Opening a dedicated thread to address a specific discussion from Near Protocol vs Internet Computer on Ai app creator.
The pertinent comments:
Should we be optimizing ICP for a hybrid approach involving verifiable agents and trusted execution environments (TEEs)?
I can’t speak to any of the relevant technicals involved. Perspectives from the better informed/Dfinity would be appreciated in shedding light on this.
Perhaps Dfinity could elaborate on the architecture of Caffeine AI and how it can be leveraged by dapps working with AI agents.
If a software run on even a trusted environment and we can even verify the compute has not been corrupted, we have no way of preventing the corruption to happen, which means it is then not unstoppable and then has no value regarding critical computation
I am still curious about what other people are thinking about this BUT
If the only reason for this questionning comes from the fact that retail may not understand the value of on chain compute, then it is really bad thinking process
If retail do not understand that is a shame but reality always win in the long run
You are sacrifying long terme reward for short term one ( child thinking process)
If we want to find use cases where verifiable compute is enough, then i am also interested but imo it will have certainly less value for the world
Latest from NEAR