HNHacker News
TopNewBestAskShowJobs

michael_nielsen

8,027 karma · joined July 24, 2007

https://michaelnotebook.com
submissionscomments
michael_nielsen··on Please learn to write
I once asked a well known science writer how many times he will revise passages. He sighed and replied "sometimes, for the difficult spots, maybe fifty to a hundred."

This made me feel much better about my own writing, which often goes through many, many drafts.

It also helped me appreciate Thomas Mann's maxim that a writer is someone who finds writing harder than other people.

michael_nielsen··on How not to ask questions at a conference
Organizers can do a lot to improve the quality of Q&A. A clever hack used by the Long Now Foundation for their Long Now talks is to ask audience members to write their question down on paper, and pass the paper to the front. The best questions are then read out.

This has many effects. It eliminates long-winded and meandering questions. It also eliminates questions primarily intended to puff up the questioner. Unfortunately, this latter category is probably the majority of questions at tech conferences I've been too.

The result is that Q&A is often as good or better than the talk, and may continue for 45 minutes. At many events this would be torture; at the Long Now talks it's a pleasure.

michael_nielsen··on Ask Valve employees: How well does the flat organization work?
After the first night of the 2008 SciFoo conference at Google, @gnat Torkington's twitter stream contained a gem: "At SciFoo opening session. 300 people standing around networking. Meanwhile, Larry Page is quietly unpacking chairs at the back of the room." (from memory, not exact)
michael_nielsen··on Show HN: iPG - YC Interview Simulator
The person who is perhaps the brightest [1] person I've ever met would - if this is a fair simulation - be rejected by YC. When answering a question he often won't even have begun to reply after fifteen seconds. I wonder how often YC rejects people such as that.

[1] I mean this in a particular way: capable of extraordinarily deep insight very, very rapidly. Just not 15 seconds rapid.

michael_nielsen··on How Andreessen Horowitz Fumbled An Instagram Investment
I think you're reading something into the article that isn't there. I saw no vitriol aimed at picplz.
michael_nielsen··on One hiring filter that works
I've never been involved in hiring at a startup, but for 8 years I was involved in hiring as a prof in academia.

In that context, there is a peculiar and consistent gap between the best hiring strategy and the strategies institutions actually use.

For hiring faculty whose main focus is research, the best strategy is targeted attempts to hire superstars -- the research equivalent of a Guido van Rossum, say, or a Linus Torvalds, or someone near that level -- or early-career people where you have very good reason to be confident they will be future superstars.

It sounds obvious to say "go after superstars". But what's strange is that in any given field only a very short list of institutions actually do it. At most academic institutions it's surprisingly difficult to convince the institution to go after superstars.

I think the reason for this aversion is instinctive. If you're hiring it's much more comfortable to go the conventional route, advertising a position, interviewing a short-list and so on. This produces a situation where most of the interviewees want the job pretty badly. This gives the hiring committee a lot of power, which is a comfortable position to be in.

Unfortunately, it also produces lower-quality candidates than identifying the best people and going after them directly. A superstar will be interviewing you more than you are interviewing them. You'll need to get creative in convincing them that you have something to offer, something they can't find anywhere else. This is much less comfortable.

A useful heuristic: if most of your job offers are being accepted, your standards are way too low.

michael_nielsen··on James Cameron begins deepest dive
Cameron hit the bottom at 14:52 PST, according to @PaulGAllen, and "all systems ok".

Here's Cameron's tweet announcing his arrival: "Just arrived at the ocean's deepest pt. Hitting bottom never felt so good. Can't wait to share what I'm seeing w/ you @DeepChallenge"

https://twitter.com/#!/JimCameron/status/184036733959143425

michael_nielsen··on James Cameron begins deepest dive
Awesome! According to @PaulGAllen, they hit at 0752 local time, and "all systems ok"!
michael_nielsen··on James Cameron begins deepest dive
The official Twitter account: https://twitter.com/#!/DeepChallenge

According to tweets from @PaulGAllen and @PeterDiamandis, Deep Challenge should arrive at the bottom around 14:40 PST.

Edit (14:50 PST): No word from @PaulGAllen since 14:20 PST, when he noted Cameron was at 32160 feet and "not long to seabead now". I hope the people on the support crew are just really busy because Cameron is touching down on the bottom!

michael_nielsen··on New: Apply to Y Combinator without an Idea
Seems similar to Alan Kay's often repeated observation that what made ARPA and PARC so innovative in the 1960s and 1970s was that they "funded people, not projects": http://www.vpri.org/pdf/m2004001_power.pdf

This is very, very rare.

michael_nielsen··on IBM shows off quantum computing advances, says practical qubit computers close
Unfortunately, the statement is incorrect. Holevo proved in 1973 that n qubits can only store n classical bits of information. See: http://en.wikipedia.org/wiki/Holevo%27s_theorem

The statement about performing logic on all that data in parallel is misleading, at best.

michael_nielsen··on IBM shows off quantum computing advances, says practical qubit computers close
The actual paper: http://arxiv.org/abs/1202.5533
michael_nielsen··on Newspaper Ad Revenues Fall to 50-Year Low in 2011
Here's Marshall McLuhan on this, in 1964: "The classified ads (and stock-market quotations) are the bedrock of the press. Should an alternative source of easy access to such diverse daily information be found, the press will fold."
michael_nielsen··on Open Access Bill for (nearly) all US publicly-funded research
The short summary is that this act (FRPAA) will require free online public access to most publicly-funded research in the US.

