rahulmehta95

Post-2002 developments that come to mind that hadn't occurred at the time of writing: anything involving Fourier analysis of the boolean hypercube, like Ryan Williams' circuit lower-bounds & the Unique Games Conjecture, and work on pseudorandomness.

More broadly, I'm always fascinated how interaction of (probabilistic) functions is the underlying concept across such a broad swath of recent developments in complexity theory. Looking forward to seeing where the next paradigm shift comes from

This isn't a serious ML result; it reads more like course notes/toy experiments rather than a serious research effort. The funny thing is, this is still only for tabular data; we don't see any actual classification results with any proper robust metrics (also the PCA setup seems a bit bizarre in this setting).

Ipe is absolutely fantastic! It's my go-to for any figures, and I'll only switch to Inkscape or Illustrator if I need to make a slightly more-involved diagram.

I was thinking that using Visitors would be a solution, which is a form of implementing a polymorphic API.

Interesting how the article makes no note of the Collison brothers' and Stripe...that's a significant Valley success story that doesn't fit this mold

Wouldn't this be a direct violation of the Outer Space Treaty?

This may be an uncomfortable fact but people have surprisingly short memories: the military funded the majority of the early advances in systems, networking, and cryptography (and especially as a large part of the latter subject area, invested heavily in fundamental, theoretical research). Not saying that I disagree with the employees' opinion, just that DoD/Pentagon involvement in artificial intelligence research shouldn't be viewed as a necessarily bad thing. Many other major powers have heavily invested in AI across all fronts (including military applications), and it would be stupid for the US to not have one of its' largest strategic assets to not be part of the process.