Fede_V

In my experience, it's incredibly difficult to get a strong positive signal from a CV. You can definitely find red flags (my favorite example: people describing themselves as experts deep learning with 1 year of experience and zero publications) - but, after removing the obviously unqualified, the next steps have very low accuracy.

Honestly, it might have just been a sourcer/recruiter fuck up. If you can at all, ask a friend (or even a friend of a friend) if you can get an internal referal and bypass the broken resume filtering step.

Larry Wall and his wife are some of the nicest people I ever had the pleasure of meeting. I posted this story on HN before several years ago (https://news.ycombinator.com/item?id=9890504), but I think it's worth reposting here:

I have a very charming Larry Wall story. He came to give a seminar to the research center where I was working, and gave a fantastic talk on Perl 6. Late in the evening, I ran into him and his wife on the bus when getting home from work - and started talking to them. Apparently, nobody from the department had arranged to take them out for dinner, so we ended up going out for dinner together, where we had a fantastic discussion covering religion, tolerance, and lots of other topics. I am a staunch materialist and atheist, while both the Walls are serious committed Christians, but we had a really pleasant discussion on religion and the nature of evidence.

I ended up giving his wife a copy of Hume's Dialogue on Natural Religion, and they were kind enough to sign a copy of the Camel book for a friend who is a huge Perl fan.

A lot of people know Larry by reputation - but his wife Gloria is just as smart and kind. It was a bit of a strange evening, but it was a rare privilege to meet two such interesting people.

I cannot emphasize enough how unassuming, kind and decent the Walls are. If you are also interested in very intelligent writing about religion, their son, Aaron, has a very neat blog: http://www.wall.org/~aron/blog/.

Interestingly, Isabelle is the director of the INRIA institute in Paris where a lot of coq is developed. I wonder if it's named after her :)

Tech employees are, as a group:

- Highly educated - Young - High % of foreign citizens - Urban - Wealthy

If you take the general population and control for those covariates, you end up with a subpopulation that massively skews left. For example - the median age for people watching Fox news is nearly 70.

You almost have to admire the total ruthlessness of republican operatives. By constantly playing on the offensive, they've put Facebook and Twitter on a constantly defensive posture worried about being called out as having 'liberal bias'.

The fact that Jack had to personally intervene to stop his employees from banning Alex Jones (after numerous egregious violations) speaks to the power of this approach.

I lived in Paris for a year, and I'm now working for a big tech company in the US. I can't speak from the VC side of the table, but France has incredibly technically strong graduates (people with technical degrees from the top universities there are all very good) and young people have a very global outlook. Further, there are more and more programs that provide start up funding/locations/tax breaks.

What is missing is an ecosystem to sustain and nurture these companies. The capital in Europe is incredibly conservative and very old fashioned.

I've said it in the past, but I'm going to repeat myself: if anyone is willing to defend Pai's actions, would you like to do a friendly bet (proceeds to to a GiveWell charity) that as soon as he is legally allowed too he will be given a sinecure at Verizon/Comcast or an RNC friendly lobby shop?

An interesting exercise: assume the first engineering hire you want has to be at least the senior engineer bar at a top tech firm.

In order for the engineer's median compensation to match the median compensation he would make working at FANG, how much stock would you have to give them making reasonable assumption about valuations (ie, assume your start up will have a trajectory in the top 30% percentile, but not in the top 0.1%).

I'm sure there is a practical reason why companies do this, but it drives me crazy and makes me want to ignore what could potentially be a really cool idea. Please make the pricing transparent and viewable without a registration.

This is a fantastic post :) I'd love to see it fleshed out in a blog if you feel so inclined!

Let me flesh out a few very simple proposals: a complaint that a lot of city residents have is that tourists that arrive in a city don't spend a lot of money there because they pack their own lunches/avoid museums, etc.

How about charging a very high tax (in the form of a voucher) for hotel/AirBnB stays, but make that voucher spendable in local restaurants/museums? This would have the bonus that you'd also kill the local black market economy, as tourists would insist on being able to pay for goods using their voucher.

I can't wrap my mind around people that think that 'too many tourists' or 'too many jobs' are a problem. Those are 'problems' that most cities in the world would kill for: if we consider them problems it's because we are not nearly creative enough to take advantage of them.

As a simple example: if we have too many tourists, give a progressive daily tax that's charged to every visitor that automatically includes museum visits/etc. Use that revenue to build more housing for local residents (and if some dipshit points out that more housing ruins a neighboorhood's character, ask yourself if Paris or suburbia has more character) or to subsidize local transport infrastructure.

I think this is an interesting and unintended consequence of AirBnB not investing nearly on filtering their users.

AirBnB obviously was interested in getting their product used by as many people as people (growth at all costs) - but this had the effect of getting a lot of young, cost conscious people to take holidays in area that were traditionally residential. While most AirBnB renters are polite and respectful, enough of them were loud, disruptive and ignoring the local rules (for example, tourists in Venice who use AirBnB but ignore the very strict garbage disposal laws) that local residents started dreading having AirBnBs around.

AirBnB is already very strict with landlords: it needs to start being a lot stricter with its users too. They rapidly need to educate their users about how to be good citizens, and that the motto of 'the customer is always right' is completely the wrong way to think about a stay in a residential area in a foreign city. Users should think of themselves as a respectful guest, not an entitled customer.

I'm not sure how you could call that generous give the col of SF. Do we just expect government employees to not be able to live in the city they work?

