How long before these models actually produce funny outcomes on a regular basis? Many people think we're going to see a quick explosion of improvement, but I t

How long before these models actually produce funny outcomes on a regular basis? Many people think we're going to see a quick explosion of improvement, but I think we're still many years away from these models actually understanding language. I could be wrong, but wondering what you all think