Many more details are available here: http://cyber.law.harvard.edu/hoap/Notes_on_the_Federal_Resea...

michael_nielsen··on Applications open for Y Combinator's summer 2012 funding cycle
The term "social enterprise" is used in many ways, and doesn't seem to have a clearly accepted legal definition. Without clarification of what you meant, it's difficult for me to comment directly on your suggestion.

However, with that said, I'd just like to emphasize that I'm certainly not criticising for-profits. I'm saying not-for-profits are different, and have a whole different set of advantages and disadvantages than for-profit enterprises.

michael_nielsen··on Applications open for Y Combinator's summer 2012 funding cycle
Any YC partners who wanted to do this could set up a parallel not-for-profit or charity. That organization might benefit from some of the easily scalable benefits of YC (say, providing access to the YC alumni network), but would be independent. It wouldn't even necessarily need to be funded by YC partners - there are foundations which might be interested in partnering with an organization like YC to fund this kind of thing.
michael_nielsen··on Applications open for Y Combinator's summer 2012 funding cycle
Would YC ever fund a not-for-profit startup, as a form of philanthropy?

Part of the reason I ask is that for-profit and not-for-profit organizations have different kinds of leverage, and for some activities not-for-profits have big advantages. It's striking that some major success stories like Khan Academy and Wikipedia are not-for-profit.

I imagine that having a YC-like model applied to not-for-profits might produce some remarkable new startups, maybe even entirely new types of startup.

michael_nielsen··on For HN: Open access to scientific papers, a summary of the state of play
Thanks for pointing this out - I'll fix it.

(On the email: yes, I sent them. Don't know what happened to yours, sorry about that. But the book is out, and available at places like Amazon.)

michael_nielsen··on For HN: Open access to scientific papers, a summary of the state of play
My understanding is that they are making a great deal of money, which they reinvest in the organization. The linked post has a link to a document which includes their balance sheet. You are, of course, correct that they're not-for-profit.
michael_nielsen··on For HN: Open access to scientific papers, a summary of the state of play
I don't have a lot to add. It's not an issue I've spent a lot of time on (not because it's not important, just because of limits on time).
michael_nielsen··on For HN: Open access to scientific papers, a summary of the state of play
At the moment the impact of these actions seems significantly smaller than policy. But perhaps it's a useful complement, showing policy-makers that many faculty support open access.

As regards your CS/math question, it's a good one, and I don't know enough about the cost structure of the journals in question to answer. I suspect from conversations with journal editors that it's not a whole lot cheaper to produce CS/math journals than other types, but that's just a general impression, not a certainty.

michael_nielsen··on For HN: Open access to scientific papers, a summary of the state of play
I'm not sure it's easy for many startups in any space to become profitable.

However, it is possible.

One open access startup that has met this challenge and come to profitability is the Public Library of Science: http://blogs.plos.org/plos/2011/07/2010-plos-progress-update... They are now making a great deal of money, which is presumably why organizations such as Nature Publishing Group are looking at replicating PLoS One.

There is a lot of online discussion of business models for open access startups. Here's one useful guide: http://www.arl.org/sparc/publisher/incomemodels/

michael_nielsen··on For HN: Open access to scientific papers, a summary of the state of play
As a blog post:

http://michaelnielsen.org/blog/open-access-a-short-summary/

I've now edited the header of the HN submission to point to this URL.

Thanks for the suggestion --- I should probably have done that initially. My thinking was that I didn't want to put it on my blog, since it's meant specificially for HN. But with a little framing, that issue takes care of itself. And, as you say, the links work a lot better in a blog post.

michael_nielsen··on For HN: Open access to scientific papers, a summary of the state of play
There's more and more pressure for that kind of thing, but it's still early days in most fields. For an example of a forward-thinking policy, consider the Wellcome Trust: http://www.wellcome.ac.uk/About-us/Policy/Policy-and-positio...
michael_nielsen··on Publicly funded science should be open science
The open access to publication argument already has nearly unstoppable momentum. In 2008 the NIH --- by far the world's largest funding agency, with a $30+ billion dollar a year budget --- adopted a policy requiring that all NIH-funded research be made openly accessible within 12 months of publication. See, e.g.: http://www.earlham.edu/~peters/fos/nihfaq.htm

Similar policies have been enacted by all 7 of the Research Councils in the UK. And there are many more such policies in place at institutions all over the world, including Harvard, MIT, Princeton, and many others.

