The groundbreaking and controversial work by Google DeepMind, with David Silver as a coauthor, suggests that future AI generations will need to go beyond training on human sources. Instead, AI agents will generate their experiences of training by learning through interaction with environments, with the accent on learning by doing and not learning by downloading.
If this scenario is legitimate, it does imply something beyond just an improvement of AI: it embodies the very assumption—a little poke at existing AI models (OpenAI and Meta, among others) that are too strong believers in centralized human-generated datasets. That method has its own limitations: legal, ethical, and computational.
But what would happen if AI were to evolve just like life? What if it could create, adapt, and optimize without our content?
Now picture this self-learning, ever-evolving AI embedded in a decentralization-true world...
Web3 is such that users own the data, not tech giants.
Training in decentralized AI occurs on nodes, environments, and agents, such that no one closes the doors behind them.
Rather than scraping data from the internet, we create on-chain experiences, simulated worlds, and peer-to-peer learning networks so AIs collaboratively evolve.
This introduces a path toward autonomous AI systems governed not by Big Tech but by open protocols, blockchain coordination, and tokenized incentives.
The Era of Experience + Web3 = A future where intelligence is not only artificial but is also decentralized.
Ready for it?
#CryptoJournalism #BuildPi2gether #Write2Earn