Broken Symmetry

What Your Computer Does While You Wait

Great piece of educational blogging by Gustavo Duarte:

Normally when the CPU needs to touch the contents of a memory region they must either be in the L1/L2 caches already or be brought in from the main system memory. Here we see our first major hit, a massive ~250 cycles of latency that often leads to a stall, when the CPU has no work to do while it waits. To put this into perspective, reading from L1 cache is like grabbing a piece of paper from your desk (3 seconds), L2 cache is picking up a book from a nearby shelf (14 seconds), and main system memory is taking a 4-minute walk down the hall to buy a Twix bar.

...

Sadly the southbridge hosts some truly sluggish performers, for even main memory is blazing fast compared to hard drives. Keeping with the office analogy, waiting for a hard drive seek is like leaving the building to roam the earth for one year and three months. This is why so many workloads are dominated by disk I/O and why database performance can drive off a cliff once the in-memory buffers are exhausted. It is also why plentiful RAM (for buffering) and fast hard drives are so important for overall system performance.

...

Finally, the diagram has various real-world throughputs for networking and other buses. Firewire is shown for reference but is not available natively in the Intel X48 chipset. It’s fun to think of the Internet as a computer bus. The latency to a fast website (say, google.com) is about 45ms, comparable to hard drive seek latency. In fact, while hard drives are 5 orders of magnitude removed from main memory, they’re in the same magnitude as the Internet. Residential bandwidth still lags behind that of sustained hard drive reads, but the ‘network is the computer’ in a pretty literal sense now. What happens when the Internet is faster than a hard drive?

High-Frequency Trading Fact of the Day

From an interesting paper on the flash crash by Easley, Mailoc Lopez de Prado, and O'Hara:

Since 2009 high-frequency trading firms, which represent approximately 2% of the nearly 20,000 trading firms operating in the U.S. markets, have accounted for over 73% of all U.S. equity trading volume.

Via Computational Legal Studies

Chicago Innovating in Technology Transfer

From the UChicagoTech blog:

One of the biggest challenges to early stage innovation is that it most needs experienced advice when it is least able to attract it, says Alan Thomas, director of UChicagoTech, the University of Chicago’s Office of Technology and Intellectual Property. “A structural problem in Chicago is that at this embryonic point there have been very few resources to turn to,” he says.

...

Start-up ventures are more likely to thrive when advised by multiple mentors with proven skills and experience. “Collectively we are stewarding discoveries flowing from between a billion and a billion and half dollars per year of basic research,” Thomas says.

Thomas and his team have a more clear-eyed view of the landscape than some of the more established tech transfer offices. The potential weak points I see in the plan are (1) that mentors won't have enough skin in the game to do more than provide very early and general advice; and related: (2) adverse selection resulting from more experienced inventors and mentors never opting into the program. Both might be alleviated if they can get a handful of high-profile anchor tenants signed up immediately. Still, it looks promising, and I look forward to hearing more as the plan is implemented.

Are Biological Systems Poised at Criticality?

From Thierry Mora and William Bialek:

The first question is why biological systems should be nearly critical. What benets does operation at this special point in parameters pace provide for these systems? For birds, we have seen that criticality confers high susceptibility to external perturbations, and this enhanced reactivity endows them with a better defense mechanim against predators. Similarly, in the auditory system, being close to a bifurcation point allows for arbitrarily high gains and accurate frequency selectivity in response weak sounds.

In neural populations, the naive idea underlying the theory of branching processes makes criticality seem almost inevitable, a middle point between death and epilepsy. However, the function of neural networks is not only to be reactive, but also to carry and process complex information in a collective manner through its patterns of activity. The observation and analysis of metastable states, both in retinal acitivity analyzed within the maximum entropy framework [20] and in the activity of cortical slices analyzed with the theory of branching processes [79], suggest that criticality may be coupled to the explosion of these states, allowing for a wider set of coding options.

Via the Physics arXiv blog.

Fiber to the Neuron?

arXiv blog provides a great summary of exciting new lines of research into how human neurons communicate with light:

In recent years, a growing body of evidence shows that photons play an important role in the basic functioning of cells. Most of this evidence comes from turning the lights off and counting the number of photons that cells produce. It turns out, much to many people's surprise, that many cells, perhaps even most, emit light as they work.

For me this ranks among the most exciting new discoveries in biology. We have incredibly sophisticated tools for manipulating and measuring these signals. What is science-fiction today (see, for example, Avatar), could be reality in our lifetime. In part because of the severe capital costs and regulatory hurdles to implementing any of these new technologies, however, I suspect that patents will play a key role in their commercialization.

Fixed Points in the Publishing Industry

From an interview of Lonely Planet's Gus Balbontin on O'Reilly Radar:

The essence of what the print people have known for 500 years is still the same in the digital world: you still have to publish great content. Yes, the tone may change a little. Yes, the experience might be slightly different. But at the end of the day, the people creating the content are humans. The people consuming it are also humans. The same humans that consume print consume digital. It's not different.

...

In terms of the next 10 or 20 years, the essence of what publishers are doing won't change at all. This is what should make publishers feel comfortable about the future. The reason why you're there and the problems that you can solve for customers won't change. That's what you need to keep focusing on.

Henry Manne on the Limitations of Behavioral Economics

Here:

The problem in traditional economics style (I think “style” a more appropriate word here than “methodology”) is a general failure to keep separate that part of economic assumptions which pertains to individual behavior and that which properly pertains to aggregate behavior. About the latter, BE has very little to say (here I am talking exclusively abut microeconomics, as I am really not sure how much relevance BE, aka psychology, has to macro areas), and Alchian simply carries the day.

