Tuesday, October 29, 2013

Visiting UC Davis

Before returning home last week I spent a day at UC Davis giving a talk.  I got the spend the day talking to various people I've collaborated with there (like Raissa D'Souza, Nina Amenta, and John Owens), and enjoyed getting to see and talk with Charles (Chip) Martel, who when he's not busy as a computer science professor is busy being a world championship bridge player.  I spent several nights up late last month watching him play (via bridge's version of an Internet broadcast) online in the world championships in Bali. 

The big news was that Davis looks to be hiring this year, and one of the areas might be some variation of "big data", with theoretical types definitely included that category.  Davis is nicely located about an hour's drive north of Berkeley (in reasonable traffic), and enjoys standard California amenities.  (The day there was filled with ridiculously nice weather and excellent food.)  If you're job searching, you should look for their job announcement.

Thursday, October 24, 2013

Mitzenmacher Drinking Game?

I've been visiting the Simons Institute for one of their workshops the last few days.  I got my advisor Alistair Sinclair to give me a tour.  I have to say, that's an amazing space they have.  A very nice building, and they've set it up wonderfully (lots of light and open working spaces).  I can't believe their location on the Berkeley campus. 

After my talk, someone told me (and now I forget who) that someone (Michael Goodrich?) said that the Mitzenmacher talk drinking game was that you took a drink every time a slide says "Bloom" or "cuckoo".  That would be a dangerous drinking game.  (It's funny because it's true.)

Sunday, October 20, 2013

NSF Coming Back Online

I was impressed how quickly the NSF got Fastlane back online.  (I think it was turned on a few hours after the shutdown was ended.)  But like many people, I'm awaiting to hear how the missed deadlines (or upcoming deadlines) will be manged.  I haven't seen word on that on the web site, but looking online I found this article.  It suggests that "BFA will establish and publish on the NSF website within one week agency-wide policies for proposal deadline extensions and other grant-related actions."  (I am not sure what the BFA is.)  That pretty much matches my expectations.  I can imagine it's very difficult catching up after being shut out of your office for two weeks, especially when having to deal with the concerns of hundreds/thousands of scientists who want to find out what's going on.  I'm not at all surprised that they have to take several days to figure out what the best way forward is.  Heck, it probably will take several days just to catch up with the weeks of e-mail and other paperwork. 

I think the article is encouraging patience on our part.  If you're awaiting word, just check into the NSF web site this week.  If you see or hear anything important, try to help spread the word.  The NSF has, in my experience, been very reasonable about dealing with things, and I'm sure they're working to figure out a way forward that is effective and fair to the people waiting to turn in proposals or otherwise participate in NSF activities.

Thursday, October 17, 2013

Simons Research Fellowships Post

Alistair Sinclair asked me to post a note about the Simons Research Fellowships positions at Berkeley.  But since that's where Justin Thaler (my now-ex student) ended up, I thought he could put in a few words about his experience to add a personal effect.  Justin says...
Having been a Research Fellow at the Simons Institute for a couple of months now, I cannot speak highly enough about the experience. Others have already given a sense of how exciting it is to be here, so I'll just briefly list some of the things I find most striking about the place.

* The atmosphere is surprisingly collaborative, even for a place specifically designed to foster collaboration.
* Any time I have a question I can't answer, there is an expert's door I can immediately knock on.
* There's a great mix of junior and senior people.
* It's particularly fun hanging out with other Research Fellows! And I look forward to every Friday, when we all meet for lunch and one of us gives a short, informal talk on, well, whatever that person wants to talk about.
* Seriously, what's not to like about having no non-research responsibilities, surrounded by dozens of top researchers in the same boat?
With that very positive description, here's the formal information:


The Simons Institute for the Theory of Computing at UC Berkeley invites applications for Research Fellowships for academic year 2014-15.

Simons-Berkeley Research Fellowships are an opportunity for outstanding junior scientists (up to 6 years from PhD by Fall 2014) to spend one or two semesters at the Institute in connection with one or more of its programs.  The programs for 2014-15 are as follows:

* Algorithmic Spectral Graph Theory (Fall 2014)
* Algorithms and Complexity in Algebraic Geometry (Fall 2014)
* Information Theory (Spring 2015)

Applicants who already hold junior faculty or postdoctoral positions are welcome to apply. In particular, applicants who hold, or expect to hold, postdoctoral appointments at other institutions are encouraged to apply to spend one semester as a Simons-Berkeley Fellow subject to the approval of the postdoctoral institution.

