Fede_V

It's such a huge topic - the modern starting point are papers by Galbraith and Tobin. Since then, there's been so much research/discussion it's hard to point to one thing in particular.

If you want good starting points for someone who isn't an economist, the Boston review had a good primer, followed by a public debate:

Article: http://new.bostonreview.net/BR25.5/vanparijs.html

Debate: https://bostonreview.net/forum/basic-income-all/fred-block-w...

I think it's great YC is funding some interesting research.

I'm a bit curious as to why they didn't partner with some established academics in this area (there's plenty of economists who have done excellent research here) rather than start a new program - but it's YC's money, and they can choose how to spend it.

How open do you plan to be about things like datasets/analysis plans and so forth? I think this would be an excellent opportunity to push forward open science http://osinitiative.org/about-osi/. Further, if you allow people who are skeptical of the minimum income to register the objections before you gather the data, you can address them before you proceed - at that point, post-hoc critiques are a lot less credible.

I would really like for this to be the case, but I kind of doubt it. Let's wait and see.

Gladwell is a wonderfully gifted writer who has a fantastic talent for weaving engaging narratives. Unfortunately, when it comes to science, this is NOT a good thing.

Scientific discoveries don't follow narratives: narratives are things we impose post-facto to make it easier to tell an engaging story. For example, the 10,000 hour rule makes for a nice story about effort, and Gladwell brings up a lot of examples: of course, if you look at the meta analysis, it turns out that it's completely wrong: http://www.smithsonianmag.com/smart-news/10000-hour-rule-not...

The problem with having a nice narrative is that it's too tempting to fit facts into your story for the sake of having a clean, linear narrative. If you were to write an honest book about the latest findings in social psychology, it would spend several chapters discussing statistical power and meta analysis, it would be incredibly cautious, and it would be full of results which are contradictory.

The only thing worse than social psychology in this regard is business school books. The numbers are even smaller, the tendency to pursue narratives even bigger, and the professors get paid a lot more.

Julia aside, Keno, kudos for being an absolute class act and replying to everything so politely.

For what it's worth, my impression of Julia has been overwhelmingly positive, and all the developers I've interacted with have been polite and friendly. I haven't made the switch from Python because:

- I prefer the Python syntax

- I like Python libraries (I know about PyCall, and it rocks)

- The increased speed of Julia doesn't really add much given numba/theano and so forth

However, I really like:

- Optional static typing for sanity checking

- Can write fast functions directly in julia - which is handy when passing callbacks or doing numerical routines like integration (although this requires timholy's fast lambda package)

While FB has released some excellent open source software in the past, I don't really find blog posts about closed source software to be that interesting.

The optimist in me says that the reason this hasn't been open sourced is because a lot of the distributed code is tied specifically to FB's infrastructure - but I guess we'll wait and see.

Have you read Bad Pharma by Ben Goldacre? One of the more interesting findings discussed in the book is that doctors were convinced they were personally not affected by drug company PR/goodies (3 day conferences in Bahamas, etc) but they were all worried that their colleagues were.

It's an excellent book and I highly recommend it.

The best part of the article:

"Dr. Curtis Wright, who led the agency’s medical review of the drug, declined to comment for this article. Shortly after OxyContin’s approval, he left the FDA and, within two years, was working for Purdue in new product development, according to his sworn testimony in a lawsuit a decade ago."

Says it all doesn't it?

The original business model of coursera/udacity was to offer the option to take most courses for free, but require people to pay for credentials.

This is not obviously a bad idea: the quality of teaching at Harvard or Stanford is probably better than that you obtain in most universities, but the one of the main added values are the network of people you meet, and the credential of having attended a very prestigious university.

However: the kind of person that takes a lot of online courses is probably already an educated professional, and for her, the main added value is the joy of learning new things, and the credentials themselves are near worthless.

I'm sure they have lots of very smart people who thought hard about how to best monetize their offerings, but, for me, I would definitely pay a flat fee per month just to be able to keep sampling every course I want.

When people talk about big data, they usually discuss datasets with lots of observations (aka, rows). Not datasets with lots of features but few rows - those are far more common in fields like genetics or omic sciences in general.

I prefer to get excellent service by being polite and courteous. I hate the idea that people are only entitled to a living wage when they work a service job if they are especially subservient.

That's probably the most unfortunate thing about the entire situation. Well that, and the people working for Theranos who will face the backlash.

The part about rate of improvement following an S-shape is interesting and I'd like to see it empirically researched, but the fact that the last exemplar of a given technology is always the most advanced is quasi tautological. Why would you build a new plane if it wasn't better (in some way) than what came before it? Or a new computer/phone?

I think this is largely a case of 'moloch' as described by Scott Alexander: http://slatestarcodex.com/2014/07/30/meditations-on-moloch/

In that long (but excellent essay) Scott points out that self organizing systems like human society can sometimes arrange themselves in horrible local minimas which can be very difficult to escape.

I think the way modern science is organized is decidedly suboptimal. Note that I'm talking purely sociologically: science, as a method, is still by far the best thing we've developed to understand the natural world. However, the incentives around publishing / grants / hiring work are broken in a myriad of ways, from small to massive.

For example:

- High prestige journals/conferences are basically a crapshoot: http://blog.mrtz.org/2014/12/15/the-nips-experiment.html - consider that getting a paper accepted in NIPS/published in Nature might completely change the way your career shapes up.

- Grant funding is incredibly competitive, and the way they are awarded is also a crapshot: https://psmag.com/why-the-national-institutes-of-health-shou...

- The pressure to publish/obtain grants drives people to either make up data to obtain fancy publications, or even to kill themselves: http://www.dcscience.net/2014/12/01/publish-and-perish-at-im...