Part of a Truth on the Market Symposium on the same.

The Coase Theorem Does Not Apply to Multilateral Agreements

In The Problem of Social Cost (available here), Ronald Coase presented the case for what became his eponymous Theorem. In short, Coase argued that in the absence of transactions costs, parties would contract around default legal rules to achieve an economically efficient allocation of resources. As a result, in the absence of transactions costs, lawmakers should only need to ensure a clear definition of property rights and the enforcement of contracts, without worrying too much about who starts off with the property rights. The Coase Theorem has been on my mind a bit lately thanks, in part, to the Supreme Court grant of certiorari in the Stanford v. Roche case. If transactions costs could be ignored, then (according to Coase) none of the stakeholders in the tech transfer system should worry too much about a reallocation of ownership rights by the Supreme Court. An economically efficient bargain would always be reachable under a new allocation of rights -- assuming the transactions costs can be ignored.

But can they? Of course not. The uncertainty associated with tech transfer investments is often so great as to swamp the costs of negotiating these agreements. In this post, however, I want to abstract away from the specific economics of tech transfer, mortgage finance, corporate chartering, insurance, derivatives, or any other economically multilateral agreement in order to emphasize just how limited the Coase Theorem is in its general applicability to the analysis of large-scale social costs.

The Problem of Social Costs makes its case using a hypothetical example of a negotiation between a rancher and a farmer over who should have to pay for a fence and/or crop damage. That archetypical case comes in the first main section of the paper, titled "The Reciprocal Nature of the Problem." Coase's hypothetical is explicitly a bilateral agreement. Coase's emphasis of the "reciprocal" nature of the problem is implicitly bilateral in its symmetry.

So what? Consider the graph below, which compares how the transactions costs of bilateral contracts and the transactions costs of trilateral contracts add up as a population of parties increases. The calculation of both is based on the assumption that each party to an agreement (whether bilateral or trilateral) will pay the same fee for entering into an agreement. The calculation of both is made assuming every party in the population will enter into a single agreement with every unique combination of counter-parties. (In detail, these are total unique combinations of two and three picks from a population of n graphed as a function of n.) The graph shows how the relative difference in transactions costs diverges dramatically even for a population of 15 parties.

The difference is even more dramatic for a population of 100 parties.

What does this mean? At the very least it means that in policy debates about the allocation or reallocation of resources whose use involve three or more distinct economic interests (i.e., stakeholders), the Coase Theorem should not be invoked, or at the very least invoked only after everybody is satisfied that the transactions costs are completely overwhelmed by the economic benefits of a bargain.

But nearly every policy question involves more than two stakeholders. Even consumer transactions -- which seem like the quintessential bilateral agreement -- have indirect effects, compensated for through indemnifications and warranties. In other words, the Coase Theorem simply doesn't apply to any practical policy problem of interest.

At the same time, the Coase Theorem was very useful because it focusses our attention on transactions costs. In many cases, it may be easier to lower transactions costs than to decide how rights should be allocated. My point in this post is not to say that The Problem of Social Cost was fundamentally wrong, or even useless. Rather, I want to point out that we can usefully analyze transactions costs by breaking them out according to the number of parties necessary to an agreement.

This point gains strength as the number and variety of multilateral agreements increases. Nothing much at all seems to happen in our economy without getting buy-in from a host of stakeholders, either formally or informally. And the biggest economic issues of our day -- from Fed policy, to mortgage financing, to tech transfer models -- are all examples of vast, multilateral negotiations. Our world is no longer Coasean.

Positional goods present an especially interesting application of this analysis of transactions by number of parties. A positional good, by definition, is a good whose value depends both on the supply available and on the demand for the good by other consumers. Every transaction involving a positional good is at least a trilateral agreement, and arguably much more multilateral than that since the other consumer might not be based on the activity of a particular person, but rather on a general perception of the activity of an entire group of other consumers. The Coase Theorem simply doesn't apply to positional goods.

Square Cofounder an Omitted Coinventor of Patent Apps Owned by Wash. U. Professor?

Techcrunch reports:

In or about February 2009, Mr. McKelvey contacted his friend of 20 years, Mr. Robert Morley, Jr., to help build a prototype magnetic card reader that could be plugged into a cell phone input jack. Mr. Morley, a professor at Washington University, had previous experience working with magnetic card readers.

Over a period of several days, working in Mr. McKelvey’s glass studio and Mr. Morley’s laboratory at Washington University, Mr. McKelvey and Mr. Morley exchanged ideas regarding the design of the prototype card reader.

Square is one of the hottest startups in Silicon Valley, backed by Khosla Ventures GP (and former Facebook and YouTube CFO) Gideon Yu, and cofounded by Twitter cofounder Jack Dorsey. A recent hire was Keith Rabois, an early PayPal exec who had a successful exit at Slide most recently. Keith was also an IP litigator back in the day.

Based only on the allegations in the complaint, it looks to me like Square has a potentially viable claim for co-inventorship and co-ownership here. I got curious to see just how deep the rabbit goes by looking up to see whether Morley had any federal funding for his research at Wash U. His CV doesn't show any recent federal funding, which is lucky for him, since whatever he might get from Square in resolving this dispute could escheat to Wash. U. depending on how Stanford v. Roche gets decided.