This comes off as pretty biased. "Run its course?" - really? Remind me, did DeepMind use Probabilistic Graphical Models to beat Lee Sedol at Go? Does any of OpenAI's research use them?