pmorais

Why? None of what he said suggests to me than an incompetent women would be hired over a competent man. The outrage over incentivizing minority hires is ridiculous to me. You’re more likely not to get hired because of random noise in the interview process than because you happened to apply at the same time as an equally qualified minority. If companies like google were actually actively discriminating against competent asian/white male developers in favor of minorities their engineer demographics wouldn’t be 80%+ asian/white male. There’s also legitimate business interests for a company to have a diverse body of engineers and managers.

It’s actually much worse than that because of reflected inertia. The inertia of the first gear as seen from the last gear is proportional to the square of the gear ratio

The US government DOES have such a fund though. It’s called DARPA/ONR/whatever other military research wing. They all invest heavily in startups.

That’s not really true. The other 3 “fundamental” forces are irreducible yet still entirely explained by the standard model.

I don’t think this kind of insight even exists into deep learning stuff (yet). One of the main reasons people even care about SVMs is that they have nice analytical properties

You’re not the intended audience. This is meant for small research labs that are just starting up and want to enter the RL/Robotics research space and can’t afford a $400,000 PR2 or $150,000 shadow hand.

You realize you have to pay an employee to do that right? Odds are it’s still more cost effective to just buy ads.

I mean, once you hit a certain salary more money doesn’t even appreciably increase your quality of life. And I think most good software engineers are already close to that asymptote

That’s revisionist nonsense.

Basically of modern convex optimization is heavily dependent on SNOPT. Deep learning research is heavily dependent on cuDNN

I feel like this hubris is why almost every google product I use is riddled with horrible bugs. The cursor move functionality on Gmail for iOS has been broken for TWO YEARS. It’s also why the tensorflow API is a horribly over-engineered abomination when Pytorch is almost at feature and performance parity with something exponentially simpler and more consistent/elegant. That “biggest baddest solution” culture makes sense for Bigtable; it’s god-awful for anything remotely customer facing.

Still missing Hangouts, Chat and GChat

I have never heard of a tech company that doesn’t require that you explicitly give up ownership rights to software written on the job, so good luck with that