Author: jlolds

Native Angeleno, professor of Neuroscience and Public Policy at George Mason University

Smarter or Dumber?

You’re recall Nicholas Carr’s very controversial article in the Atlantic over the summer that suggested that on-line searching was making us stupid. Today, CNN.com reports on a study which purports to show the opposite! Dr. Gary Small used fMRI showed that:

Members of the technologically advanced group had more than twice the neural activation than their less experienced counterparts while searching online. Activity occurred in the region of the brain that controls decision-making and complex reasoning

Meanwhile cross town rival…

Liz Zelinski, a professor of gerontology and psychology at the University of Southern California, said the findings about the brain activity differences aren’t surprising and offered this analogy: “If you wanted to study how hard people can exercise, and you take people that already exercise and people that don’t exercise, aren’t they going to be different to start out?”

Don’t you think fMRI is getting a bit oversold these days?

Jim

Another Nobel Prize for MBL in Woods Hole

This time for the discovery of Green Fluorescent Protein or GFP–used now ubiquitously in all fields of biosciences:

Japan’s Osamu Shimomura and Americans Martin Chalfie and Roger Tsien shared the prize for their research on green fluorescent protein, or GFP, the Royal Swedish Academy of Sciences said.

And where does Shimomura work?

Shimomura, born in 1928, works at the Marine Biological Laboratory in Woods Hole, Massachusetts, and the Boston University Medical School.

Congrats to Director Gary Borisy and all the good folks at the MBL!

Jim

Financial Times on Machine Learning

Alane Cane has written a spectacular piece in today’s FT regarding the limitations of AI and cognitive computing. Featured in the article is IBM’s Dharmendra Modha:

IBM was a pioneer in the field and today continues to invest heavily in AI research. Dharmendra Modha, a scientist in the company’s California research laboratory is working on cognitive computing, which he defines as a computer model that simultaneously exhibits characteristics seated in the human brain, including perception and emotion.

His aim is to discover how the brain works, not how the mind works, he is quick to emphasise.

Last year, his group achieved a milestone by managing to simulate the operation of a mouse brain on an IBM Blue Gene supercomputer.

He notes: “We deployed the simulator on a 4096 processor Blue Gene/L supercomputer with 256 megabytes of memory per processor.

We were able to represent 8m neurons and 6,300 synapses (connections) per neuron in the one terabyte main memory of the system.”

There will be, of course, a considerable time lag before the benefits of this research are seen in actual products.

Mr Modha thinks it could be 10 years before cognitive computing of the kind he is working on makes its debut in productivity and security systems. It is, however, a giant leap from 1956 when an IBM supercomputer of the day simulated the firing of a mere 512 neurons

Read the whole article!

Jim

Neuroethics and fMRI

Here’s a sociologist worrying that we’ll be classifying kids on the basis of their fMRI:

At the annual meeting of the American Sociological Association in 1968, the social activist Martin Nicolaus leveled a blistering critique at traditional criminologists and sociologists. He said — and I am paraphrasing here because it was a panel discussion, not a paper — you people have your eyes down and your hands up, while you should have your eyes up and your hands down.

He was speaking language that social scientists of the era would have easily understood. “Eyes down” meant that almost all the research on deviance and crime was focused on the poor and their behavior, while “hands up” meant that the support for such research was coming from the rich and powerful — from foundations, the government, and corporations. Conversely, of course, “eyes up” meant turning one’s research focus to the study of the pathological behavior of the elite and privileged, and “hands down” meant giving more of a helping hand to the excluded, impoverished, and disenfranchised. What was true of the social scientists’ almost monomaniacal attention to the deviance of the poor in the 1960s is even truer of the neuroscientists of the last decade’s vanguard research.

A common tactic is to do brain scans on people who have already displayed violent or aggressive behavior. While this will net the adolescent boys in juvenile correctional facilities, it will surely miss the spawning white-collar bandits in energy markets or subprime home loans.

This is from the Chronicle of Higher Education. You can see where this is going can’t you? Let’s brain scan Hank Paulson. Somehow I don’t think that’s going to happen.

Jim

Goldilocks on Wall Street

Greetings to all. I am the new Chair of the Dep’t. of Computational Social Science at the Krasnow Institute for Advanced Study at George Mason. Jim has graciously invited me to contribute to his blog. I am an economist with broad interests in economic theory and policy, behavioral economics (how people depart from perfectly rational behavior), cognition (thus my Department’s natural connection to Krasnow), and computation (my Ph.D. is from Carnegie Mellon).

Under normal circumstances I would comment here on matters of basic research and science policy. But the current financial environment is sufficiently grave that I will take the liberty of discussing the proposed bailout of Wall Street.

