In fact that is exactly the case. Without going into details, I've done a whole lot of AI work professionally and I'm now working harder than ever - without much support, except a couple of my ex coworkers are working with me and I get some licensing fees on my patents - trying to figure out how to build a system with consciousness and self-awareness. Those are traits which both create new kinds of existential risk, and which can help mitigate both that risk and others.
If such systems are created and controlled by people who want power over others, or to take money from others, then we are doomed. If they are created by such people and then those people lose control, it's a gamble with very bad odds. The people working on FAI are living a fantasy life if they think their proposals to deceive or hobble these systems will work; they've made themselves irrelevant. So three of us are spending close to 60 hours a week working to try to create a third option - also a gamble but hopefully one with better odds. And that is a system with consciousness, self awareness, and empathy.
Governments or large investors would not support us; our project, if successful, will not work more to their benefit than anyone else's. And as I said, it is a gamble. The FAI people are selling a fantasy that they can find a way to do this that's completely safe. They can't. In fact any of their proposals I've seen, put into effect, would be either useless (preventing any powerful AI at all much less a friendly one) or actively harmful (creating an AI that is more likely to be hostile because it's learned its behavior from people who treat it as hostile). But to the extent they sell people on the fantasy of perfect safety, those people might decide to lock us up.
And we could be wrong. That's the hell of it; we just don't know. We can point at a bunch of things and explain why they won't work, but just because we're pursuing something different doesn't mean that it will. We're doing this not because we're sure we're right, but because time is running out.
If such systems are created and controlled by people who want power over others, or to take money from others, then we are doomed. If they are created by such people and then those people lose control, it's a gamble with very bad odds. The people working on FAI are living a fantasy life if they think their proposals to deceive or hobble these systems will work; they've made themselves irrelevant. So three of us are spending close to 60 hours a week working to try to create a third option - also a gamble but hopefully one with better odds. And that is a system with consciousness, self awareness, and empathy.
Governments or large investors would not support us; our project, if successful, will not work more to their benefit than anyone else's. And as I said, it is a gamble. The FAI people are selling a fantasy that they can find a way to do this that's completely safe. They can't. In fact any of their proposals I've seen, put into effect, would be either useless (preventing any powerful AI at all much less a friendly one) or actively harmful (creating an AI that is more likely to be hostile because it's learned its behavior from people who treat it as hostile). But to the extent they sell people on the fantasy of perfect safety, those people might decide to lock us up.
And we could be wrong. That's the hell of it; we just don't know. We can point at a bunch of things and explain why they won't work, but just because we're pursuing something different doesn't mean that it will. We're doing this not because we're sure we're right, but because time is running out.