Further details and application instructions can be found at http://simons.berkeley.edu/fellows2014.  Information about the Institute and the above programs can be found at http://simons.berkeley.edu.

Deadline for applications: 15 December, 2013.

Wednesday, October 02, 2013

Some Advice on Entrepreneurship from the AH meeting

At the Andreessen Horowitz academic round table (see past post), there was various advice, some of it contradictory, for professor-types interested in starting companies.  I should start by saying that all of this is my interpretation of the advice, and the various people involved are not responsible if I've gotten it wrong.  Certainly further opinions are welcome.  

At the research level, Nick McKeown expressed some of his rules for research projects.

1. Pick a problem that is intellectually interesting.
2. And improves the practice.
3. And industry doesn't like (yet).

His idea (my interpretation!) was that if industry liked the idea, then the problem wasn't out there enough in terms of a time horizon.  Also, given the resources in industry, if industry really liked an idea, they could throw more resources at it than researchers at a university.  This idea had some pushback.  For example, Michael Franklin said the AMPLab at Berkeley had a lot of enthusiasm from industry, and that the industry support and interest was very positive in terms of their success.  (AH very recently funded a startup that came out of the AMPLab.  And the AMPLab is very impressive in terms of its resources, including people -- lots of great faculty.)

I will say that part of Nick's conception resonated with me.  When I've expanded my research into new areas, I've found at times that people in that area can be very negative.  And when that happens, it often turns out to be the most interesting research.  The work on low-density parity-check codes from way back when was roundly criticized by multiple old-guard coding theorists when we first presented it, and then suddenly there were entire sessions at coding theory conferences on the subject.  If you're inspiring angry reactions, you may indeed be on to something in your research.  (Of course, as he also acknowledged, you may also just be wrong.)

Another key issue that arose was "commitment".  The VCs at AH expressed some skepticism for professors who wanted to take a year (or maybe two) off to help set up a company but then hand it off and go back to their regular job.  Besides investing in an idea, they're investing in a team, and it's not a great sign if the team leader isn't committed.  Also, there's the feeling that that sort of change in leadership can have a huge transition cost.  (Also, I think, as I mentioned previously, they really seem to like working with tech CEO's.  Handing a company off may remove the "tech" leadership.)  They were fine with a model where it was professors and graduate students starting a company, and the commitment was coming from the graduate students;  in that case, a "part-time" professor founder could be workable.

I personally think the "commitment" issue can be a challenge.  It's a problem for me, with liking my regular job so much.

There was various talk about patents.  Most of the crowd were against making them a priority in starting a business, and recommended not getting them.  A patent, it is said, is just a license to sue, and who needs or even wants a license to sue?  Making your work open-source to get excitement and interest, and then commercialize after that point, can be a very successful business model, and maintains the academic desire to get the basic work out to a wide audience.  But perhaps most importantly, as an academic, patenting your work means dealing with your school's version of the Technology Licensing Office, and nobody says anything good about their Technology licensing offices.  (Even the west coast schools -- the best anyone said is that generally theirs would stay out of your way.)  A patent gives your TLO license to ask for whatever percentage of your soon-to-be company they feel like asking for, and until they sign off, it can be hard to get a VC interested in a company where another entity holds the patent.  And generally speaking, your success is not their performance metric.

(Two quick additional points related to TLOs.  Many noted that most TLOs have been brought up by the medical school/biology groups in the university, where patents matter a lot and licensing is a strong way to go.  That's much less so in CS tech.  Also, while you'd think TLOs would be thrilled to have professors/graduate students start companies, get very wealthy, and donate back to the university -- that doesn't seem to be their model, as enlightened as it might seem to us.)  

I certainly know some cases where patents at least seem to me to have been important to "startup" companies.  (Akamai/Limelight?)    But the rest of what was said seemed to make a lot of sense.   

One thing the VCs emphasized is the importance of timing.  They said the idea behind many startups had actually been around for a while, the subject of study in universities or labs, but often the timing has to be right to make the move into the product space.  Sometimes other technology has to catch up, or something else in the environment has to go your way.  A lot of "failures" may not be failures of the idea, but just not the right timing.  