For the last year or so banks have been sufficiently worried about the risks posed by the mortgage-related assets of their peers that most interbank lending has become expensive (as measured by the so-called TED spread, for instance). Indeed, in the past couple of weeks such interbank activity has nearly dried up—this is the ‘credit crunch.’ Such short term lending is an important activity in the day-to-day operations of major banks, and its demise has led to the Fed providing short term credit facilities to keep overnight and other short term lending and borrowing going. But what we don’t really know (yet) is just how bad bank balance sheets look. It is reasonable to assume that many are weak—thus the urgency with which bailout legislation has been crafted and voted on—but just how weak is not known. The big problem, as I see it, is that bank solvency is not really knowable ex ante, i.e., before a rescue plan is implemented. Here’s why:

Economists are wont to focus on prices to explain the inner workings of the economy. An important part of proposed bailout/rescue plans is use of certain auction mechanisms in order to ‘discover’ the proper prices for ‘toxic’ mortgage-backed securities and related financial assets that troubled firms wish to take off their balance sheets. The functioning of these auction mechanisms is critical to the performance of the bailout, but the exact nature of these mechanisms is not yet specified. Indeed, there is a ‘Goldilocks’ flavor to how such auctions have to function in order to relieve the credit crunch. Currently there is no market for the most toxic of these assets—they are illiquid and their value may be very close to $0. If a government auction for them produces very low prices for such assets then the troubled banks selling them will benefit little from such sales and may be made insolvent. If these auctions produce prices that are too high then taxpayers will be on the hook in a big way and banks will get recapitalized from a true government bail out—welfare for Wall Street even though these same institutions have earned staggering profits in recent years. Instead of auction prices that are either too low or too high it is hoped that the Treasury auctions find prices right in the middle, sufficient to keep the banks operating and thus restoring confidence so interbank lending can resume at reasonable rates. Thus Goldilocks.

The trouble with this whole approach to finding the ‘right’ price for a particular mortgage-backed security (MBS)/collateralized debt obligation (CDO) is that the real determinant of the value of the underlying mortgage assets depends on the health of the overall economy. If the near term recession is mild or even non-existent then stable home prices would result, meaning higher auction prices for distressed mortgage-related securities would be in order. If a longer, deeper recession is in the cards then lower auction prices for MBS/CDOs are justified, even though this would likely ‘take out’ several large banks. In the case of a big recession/depression then the value of many mortgage-related assets is probably close to $0 and many banks will be insolvent.

If one looks at the economy in a static equilibrium way, as is conventional in economics, then one might say there are multiple equilibria in the picture I have laid out: low housing prices and insolvent banks represent one equilibrium and higher housing prices and stabilized institutions are another. Unfortunately, unless one has a reasonably accurate model telling you the effect of a bailout on the real economy, it is going to be guesswork as to which kind of bailout plan leads to which equilibrium. In an economy with multiple equilibria, a specific bailout plan, in essence, selects the final outcome for the economy by picking prices paid for ‘toxic’ securities. Prices that are too low might in advertently ‘kick’ the economy down to a deep recession or depression equilibrium.

The real problem, it seems to me, is that we have only the crudest understanding—zeroth-order models and gray beard guesses—for the overall effect of alternative bailout plans on the real economy. The great 20th Century Economics Nobelist Herbert Simon often argued that we, as a nation, need much more support for basic research in the social sciences. Imagine the ‘return on investment’ today from a few million dollar research program that had created a reasonably-accurate, quantitative model of the current financial system? Maybe it would be saving us a few trillion dollars right now.

Reverse engineering the brain: not to be taken literally

There’s been some angst among fellow scientists about the use of the above term to describe the current attempts to use experimentation and computational neuroscience together particularly in the context of robotics. The worry is that if we can’t reverse engineer the worm C. elegans or the sea snail, Aplysia–which we can’t at present–then we certainly aren’t in a position to reverse engineer the aspects of higher cognition (e.g. spatial navigation) that are interesting about mammalian brains–especially our own.

I agree with the worriers in that, the prospects of building a robot with fully human-like capabilities is not in the near future. Where I disagree is with the notion of setting a scientific agenda that uses the principles of reverse engineering to build up (incrementally) an understanding of how aspects of higher cognition emerge from neuronal activity. An example of very productive work in this field can be seen in the research coming out of Gerry Edelman’s Neuroscience Institute in La Jolla where computational neuroscience models interact with the real noisy environment. Jeff Krichmar’s work there (he’s now at UC Irvine) is an excellent example of this approach.
And so, when the idea of reverse engineering the brain comes up, from my perspective it is as a metaphor. As far as actual reverse engineering of brains, this will happen in the worm or the fly long before it becomes an actual possibility for vertebrate brains. But the term is a useful metaphor for an approach, whereby experimentally determined characteristics of real brains can be applied to real life engineering problems–like creating better autonomous vehicles.
Jim

Meeting Daniel Schorr somewhat by accident

I had signed up to attend Daniel Schorr’s lecture yesterday evening at the Cosmos Club. There was a reception and in typical fashion, my wife and I picked a table in the very back of the Powell lounge to enjoy some cheese and wine before the lecture. A few minutes after sitting down, an elderly gentleman sat down at our table and introduced himself as Dan Schorr. And so we began to talk about the current financial crisis, foreign policy and Mr. Schorr’s NPR reports. He is really a remarkable journalist–of the old Edward R. Murrow school.

Only in Washington.

Jim

I wonder if this is true with Amazon’s Kindle?

Regarding reading words from your computer screen, from the Chronicle of Higher Education….Money quote:

people took in hundreds of pages “in a pattern that’s very different from what you learned in school.” It looks like a capital letter F. At the top, users read all the way across, but as they proceed their descent quickens and horizontal sight contracts, with a slowdown around the middle of the page. Near the bottom, eyes move almost vertically, the lower-right corner of the page largely ignored. It happens quickly, too. “F for fast,” Nielsen wrote in a column. “That’s how users read your precious content.”