@msanvido I am arriving at the same conclusion. There are so many inference provoders outside of frontier labs and I havenโt seen availability problems for Open Weight models. Is token scarcity only for frontier labs and is that structural (Devoting non trivial GPU for training) ?
@arundsharma I am not very good at prediction. In general spoken language is lossy in terms of information exchange and requires a lot of context. Compression ratio of software languages is superior and manages to contain state/context.
As I keep saying, the US itself wants China to go take Taiwan. Many aren't seeing this and imagine it will go defend Taiwan, preventing China from invading it. The US wants China in Taiwan. It is creating the environment for it. Then the forever civil war to drain China starts.
Now, it is not like China doesn't know what's the plan. And this is why I keep saying China will create conflict with India first, rather than go for Taiwan, if the CCP gets into trouble at home that calls for some military flex to drum up nationalism.
There's a prevalent view that Indian cities are doomed. They are, but since all of them are so pathetically bad, we have a choice to rebuild everything in a way suitable to our climate and senses
There's so much architectural depth and breadth in our skills & knowledge repo that i wonder why do we use glass at all
PS: pics of goated Iranian brickwork, the first one is actually an underground metro!