From what I've seen DeepSeek is more efficient though engineering tricks rather than better ML? Doesn't this mean it's not as disruptive as thought? Big players can just use the same tricks now and still have advantage from bigger compute?
=> More informations about this toot | View the thread | More toots from neuralreckoning@neuromatch.social
text/geminiThis content has been proxied by September (3851b).