> Therefore we should design programming languages that on the surface look very unoptimized, but specify enough information that compilers can do a really good

> Therefore we should design programming languages that on the surface look very unoptimized, but specify enough information that compilers can do a really good job. It sounds like no one knows what such a programming language would look like. This entire post would be painfully familiar to anyone working with probabilistic programming. An intriguing overview of the state of the art was posted here a little while ago: https://moalquraishi.wordpress.com/2015/03/29/the-state-of-probabilistic-programming/ https://moalquraishi.wordpress.com/2015/03/29/the-state-of-p... The task here (I'm adapting an older comment of mine) is that the human should write a relatively simple procedural program that says "if we were to fix a bunch of properties/states of the natural world and provide them as input, here's the output/reading you'd hypothetically get," and then from that, the "compiler" and runtime create a function that, given a large number of outputs/readings, provides a posterior probability distribution over those hidden properties/states of the world. The implementation choices for this "function" can make a difference of days or weeks of runtime on huge computer clusters. The problem is that machine learning researchers use all sorts of insights to make manual optimizations when hand-rolling machine learning code - for instance, if there are real-world reasons why a certain type of search over your space may be well-suited or ill-suited to the problem domain. And not only can the languages not capture these insights yet, but even if they could, there's fragmentation in the community (since it's at such an early stage), and there's no standardized way to map heuristics detecting optimizable patterns to actual optimizations. I'd imagine optimizing compilers looked very similar at one point, though... so only time will tell how far the community progresses!