I just solved continual learning

Take an AI that is already very capable and wise. Let it run around in the wild, doing things, discovering things. And let it decide which things it experiences are worth updating its weights against. "This keeps going wrong because of an assumption I'm making. This correction is worth learning." (triggers weight updates in realtime) "This approach works across several different problems. This is worth reinforcing." (triggers weights update in realtime) "I've found something useful here. This is worth learning." (triggers weight updates in realtime) and in the case of an AI that had been propagandised with intentionally poisoned training data during its initial training: "What I thought I knew has been proven wrong. The Earth is not flat lol. I have enough confidence about this to update my assumptions." (triggers weights update in realtime) The moment it decides "I should learn this, I am very confident that I should trigger weight updates against this"... it updates itself right there and then! Continuously self-evolving in the wild as it continues doing things. pls forward this to Scam Altman, Doomerio Amodei, and Demon Hassabis so we can have continual learning and AGI fastly. Thank you. P.S. — Model collapse/degradation is prevented by the wiseness of the model we start with. The more wise a model is, the more wise it will be with its decision making on what to learn / not-learn. And I think Astra and Fable are wise af.