misuse.
AI is like global warming or microplastics. It has all the necessary traits: The danger is in the future and it could be something bad, so it's an urgent problem and the government must become even bigger to fix it.
Maybe we should encourage this. Maybe if they're all scared of AI they'll stop pushing their fear of global warming or microplastics or 5G radiation onto us. They're talking about deepfakes, also known as cartoons, which would cut the credibility of our news media in half—from 1 percent to oh, say half a percent. Or maybe, they're saying, AI could write computer scripts or trojans that are so sophisticated they can rewrite themselves to escape detection—and somehow still fit on your computer.
I'm coughing as I write that. So far, all AI does is steal code from the Internet, change it a bit to make sure it doesn't work, and then apologize when you point that out.
The only thing that is actually scary would be putting an AI on a drone to make it chase people down and shoot them. I hate to tell you this, but drones have been doing that for a couple years now. Ask the Russians and Ukrainians.
If the scaremongers keep it up, they'll discredit their cause and we'll have to go back to being scared of tiny pieces of plastic again. And that is indeed a scary thought.
may 25 2024, 5:18 am
