Once computers can match human-level intelligence, that’s not all. With deeper knowledge, machine-level mathematical ability, and better algorithms, they will create superintelligence, right?
Well, there’s no question that machines will eventually become smarter than humans. I don’t know how long that will take. It may take years or even centuries.
in that By the way, do you need to batten the hatch?
No, no. It’s going to be like we all have AI assistants and we all have super smart people working with us. They cease to be human. Humans feel threatened by this, but I think they should also be excited about it. What excites me most is working with people who are smarter than me. Because it amplifies your own abilities.
But if computers become superintelligent, why do they need us?
Just because AI systems are smart, there is no reason why they would want to control us. Those who imagine that AI systems have the same motivations as humans are mistaken. they don’t. We will design it so that this does not happen.
What happens if humans don’t incorporate these drives, and superintelligent systems end up harming humans by single-mindedly pursuing their goals? Philosopher Nick Bostrom suggests making paperclips no matter what. Like the example of a system designed to conquer the world to make more paperclips.
It would be very foolish to build a system and not build guardrails. It’s like building a car with a 1,000 horsepower engine and no brakes. Incorporating drives into AI systems is the only way to make them controllable and safe. I call this goal-driven AI. This is kind of a new architecture and there is no demonstration of it at this time.
Is that what you are working on now?
yes. The idea is that machines have a purpose that must be met, and they cannot produce anything that does not meet that purpose. Their purpose may include guardrails to keep out dangerous objects, etc. That’s how you make AI systems secure.
Do you think you will live your life regretting the consequences of the AI you helped?
If I thought so, I would stop what I was doing.
Are you a big jazz fan? Will what AI generates match the elite, euphoric creativity that only humans have been able to produce? Can it produce soulful work?
The answer is complicated. Yes, in the sense that AI systems will eventually produce music, or even visual art, etc., with a technical quality that is equal to, and perhaps better than, what humans can do. However, AI systems do not have the essence of musical improvisation that relies on the transmission of human moods and emotions. At least not yet. That’s why you should listen to jazz live.