I think the broad shift comes from getting to see your expectations and perceptions against how things play out, repeatedly. For example I've always been quite anti-war and anti-interventionism. I voted for Obama, once, in significant part because of his promise to close Guantanamo on day 1. On day 2 he signed an executive order ordering it closed within a year. I got goose bumps, because I thought we'd finally entered into a new era. Suffice to say it's still open to this very day.
And then he'd go on to bomb a ton of countries, sign indefinite detention without trial of anybody accused of terrorism into law, and more. And this has played out repeatedly. Similarly for the crisis of the day. Something is always an imminent threat unless some urgent action is taken over some timeline that's short enough to be scary, but long enough for most people to forget when the timeline passes and nothing happens. It all destroys political optimism and breeds cynicism.
---
One of the many reasons I'm optimistic about LLMs is that the tech can't really be monopolized. Right now you can run open weight frontier level models on a $1600 'tier' video card, and that cost tier keeps going down rapidly. In the near future we're going to be running frontier models on plain old computers, if not phones. For the effects of this you need to consider how the economy will adjust in response to cognitive working being able to be automated at a price approaching $0. All that means is that the economic value of said work is also going to approach $0. It's going to radically reshape the economy in much the same way the Industrial Revolution did.
People in an agrarian society prior to the Industrial Revolution might have predicted 'super farms'. And those super farms did come to pass. But the value of agriculture (as a percent of the economy) began rapidly trending towards $0 to the point that those farms became a practically inconsequential aspect of the new structure of the economy. I expect this is what we'll also see with LLMs. The things we expect to dominate the world in the future will likely become ever less relevant. The tricky part, and trillion dollar question, is in our new economic equilibrium, what will be valued?
And then he'd go on to bomb a ton of countries, sign indefinite detention without trial of anybody accused of terrorism into law, and more. And this has played out repeatedly. Similarly for the crisis of the day. Something is always an imminent threat unless some urgent action is taken over some timeline that's short enough to be scary, but long enough for most people to forget when the timeline passes and nothing happens. It all destroys political optimism and breeds cynicism.
---
One of the many reasons I'm optimistic about LLMs is that the tech can't really be monopolized. Right now you can run open weight frontier level models on a $1600 'tier' video card, and that cost tier keeps going down rapidly. In the near future we're going to be running frontier models on plain old computers, if not phones. For the effects of this you need to consider how the economy will adjust in response to cognitive working being able to be automated at a price approaching $0. All that means is that the economic value of said work is also going to approach $0. It's going to radically reshape the economy in much the same way the Industrial Revolution did.
People in an agrarian society prior to the Industrial Revolution might have predicted 'super farms'. And those super farms did come to pass. But the value of agriculture (as a percent of the economy) began rapidly trending towards $0 to the point that those farms became a practically inconsequential aspect of the new structure of the economy. I expect this is what we'll also see with LLMs. The things we expect to dominate the world in the future will likely become ever less relevant. The tricky part, and trillion dollar question, is in our new economic equilibrium, what will be valued?