Extreme Points

Ex-FCC Commissioner Reed Hundt on Telecoms Policy

2008-03-01T16:56:00.000-08:00

Reed Hundt gave an interesting interview to Telephony Online, covering a range of topics including broadband deployment and wireless spectrum policy.

Video Traffic and Net Neutrality

2008-02-12T09:10:00.001-08:00

The Associated Press has an article about the impact of increasing video traffic on the network neutrality debate.

Facts and Fiction About Cable Cuts

2008-02-06T22:57:00.000-08:00

See this Wired post for a description of some of the issues around the Middle Eastern cable cuts. One surprising statistic is that cable cuts happen on average once every three days.

Over $4.6 Billion

2008-02-06T22:56:00.001-08:00

The reserve has been met in the FCC auction...

Stalled at $4.3 Billion

2008-01-30T21:56:00.000-08:00

After rising to $4.3 billion, no additional bids were made on the C block today. As this NYT article explains, tomorrow is an important day for the C block auction.

IEEE Spectrum 700 MHz Auction Primer

2008-01-29T08:28:00.000-08:00

IEEE Spectrum Magazine has a primer on the 700 MHz spectrum auction.

Do You Believe the MPAA?

2008-01-26T11:13:00.000-08:00

John Heidemann, a researcher at USC (and coauthor of my colleague Ashish Goel), is quoted in this Cnet article about whether we should believe the MPAA's claims about the extent to which illegal filesharing takes place in college campuses.

Good Mathematical Modeling

2008-01-25T14:23:00.001-08:00

Sitting in a program committee meeting as I write this...

I interpret mathematical modeling as the act of representing a "real" system through a mathematical formulation, for purposes of both quantitative and qualitative insights--into design, operation, and performance. One of the key features of good models is that they retain enough connection to the system to make the qualitative insights robust, and yet remain parsimonious enough mathematically to be tractable. I am thus often surprised that the reaction to good modeling papers is to claim, "But they didn't include Feature X of the system!" Now, if Feature X is necessary to have a robust model, then this is a fair claim. But often, Feature X is an orthogonal feature, that only adds extraneous detail and complexity to an otherwise elegant system representation. As an advisor of mine once said, the best work is done at the boundary between the trivial and the intractable; I say kudos to those who walk this line successfully.

Nice FAQ On 700 MHz Auction

2008-01-25T14:22:00.001-08:00

The Wired article cited in the previous post also has a link to a nice FAQ on the 700 MHz auction.

First Day of 700 MHz Auction

2008-01-25T13:44:00.001-08:00

Bidding in the 700 MHz auction has begun; see here for stats after the first day. The FCC required $4.6 billion to be bid on the C block to have the open access regulations bind; currently the bidding on C is at $1.5 billion, though Google has proclaimed they will ensure the reserve is met.

Interesting Comments From FCC Commissioner

2008-01-25T13:39:00.000-08:00

This post on CNet relays thoughts of an FCC commissioner about recent moves by providers such as Verizon to preemptively implement open access policies on their networks. Interesting questions of whether it is all smoke and mirrors on the providers' part or not...

OpenCourseWare and Walter Lewin at MIT

2007-12-18T22:28:00.000-08:00

This NYT piece details Prof. Walter Lewin's Internet popularity thanks to physics lectures made available through MIT OpenCourseWare.

The End of Industrial Research Labs?

2007-12-18T09:55:00.000-08:00

This piece in the New York Times details the changing nature of academia-industry partnerships, a big theme in the "Silicon Valley comes to Cambridge" event I attended in mid-November. I was an intern at Bell Labs and must say the environment there was remarkably close to a university. It is interesting to watch research change inside of corporations -- for example, Yahoo! and Google take dramatically different approaches to research. The former is much more academic, while the latter seems to believe it is better to "do first, prove later."

Efficient Capital Allocation

2007-11-29T16:59:00.000-08:00

I had the opportunity to talk with a couple of venture capitalists on a recent trip to the UK (part of "Silicon Valley comes to Cambridge"). I found the numbers interesting: a typical VC will look at several hundred prospects each year, seriously consider perhaps 10%, and fund 2-3. The lifetime of an average company is 6 years before the VC will exit, which means a relatively small steady state portfolio.

I wonder about the welfare effects of this approach to capital allocation. Essentially a single decision maker investigates an investment in the setting of a VC, and the variance of return on any given investment is quite high. I'm curious whether an economy that increasingly employs venture capital to fund innovation is essentially choosing a huge increase in variance on its return on investment.

Verizon Wireless

2007-11-29T16:56:00.001-08:00

Exciting news this week: appearing to preempt the license rules that are likely to be enforced on the C block auctioned in January, Verizon Wireless announced an open application and open device policy on their network beginning next year.

Details are here.

Google's Open Handset Alliance

2007-11-07T22:42:00.000-08:00

Since I was away at the INFORMS Annual Meeting, I wasn't on top of Google's press release Monday about the Open Handset Alliance, their new initiative to encourage creation of "open" wireless devices. They have on board an impressive list of 34 companies ranging from service providers to device manufacturers. See here for details.

Facebook and marketing

2007-11-07T13:09:00.001-08:00

This New York Times article has some interesting info about Facebook plans for "hooks" for advertisers.

OpenSocial

2007-11-01T08:53:00.001-07:00

Another Google related post: together with a host of other companies, Google announced an "alliance" for an API to counteract Facebook's platform. See this article for details.

gPhone

2007-11-01T08:47:00.000-07:00

Lots of chatter now about Google developing a platform for wireless devices. See, for example this Wall Street Journal article.

