Here's what's super exciting about DeepSeek: if you can vastly improve results with 10% of the compute resources, this will continue to scale when you have 100% of the compute resources.
Even in the most naive implementation, you can use these optimizations to get results that are 10x fresher.
These are exciting times.
=> More informations about this toot | View the thread | More toots from Lee_Holmes@infosec.exchange
text/geminiThis content has been proxied by September (3851b).