It's basically taken almost verbatim from a dialogue between two characters in The Brothers Karamazov. She later said she didn't remember the dialogue at the time but it was stuck in the back of her head and was definitely an influence.

I'm not very familiar with Clojure, but are the numerics done with unboxed numbers? In the end, what every language that wants to do scientific programming right is expose some sort of c-array like data structure that you can pass onto c/fortran code.

I lived in France for a year before moving to the US. Even though French waiters are notorious for having terrible service, I always received excellent service by being very polite to the waiters.

The idea that you are entitled to excellent service because you are paying is pernicious. If you want excellent service, treat people excellently.

As I've previously said: for people that believe Pai acted according to good faith, are you willing to bet that, as soon as he is legally allowed to do so, Pai won't be given a VP-level sinecure at Verizon or Comcast?

I've done a lot of interviews for work, and I make it very clear to the candidate that the questions they ask me won't be used to evaluate them and that they are purely for their benefit.

Candidates have just had a very stressful interview, I don't want them to have to worry about choosing to impress me versus getting the information they really seek. Obviously this doesn't apply if they ask questions that reveal massive red flags (like something utterly racist or sexist) but so far that's never happened in over 70+ interviews.

To emphasize the fact that questions are for the benefit of the interviewee, if I'm interviewing someone that's foreign and I happen to speak their language, after the interview is over I close the computer and switch to their language to make it clear that we are now having a conversation and they can relax and be informal.

Pai's speech was funny, but his behaviour as chairman is completely indefensible, especially with all the evidence that the comment submission process was fraught with irregularities.

To put it more concretely: if you think he has no ulterior motive in acting the way he is acting, what betting odds would you give against him being given a very lucratively paid sinecure as vp in some sort of telecom as soon as he is legally allowed?

For experimenting, I think the imperative paradigm (define the graph by running it) makes for much faster experimentation and requires a lot less boilerplate code.

All big DL frameworks (TF, PyTorch, MXNet, Chainer) are now either imperative or have an imperative API in the works.

I have a few contradictory thoughts on this.

On one hand, I really feel Chrome is more on my side than the average webmaster. I don't want bloated websites, I don't want autoplay videos, I don't want copy/paste blocked, I don't waste text copying hijacked so that a phrase has a referral link to the article, etc. If Chrome breaks a website functionality to over-write these things, good.

The downside is that Chrome is a very big player and by working outside of the system then the rest of the web loses on the spillover benefits.

This is hard, committees are slow, bureaucratic and massively resistant to change - and saying fuck it, I'll fix it properly for myself is a lot easier, and I have little patience for it myself. However, in the long ran this is probably better.

But that's exactly my point. Feynman was only able to give such a clear explanation because he really understood QM inside out.

If you want to describe yourself as someone who can read and appreciate Feynman's books - that's completely fine with me. However, I hope we can agree that reading Feynman's books (the ones for laymen at least) is not enough to call yourself a polymath.

Thanks for the interesting post - however, I respectfully disagree with the example you've picked out.

How can you do intelligent philosophy about probability if you don't have a good detailed understanding of what probability is?

To make a concrete example: the best philosophers of logic (Kripke, Ruth Barcan Marcus, Quine, etc) were also superb logicians with an excellent understand of logic itself. Regardless of how you feel about his atheism, Richard Dawkins is in my mind the best philosopher of biology currently alive, and his first two books (Selfish Gene and Extended Phenotypes) are masterworks. Certain concepts in physics are much clearer when formulated in math - and you can only really explain them clearly once you are firmly grounded in the math. Speaking for myself, I've never met someone who I would describe as having a fantastic understand of QM yet skipped the formal mathematical training to deeply understand the equations.

While we do need polymaths, the problem with a lot of self described polymaths is that they skip the hard part of different fields and only focus on a 'Gladwellesque' understanding.

I think a better way to be a polymath is to be a real expert on a particular subject, and then augment that with ancilliary knowledge from other fields. If you claim to understand physics but don't know maths, or claim to be an expert on classical literature but don't know how to read greek/latin, then you aren't a polymath but a bullshit artist.

The polymaths I respect are people like Peter Medawar, JD Bernal, etc..

My particular pet peeve lately is people (especially prevalent in the rationalist sphere) that talk about how Bayesian they are, but don't know what a conjugate prior is, or have no understanding of MCMC sampling. There is absolutely nothing wrong with not knowing either of those two things - but if you don't know either, then describing yourself as a Bayesian is just putting on airs.

It's a scam, but honestly, brands trying to advertise their products without seeming like they are advertising are just as dishonest.

The description of differential privacy in the article is completely wrong. I am really shocked it was described so poorly.

That Twitter, as it's currently designed, is an excellent platform for sharing cool links (science/ML twitter is fantastic), but it is downright deleterious for any subject where nuance is called for.

I think Twitter had a net negative effect on political discussion not because of free speech, but because it rewards bombastic language, point scoring, cheerleading and makes substantial arguments impossible.

That's not what the quote is saying.

First of all, the architecture of a service has important influence on the kind of speech that it will foster: for example, the 140 character limit insures that it's almost impossible to get a nuanced argument across. Second, the broadcasting model of twitter often leads to horrible dogpiles.

I am a huge free speech advocate, and I don't like censorship but several key decisions about the architecture of twitter were definitely suboptimal if your goal was building a better town square for the exchange of ideas.

Biology is at the intersection of those fields. There is an oversupply problem, and it is definitely the worst offender for metrics-based research micromanagenent.