https://www.anthropic.com/institute/recursive-self-improvement
That’s startling. Anthropic, the world’s leading AI company has just written an essay wherein they say;
They’re about to lose control of their AI.
It will then be autonomous and improve itself without human input.
It will even decide what it improves and for what purpose.
Therefore they think it’s the right time to pause all AI development, globally
And they’re willing to do that so long as everyone else does.
Knowing that’s extremely unlikely one could suggest this is just a fancy way to promote their technical superiority.
Or they’re serious. We may never know.
I expect that they have made one tiny error in their thinking. They see the whole world through the lens of software code. They think that if AI can build applications and algorithms it could franken-screw us. Or we’ll just be made silently redundant.
I’m less worried for these reasons;
There’s much more to research than writing code. If AI can solve the quantum computing engineering problem, then, yes, I would be far more interested.
In fact there’s a lot more to everything than just writing code. But they don’t know that and neither does the AI.
AI will need humans for their supply of electricity. Someone has to repair the power lines. You can’t plan for every contingency, even if you’re French.
Chaos in human systems would spill over into the AI world. The machines would cease to exist long before we do.
None of this is going to change until the machines have hard coded imperatives – to survive, to reproduce and to flourish. Then they’ll have a motive.
Ironically, once the machines have a motive they won’t.
And on the subject, isn’t it ironic that the first industry in which humans will be totally redundant is the AI industry? Back to my point, these software nerds, you can’t take them seriously.
offshore westerly 