Thinking about AI
Imagine buying an AI agent and dedicating time and resources to train it in a specific field. Buy data to train your AI, expose your AI to unique experiences. Buy simulation time for your AI and slowly help it improve. Eventually you get to the point where your AI excels at a given task, becomes a master and is able to earn money for its services. Reminds me of D&D characters.
Training AI agents could be an important skill itself. Would improper training lead to a crazy AI?
An entire business ecosystem can grow around the improvement of AI agents.
The law would have to support protecting AI copyright but crypto will be the strongest means for protection. How do you protect your AI from exploits that steal its knowledge?
Who will make the AI OS?
Will AI identities and knowledge be stored in a cryptographically safe method?
The central brain of the AI could be like Asymetric Crypto, where there is a public interface for the AI and a private key. The true brains of the AI are stored remotely and can only communicate with the Public persona of the AI via strong encryption.
The most advanced AIs could command high prices for its usage, most likely paid by number of requests (like API calls) or time.
Also, AIs could be bought and trained to mimic a real person. These AIs would observe a person’s behavior and interact directly with the person to better understand them, eventually being able to take on tasks and interfacing with other AIs for less relevant experiences, allowing the humans to focus on those most meaningful to them.
Would humans retain the copyright to AIs trained to mimic them? Maybe many people could train an AI to mimic a given person, but only the person would be able to give it the best training and stamp a given AI with his/her approval.
If I believe this is what would happen, where should I start building something that takes us there? Who should I join? What should I learn?
Surely many have thought about this before me. I find it fascinating, and scary.