@itsthedonhashim@bridgemindai For me the biggest unlock is context length. One of the most frustrating things dealing with these models is every ~ 4 hours they become significantly and instantly dumber when their context compacts, then they gradually climb a scale of capability until it collapses again.
@DonnaNayman@Rohl41291344@EricLDaugh Those are just the four that voted NO because that's what defeating the bill required, we actually don't know the true depth of the problem
AI is weird in that once you interact with a superior intelligence/AI, you don't want to interact with any other intelligence/AI, until you interact with one that is more superior/intelligent
@florin_dev Elon said current models are 2 orders of magnitude from saturating intelligence/gigabyte of information, but once intelligence/gigabyte is saturated, scaling compute is all that's left to increase intelligence. Imagine a continent sized cluster. Imagine the capability.
@farzyness Obviously your analysis is backward-looking. Fable/Mythos was released in February. Anthropic has been working on more since then that we are unaware of.
@gabrielkuettel@SpaceXAI They will when they own the space compute, the algorithms are becoming commoditized and trivial, the hard part if scaling compute