@matt_martensen There are pros, It would be verifiable, composable and censorship resistant.
But there are also cons, It’s expensive, slow and hard to scale.
Maybe a hybrid could be a good start.
Is it just about improving LLMs with external symbolic systems or does it make sense for LLMs to learn symbolic representations?
I will start working on my masters thesis soon, I have explored neurosymbolic AI this year with some symbolic methods and ontologies subjects, would appreciate any guidance on which specific problem to tackle
@matt_martensen Now there are conversations about developing US open-source AI, imo It should be a worldwide network, and on-chain might be the best infra.
@TheAhmadOsman Precisely, but imo this is not about a cybersecurity threat, the US is basically protecting national companies by limiting chinese competitors. Similar thing happened with EVs.