[falcor84]

Really, there's "no reason"? For me, watching ML gradually get better at every single benchmark thrown against it is quite a good reason. At this stage, the burden of proof is clearly on those who say it'll stop improving.