Open Platforms and the 700 MHz Spectrum Auction

2007-10-18T10:39:00.001-07:00

As is well known by now, on October 5 the FCC announced the details for Auction 73 for the 700 MHz spectrum, to be held January 24, 2008. Some observations about the C block from the FCC auction auction rules document and the license rules document here:

The C block includes spectrum at 746-757 MHz, and 776-787 MHz.

Spectrum is paired across the two bands.

There are 12 licenses total in the C block: 8 regional licenses (across the 50 states); 2 "Atlantic" licenses (US Virgin Islands, Puerto Rico, Gulf of Mexico); and 2 "Pacific" licenses (US Pacific territories). Package bidding is allowed on the national licenses, on the Atlantic group, and on the Pacific group.

The C block has a reserve of $4.6 billion; if this reserve is not met, then the spectrum will be re-auctioned (as part of Auction 76), without the open platform requirement.

Network Neutrality and Provider Investment Incentives

2007-10-12T19:40:00.000-07:00

Just returned from the Bay Area Algorithmic Game Theory Symposium, with some nice talks; in particular, Jean Walrand gave a really nice presentation of a simple model to investigate provider incentives in the network neutrality debate. I thought his presentation was a good example of why simple mathematical models can be useful: if you ask the right questions, then the insight delivered makes the abstractions worthwhile. His paper can be found here.

Social Networks: Fad or Fundamental?

2007-10-12T07:00:00.000-07:00

This post from the New York Times is very appropriate for my current frame of mind. Lots of researchers with my background seem to be drawn to the broad area of marketing on social networks, ostensibly because of the intersection of interesting math, statistics, and computer science with a practical problem focus. It helps that a large body of recent research is emerging that aims specifically to improve our ability to solve large scale optimization and inference problems on an underlying network structure.

All that said, I wonder whether we are working on a set of problems that are really fundamental; in particular, given that success in viral marketing probably involves seeding a lot of experiments (i.e., that success probabilities are rare), is it likely that we can design algorithms that significantly leverage social data for marketing purposes?

Bay Algorithmic Game Theory Symposium

2007-10-09T22:47:00.001-07:00

The biannual Bay Algorithmic Game Theory Symposium will be held this Friday, October 9, 2007, at Yahoo! headquarters in Sunnyvale. See here for complete details.

This event has been steadily diversifying, including a wide range of researchers from computer science, economics, operations research, and electrical engineering. One of many reasons it is great to be an academic in the Bay Area!

Blue Angels over San Francisco

2007-10-09T22:29:00.001-07:00

The Navy Blue Angels were visiting San Francisco last weekend; here's a shot of the six of them in delta formation over the Golden Gate Bridge.

I've gone to see them each of the last three years--I think from a pure engineering standpoint, some of the feats the F/A 18's manage are rather amazing. However, it bears mentioning that not everyone in SF is a Blue Angels fan, for a variety of reasons ranging from pure pacifism to a desire for peace and quiet. See here for details (scroll down to "Blue Angels - or Demons?").

University lectures and iTunes

2007-10-08T19:10:00.000-07:00

As you can see on the right side panel, I had the good fortune to have five lectures I gave on "The Future of the Internet" posted on iTunes (through Stanford Continuing Studies). When I was asked if Stanford could record them, I didn't think twice about it--it seemed like a great opportunity, and since then I've received e-mails from all over globe with a variety of insights, comments, and (thankfully) praise.

Over the weekend, I had an interesting conversation with a former university legal counsel about some of the issues involved here. Not surprisingly, university agreements about posting course content online can be fraught with legal danger. In the case of iTunes, the problems begin with ownership: it is the lecturer who owns the lecture (the audio, lecture notes, etc.), and thus if the university makes agreements with third parties (such as iTunes) to distribute course content, it must also make sure that lecturers have the right to "opt-out" (although technically speaking even this is not good enough, since the university doesn't own the content in the first place...)

The issue gets thornier from there. As the lawyer I was speaking with pointed out, a university has at least three key arguments against free dissemination of recorded course material:

1) Brand dilution: This argument basically says that by posting its content online, a university is reducing the value associated with the product delivered by the institution--namely, a degree.

2) Student participation: In many cases, students are recorded on posted lectures; in some cases students may not want their questions or comments to be posted with such recordings.

3) Third-party content: This is a big one, and the administrative headache can easily become a migraine. It is in principle illegal to post any lecture that uses content where the copyright owner has not given authorization. For purposes of lectures in the classroom, e.g., where an audio recording or video is used as part of the classroom lecture, typically such issues can be avoided by appealing to "fair use." But this exit can't be used for posting of content online.

What surprised me most about this conversation was the extent to which I hadn't thought through the issues myself before agreeing to have my lectures posted on iTunes. I am generally in favor of open course content, but at the same time this conversation made me think harder about the depth of issues universities have to deal with in posting course content.

Welcome to Extreme Points!

2007-10-07T08:33:00.000-07:00

Over the last year or two I've seen a variety of research activities that did not lead to publishable insights--either because the results were not novel enough to disseminate, or because "insight" is not necessarily the same thing as "research." So, I'll be using this blog to store those thoughts, and hopefully others will find them helpful as well. Given my broad range of interests, I expect posts to cover the gamut from applied mathematics to networking technology to policy and economic issues.

(I also reserve the right for the occasional off-topic posting... )

On the title: The term extreme point has its origins in the mathematics of convex sets. (A set is convex if it completely contains any line segment between two points of the set.) Formally, a point is an extreme point if it never appears on a line segment between any other two points in the set. In the context of this blog, my "extreme points" represent solely my opinion, and not that of my coauthors, research sponsors, or Stanford University.