jmcohen

Is there an email address where I can send a resume / cover letter if I'm interested in ML at Spotify? Thanks!

Yes, every engineer there has a dev server running a copy of Facebook.

plug for a friend's similar whimsical project: http://www.dayhue.com/

"Likelihood" can refer to both p(param|data) and p(data|param). See [1] for an example of the first, and [2] for the second.

[1] http://en.wikipedia.org/wiki/Likelihood_function

[2] http://en.wikipedia.org/wiki/Bayes%27_theorem#Events

Both first-come-first-serve and random selection are better ways to admit people to a hackathon than "let's play college admissions committee."

HackMIT came up with a nice hybrid solution this year [1]. Most spots were filled by lottery, but a few were reserved for the first people to solve a puzzle. This left a place for merit --- but, like, actual merit, which is not well-assessed by a 19-year-old college student scanning resumes for "placements at major hackathons."

[1] https://medium.com/hackmit-2014/joining-the-fancycat-club-ha...

Yes, I think that the idea that a hackathon can be "won" is silly too.

Penn's hackathon is by application only. I think this is pretty silly. Here's a quote from a Medium post by one of the event organizers announcing this year's low acceptance rate of 30% [1]:

  The people that got in had almost all either won major hackathons before or 
  worked on multiple projects that each blew our socks off in a spectacular fashion.

Organizer #1: This team looks promising. They placed second at the State U hackathon last spring!

#2: That's not impressive. Everyone knows that StateUHacks is a third-tier hackathon. A monkey with a rotary phone could win the Twilio prize there. Who do you think we are, Cornell? This is UPenn; we accept elite hackers only. Only the cream of the crop from HackMIT is good enough to don Dropbox t-shirts with us.

#1: Ok, how about this girl? The side project she submitted under the "Supplemental Materials" section of our application looks pretty good.

#2: As IF! I'd give that side project a 4/10. No infinite scrolling? What a waste of my time. And look at that .ly TLD. It's like, "HELLO, 2012 is calling and asking for its domain name back," am I right?

[1] https://medium.com/pennapps-x/pennapps-x-application-stats-6...

Right, the choice of 40 as the number of factors was totally arbitrary and unscientific. We tried factorizations with 30, 40, and 50 factors, and chose the one that produced the most interpretable latent factors.

All evidence points to the contrary. 34% of Facebook's employees are Asian, but only 19% of managers are.

NYT article: http://bits.blogs.nytimes.com/2014/06/25/facebook-mirrors-te...

Facebook's "Equal Employment Opportunity" report: https://fbnewsroomus.files.wordpress.com/2014/06/facebook-ee...

One reason why bankers might become more dishonest after being primed to think about their jobs is that being a good banker, like being a good used-car salesman, often requires dishonesty.

I’m thinking of a scene from the (non-fiction) book Liar’s Poker where an investment bank finds itself stuck holding a bunch of bonds whose value is rapidly declining. Does the firm swallow the loss? Of course not! Management tells the sales team that priority #1 is unloading the low-quality bonds onto the bank’s gullible customers at an inflated price.

Another example that might hit closer-to-home here is that of equities analysts on Wall Street during the .com bubble who became unreasonably bullish on certain tech companies in order to gin up IPO business for their firms’ profitable mergers/acquisitions divisions.

In banking, the interests of the firm and the customer are often misaligned, and, well, it’s not the customer who's going to pay the banker’s bonus.