Full threadalpark3·Is it reasonable to expect that some companies out there are training LLMs powered by Hyena instead of attention? Like a GPT-3 level size to bet on ultra-long context length?View on HN