It was expressed that startups that have a mission to change the world in some interesting way were better off than startups whose mission was just to make money.  In particular, it can help recruit the best and the brightest, creating a very powerful positive feedback loop.  Simply stated missions -- Google says theirs is to "organize the world's information and make it universally accessible and useful" -- can be quite powerful.  The cynical might question people's motivations and whether money is really what's behind the mission, but either way, missions can be useful.

Finally, and perhaps this isn't so surprising, but the best way to connect with VCs is probably through personal connections.  If you know someone that's worked with the VCs, get them to introduce you.  The VCs apparently get thousands of proposals sent to them a year, and very few of those get consideration.  Having someone they trust vouch for you means a lot in terms of them being willing to make time to listen to you.  That's not unlike aspects of academia (the importance of letters for jobs/graduate school;  in some cases connections being helpful for getting funding from industry).  While not surprising, it seems worth saying. 

Tuesday, October 01, 2013

Entrepreneurship and the Curriculum

At the Andreessen Horowitz academic round table (see last post), the issue of how to promote student entrepreneurship through the curriculum arose.  The VCs at AH (which I'll use for short hereon) want there to be more tech-based CEOs.  As they put it, it's easier to teach a tech person what they need to learn about business than to teach a business person what they need to learn about the tech.  Somehow, in most universities and I believe in the world at large, the culture has developed that the business people think they're the powerful ones, not the tech people who build the things that consumers love.  The business people think they're the ones delivering the value, and then divide the value accordingly.

Don't believe me?  Go see http://whartoniteseekscodemonkey.tumblr.com/ , a site (that came up in the discussions) devoted to the e-mails sent by Wharton business school people looking to hire (undergraduate) programming and engineering talent.  As a faculty member I get bunches of these sorts of e-mails a month, and I'm sure the computer science students do as well.  The underlying message is that the tech people are commodity cogs to be plugged in as needed.  That's not the message we want our students to get, and not how things really work in most successful startups.  

So AH says they believe in and support the tech CEO, and want to encourage that.  What does that, and entrepreneurship generally, mean for our curriculum?  Should CS departments have courses on entrepreneurship (or give credit for classes in other departments on the subject)?  Should we teach computer languages that are the latest on the start-up scene (in preference to those that, arguably, provide a larger conceptual framework or encourage certain ways of thinking)?  Should we have an "entrepreneur track", like we might have a theory track or AI track or computer science and engineering track?  What is the school's role at the department's role in encouraging entrepreneurship?  Some people thought CS departments perceive themselves as professors in the business to make more professors, and thereby ignore the potential CS has to change the world via business.

These are tough questions.  One issue that makes it even more problematic for CS is that these problems are not faced by many other parts of the university -- literature, history, and even most of the social sciences don't have a significant start-up scene -- which means in some ways, we're on our own.  Indeed, significant parts of the university might actively resent an emphasis on entrepreneurship, which they might argue does not always fit so well with the university's educational mission.  (Or,  perhaps, it just represents self-interest on their part.)

Aged fuddy-duddy that I am, I'm somewhat sympathetic to this view.  Computer science is science.  I want to educate students about the great questions (and answers) of computer science, and I am thrilled to be educating the next generation of scientists, especially computer science.  But yes, computer science is also engineering (in the practical sense of the word), providing the ability to solve immediate problems, yielding economic benefits to the users and of course the developers of the solutions.  I see striking the right balance as a challenge;  greedily, I do somehow want both.

At Harvard, I feel we've been pushed and pushed ourselves to make the requirements for the major quite minimal (in terms of the number of classes), so I want those required classes to be on the "science" side.  I want computer science graduates to have both breadth and depth in computer science.  Much of the entrepreneurship can naturally fall outside the curriculum -- there are now a number of student organizations, and university-level initiatives, to promote entrepreneurship.  (Harvard, I think, has been finding a way to expand the concept of entrepreneurship beyond just "business" -- into the larger concept of innovation -- to make it more appealing throughout the university.  For example, check out the i-lab.)  At the same time, I'm clear that having all the entrepreneurship activities fall outside the traditional curriculum potentially pushes a set of students away.  Again, we're left with finding the right balance, for us.

Sadly, the meeting's discussion on this only lasted a short while, and I felt left with more questions than answers.  Feel free to discuss your thoughts here.