A treasure trove of information about open access is Peter Suber's website. See, for example, his timeline: http://www.earlham.edu/~peters/fos/timeline.htm

The timeline stops in 2008 (he changed jobs), but if you look at progress up to that point, it's clear that this has enormous and rapidly increasing buy-in at grant agencies, libraries (who are the main "customer" for journals, in that they pay) and major universities. It will be very difficult for other parties to stop this change, provided people continue to work for it.

At this point, momentum is starting to build for more open data, and also for other kinds of systematic online sharing of research information. Why not give researchers an incentive to (say) upload online videos showing how certain lab experiments are done? At least in some labs there's stuff that's very difficult to describe in a paper, but easy to show in a video. And so this would be very valuable. (There are initiatives underway to do this, e.g.: http://www.jove.com/ )

I entirely agree with your final point - "The system of funding and promotion is going to have to undergo a very big shift before this dream comes to life". There's enough traction now within universities to start to have this discussion. Certainly, some senior scientists are dismissive; but many are extremely interested, and engaged.

michael_nielsen··on Publicly funded science should be open science
This factor is not left out, it's at the core of the article:

"Even if you personally think it would be far better for science as a whole if you carefully curated and shared your data online, that is time away from your "real" work of writing papers. Except in a few fields, sharing data is not something your peers will give you credit for doing."

There is no question that in most (not all) fields it is not yet in an individual scientist's best interest to spend a lot of time systematically sharing data. The question is: might it sometimes be in the community's best interest?

This argument played out, for example, with the human genome in the 1990s, where many scientists were initially reluctant to share data (little individual reward for the time and effort). Ultimately, grant agencies required very early-stage sharing, something almost everyone would agree has benefited science as a whole.

The point of the article is that there needs to be a change to bring individual incentives for this kind of systematic sharing into alignment with the community's best interest.

michael_nielsen··on Who listens to scientists? Mostly just other scientists.
In the US, you're probably thinking of the NIH's open access policy. The NIH is the largest research agency, with a more than $30 billion dollar budget, so this does cover a lot of US research, but it doesn't cover research from many other agencies.

For much more on the NIH policy, see the links in this article: http://www.earlham.edu/~peters/fos/2007/12/oa-mandate-at-nih...

There has been a concerted (though not yet successful) effort to extend this type of open access mandate to all US federal agencies with budgets over 100 million. See, for example: http://www.arl.org/sparc/advocacy/frpaa/index.shtml

With a lot of work, I believe these efforts are likely to be successful. But there is real opposition, not just from some journal publishers, but also from some politicians. See, for example, the efforts by John Conyers (D-Mi) to roll back the NIH open access policy: http://www.earlham.edu/~peters/fos/newsletter/10-02-08.htm#n...

In the UK, all seven UK Research Councils have similar open access policies. The last one to adopt such a policy was the Physical Sciences and Engineering Council, which I believe is the one relevant to computer science research. That policy was adopted in 2009, I believe, and so we're only starting to see the effects now.

michael_nielsen··on Perfect eavesdropping on a quantum cryptography system
I haven't verified the results of this particular paper, but there is a broad class of potential problems along these lines with quantum cryptography. Here's a short piece I wrote about these problems in 2004: http://michaelnielsen.org/blog/whats-wrong-with-those-quantu...

The short summary is that side channel attacks are a real problem with quantum cryptography, potentially moreso than with conventional crypto. The reason is that only a limited number of people have the resources to seriously test (and thus improve) real-world implementations of quantum cryptosystems systems. I.e., quantum crypto greatly raises the bar to do white hat hacking.

michael_nielsen··on Pregel - distributed graph processing | DDI
I haven't heard of any publicly available implementations of Pregel, per se.

The Pregel paper does describe some other related and publicly available frameworks for graph algorithms, but I haven't played with any of those frameworks. Details are in the second last section of the paper: http://www-bd.lip6.fr/ens/grbd2011/extra/SIGMOD10_pregel.pdf

michael_nielsen··on The Clock in the Mountain
I also recommend Stewart Brand's book about the (then-hypothetical) Clock: http://www.amazon.com/Clock-Long-Now-Stewart-Brand/dp/produc...

In the book, Brand argues that: (1) it's very difficult to have a general conversation about the 10,000 year future without descending into vapid generalities; but (2) with a concrete engineering goal like building a clock, it's possible to have much more focused and interesting discussions.

The Long Now seminar series is also excellent: http://longnow.org/seminars/

← PreviousPage 8 of 17Next →