- We reward researchers for brilliant clean discoveries, not brilliant methodologies. However, the outcome of a serious research project is the one thing that a scientist cannot control: science investigates the unknown, if someone explores a plausible hypothesis in a clever way, and the hypothesis turns out to be wrong, the scientist didn't do anything wrong.

- PhDs are often used as cheap labour. A very famous PI once bragged that before they got their new fancy robot, they just had 10 chinese postdocs handle all the plating.

- Private companies make an insane amount of money from publishing scientific research that's carried out with public money, and that's reviewed by scientists (largely) paid with public money who donate their money for free.

I could go on for ages - and so could almost every researcher/ex-researcher on HN. There's a few no brainer changes - but the problem is that almost everyone in power who could affect change stands to benefit from the current system.

That's a fair point - I was also surprised that the essay appeared in firstthings. However, the essay is really good, and the author is spot on about a lot of his critiques.

A better way to think of it I think is that RNA is both software and hardware.

A lot of those molecules are bound up in cells, they are not freely diffusing in blood. Even those diffusing in blood are not uniformly distributed, blood is highly dishomogeneous.

Are you trying to model metabolites as perfectly diffusing molecules with D^2 proportional to an idealized Stokes radius?

Actually, yeah. The variance in the amount of quantities present in the blood draws by Theranos is too high to get useful measurements (without very many replicates, defeating its purpose) according to published research: http://ajcp.oxfordjournals.org/content/144/6/885.abstract

I was also utterly astonished when I read this. Even if the experimental error of the instrument was zero, you cannot reduce the biological variance short of drawing more blood or doing multiple measurements. However, this is such a no brainer that the idea that Theranos skipped this step seemed too crazy to contemplate - apparently not though.

C-API compatibility, I'm guessing. I agree it was of a strange choice though.

Man, the shoe fell really, really quickly. Also - huge congratulations to the reporting team of the WSJ, that did an incredible job of unconvering this story back when Theranos was still considered a golden goose. John Carreyrou in particular did some exceptional reporting.

I think you are exactly right, and in a healthy market that's what would happen. In practice, I believe we are so far from equilibrium that it would take such a long time for the trickle down effects to push housing costs in a range which is affordable for people that aren't well off.

Telling people that within 10-15 years if you keep building several hundreds of thousands of houses eventually rents will be affordable is not a winning slogan :)

Housing in SF is an incredibly powerful illustration of the danger of letting things get so bad that the usual remedies no longer apply.

Basically:

- Traditionally, the proper fix for a shortage of housing is to build more houses. Just lift the existing regulations, and let the free market do what it does best: take care of the demand at an efficient price.

- However, after years and years of very slow building, the amount of pent up demand is such that even if builders were allowed to construct houses, the houses would almost all be built for the very well off. The amount of houses you'd have to build before you started constructing truly affordable housing is immense.

- People that are in rent regulated apartments have absolutely nothing to gain from more construction. However, even poor people that have no stable housing also have almost nothing to gain from new construction because their only realistic chance of having affordable rent is through subsidies or semi-public housing in the medium term.

- The people that have the most to gain are the moderately well off who can afford SF rents, but cannot find housing at the existing market rate.

Basically - politicians and residents let the situation with housing in SF get so absolutely bad that free-market based solutions would take a crazy amount of time to create affordable housing for poor people. At this point, I'm not sure there even are good solutions, especially if you limit yourself to what is politically feasable.

The lesson to draw from this is that you should never let the situation get so bad that people at the bottom stop caring completely about improving policy. In this case, the right policy (build A LOT more houses) would barely help poor people because the supply/demand curve is so out of wack. The very people that should benefit the most from improving the housing situation are instead opposing it - for fairly rational reasons.

I am really curious to see how this plays out.

Kudos for the 'be nice' rule - rejections really really hurt, and if they are in public, even more so. Anyone who is willing to put themselves out there like that deserves to be treated with kindness - even if the actual feedback is critical. Snark is very easy.

Out of curiosity: has anyone ever tried to do a correlation about the performance of a hard-tech company vs the ratio of technical/political people in their board of directors?

This is probably a strong prior on my part - but when I saw the composition of the board of directors of Theranos (secretary of states/generals/congressmen etc) I was immediately incredibly skeptical. It seems like the board was chosen to provide lucrative contacts instead of actually useful advice and supervision.

That would be fantastic!

I can only offer my point of view as a complete rust novice. Lifetimes are the most intriguing and new aspect of programming in rust - this unfortunately means that when people first start with the language and they run into non-obvious errors, they will need a little handholding.

Google/stackoverflow of course help, but, as an expert, I think you may be a bit biased by your own experience as to how helpful the compiler errors actually are.

Cool blog. I tried playing around with Rust as well on some toy problems, and had a similar experience: cryptic error messages that only made sense in hindsight.

It's an awesome language, but I wish there was a bit more effort put into into making it 'ergonomic'. Of course, it's an open source effort so nobody is owned anything - but useful error messages are incredibly important, especially when the language pushes so many new concepts.

Thanks for taking the time to explain that :)

Hey Sam, semi technical question here. NIPS, ran an experiments a few years ago where for 10% of the papers they had two completely independent teams of reviewers assess papers and score them - the agreement between the teams was very low: http://blog.mrtz.org/2014/12/15/the-nips-experiment.html (barely better than chance, actually).

Would you consider doing something similar for YC? Obviously, you already have multiple people looking at websites/applications, but I'd be really curious to see what the agreement is like - and you have some absolutely exception AI/data scientists people at openAI who could crunch all the numbers.