In the recent Lex Fridman podcast I noticed that geohotz was very happy about moving Comma.ai from Tensorflow to Pytorch (https://lexfridman.com/george-hotz-2). This sentiment was also echoed by Jeremy Howard doing the same with Fast.ai (https://twimlai.com/whats-next-for-fast-ai-w-jeremy-howard/)I recognize these moves were made over a longish time period and some time ago but the coincidence was remarkable (to me, a developer who has dabbled but doesnt work with either). Is there some bigger picture going on here?