fluxty

This is pretty cool. As a runner, I try to stay indoors when the weather is bad, but I also wonder if there are highly local pollution patterns related to traffic. Ideally, you could identify something like that with a tool like this.

It is kind of interesting, given that Princeton's not really at the top in AI (but it would be difficult to be, because they're so tiny--departments like Berkeley have a huge number of AI faculty).

But, at the same time, Google already has connections with Princeton, both through these individual profs and through the NY Tech campus (they shared an office with Google before moving to Roosevelt Island).

It is kind of interesting in the trend of professors being more connected with industry and spending time actually working at tech companies. It's both interesting from the perspective of engagement (profs can work on real problems) and frightening from the prospective of independence--the companies will inevitably sway which problems profs work on, and of course their students won't enjoy their advisors' attention being scattered elsewhere.

Clifford Stoll, I suppose, had a certain amount of skepticism, seeing as he was the person to help track down and prosecute the first recorded case of computer hacking. A large part of the Cuckoo's Egg was about how there simply weren't systems for dealing with hacking--he'd go to the police, and they'd be like, "Cool..." and do nothing.

I used to be into this stuff in college--we were obsessed with the idea of generating very large or else very large numbers of odd numbers, such as weird numbers or untouchable numbers. I was into HPC and writing programs that run on clusters, and this gave me some good practice.

We eventually realized that no one really cared about the numbers we felt confident we could find--now Mersenne primes, though, people really care about those, but they're extraordinarily difficult to find. Always a cool day to see a new one.

Nice. That's some response time.

I wonder what topology this has--it definitely seems reminiscent of older supercomputers like the famous Thinking Machines CM-5, which used a hypercube.

Had to read your answer twice... "High latency???" But I see you consider 1-2us high latency. :P

It is an interesting question what people who don't have access to a supercomputer, but would like to learn and optimize for HPC-style distributed memory programming should use.

I've found AWS to be pretty nice, except there are no RDMA drivers for the elastic NIC and the BW is a bit low. (25Gbit vs. 100Gbit). For MPI bulk synchronous programs, it's probably a pretty close model, though.