Juan Lazaro IV/IRRI Photo Bank

A comparison of different rice types.

From the NSF, a story on a federally-supported project aimed at harnessing the power of distributed computing to alleviate hunger.

The project, dubbed Nutritious Rice for the World, runs out of the University of Washington, but pieces of the research could be unfolding on a desktop near you. That’s because the research is one of five projects currently part of IBM’s World Community Grid. The grid allows volunteer computer users to run a small program that takes advantage of unused processing power to predict the structure of desirable rice proteins.

Understanding the protein structures for the major strains of rice will allow researchers to help farmers cultivate healthier crops, feeding more people and earning growers more income from their harvests. According to the Food and Agricultural Organization of the United Nations, rice is the primary staple food for more than half the people on the planet and accounts for 20 percent of the total caloric intake of everyone in the world.

Ram Samudrala, associate professor and computational biologist, is the project leader at UW, and explains that plants better suited to local environments mean that food doesn’t have to move across large distances before it get to the people need it:

“The fundamental problem with food shortages in the world is one of distribution,” Samudrala says. “Creating distribution chains costs money. We overcome that by designing new crop species that indirectly address this problem by providing higher yields but also better nutrition and adaptability to local and global environments.”

This is doubly significant in the wake of this year’s food price spikes and the resulting crisis. One of the catalysts of the price increases was the skyrocketing cost of transportation fuel.

Press coverage of the project has been significant since the launch in May of this year. Got some unused CPU power under the hood of your PC? Anyone can register for the grid.

SOURCE: Complete Genomics

Image of a plate of DNA sequence markers.

Earlier this week, Complete Genomics announced that it will offer complete human genome sequencing for the low, low price of $5000. To put this in perspective, Technology Review explains that “with a price tag of $100,000 to $1 million over the past two years, only a handful of human genomes have been sequenced to date.”

Technology Review also offers a detailed description of the sequencing technology, along with a helpful slideshow. The graphical explanation illustrates the bridges between cytogenetics, nanotechnology, and computer science that enable the breakneck pace of this sequencing technology.

The new technique also emphasizes the fact that Complete Genomics is set up not as an instrument company; rather, the company is set up as an information service provider:

Beyond its unique technology, Complete Genomics has also chosen an unusual business model: rather than selling instruments, as most sequencing companies have done, it plans to offer sequencing services through a commercial-scale genome center….The company is now building a massive data center to manage the immense volume of information it expects to generate; it’s planning to have a computer cluster containing 60,000 processors online by 2010.

And unlike sequencing companies like 23andMe and Navigenics, which market their services to consumers (and do not offer complete sequencing), Complete Genomics will not be encouraging Spit Parties.

Over at Genetic Future, there’s a provocative analysis of the near-term future of how companies in this market will likely evolve:

If Complete Genomics does indeed have the edge over its next-generation competitors, you can guarantee it won’t last long – all of the platforms are constantly being tweaked and improved, and there are new competitors on the way…. By focusing on service provision using a single, cutting-edge technology, the company may present an attractive option for pharmaceutical companies and other corporations looking to outsource their sequencing needs.

But here’s the kicker, as Genetic Future again points out—the likely result here is for profits to flow to companies that can offer the best interpretation of genetic information, not just the fastest and cheapest sequencing:

There’s an important message here between the lines: as technology drives the price of sequencing down, massive competition between platforms and service providers will almost certainly drive down the profit margins of sequencing providers. The real money will then be in providing sophisticated, up-to-date and easily understandable genome interpretation services. The best interpretations will come from the companies with the largest databases of genetic information, and with expertise in putting that complex information in the appropriate context for lay consumers. It seems to me that right now deCODEme has an advantage in the former race (given the formidable genomic data-sets assembled by its Icelandic parent company deCODE genetics), while 23andMe is winning the latter – but if 23andMe succeeds in attracting a wave of customers (and particularly disease patients) with its new low low prices, they may well ultimately gain the edge on both counts.

The historical parallel that springs to mind here is the bet Microsoft made in the late 70s to focus on software rather than hardware. The path to blockbuster success in the computer world didn’t turn out to be making the speediest desktop or the biggest hard drive; hardware costs for those products are still dropping after decades. The money was in the tools that helped people manipulate, interpret, and share information.

The not entirely unintuitive fact that the sweet spot for profits in genomics will be in interpretation services emphasizes the challenge for public policy that balances consumer protections with advancing cures. On the one hand, better genome studies will allow researchers to further understand the tangled relationship between genes, environment, disease, and disability. But as regulators have been sluggish to do their genetic due diligence on what Art Caplan calls “spitomics” companies, you can probably bet that spurious claims from companies offering misleading interpretations to consumers will multiply in the future.

As a climate modeler, you are always working with the best of what’s available—whether that means the best data, best infrastructure, or best science.

To be fair, “denigrate” might be a little too strong of a word to use to characterize the often-legitimate criticism that has come climate modeling’s way. The critics’ main point of contention? Quite simply that models cannot—and likely never will—accurately represent the whole climate picture. There are simply too many known unknowns and unknown unknowns—pardon the reference—for even the most skilled modeler to wrap his head around. On a more basic level, does anybody really think that a collection of models, let alone a single model, can fully reproduce Earth’s complex inner workings? No, of course not, and that’s a point any climate modeler will readily concede.

As a climate modeler, you are always working with the best of what’s available—whether that means the best data, best infrastructure, or best science. And since all those variables are subject to frequent revision, it’s rare to find a robust model that is able to withstand years of new findings. Scientists often relish poking holes in them, using the results from a recent research expedition, for instance, to undermine a single component—regardless of how well the model otherwise captures the environment. While some of this criticism may seem gratuitous, or even childish at times, it is often done with good reason.

Take a recent study published in the Proceedings of the National Academy of Sciences, whose findings risk invalidating over 60 percent of the so-called “climate envelope” studies. Climate envelope models help predict where species will live under conditions of future climate change by using their current distributions to make up a set of climatic conditions—the “envelope”—that closely approximate their needs.

In the past, these models have come under withering criticism for failing to take into account a number of other factors, such as anthropogenic activity or species-species interactions, that figure as prominently, if not more so, as climate change in influencing species distribution. Despite some of their limitations, the Intergovernmental Panel on Climate Change put its stamp of approval on their use in its 2007 report, noting that they “offer the advantage of assessing climate change impacts on biodiversity quantitatively.” Colin Beale of the United Kingdom’s Macaulay Institute of Land Use Research, the PNAS study’s lead author, found that the models performed no better than a simple roll of the die—pure chance—in approximating several bird species’ natural habitats.

To be clear, Beale’s study would not be the first to take such a dismissive view of climate envelope studies—many scientists argue that they still play an important role in predicting future species abundances, if not their exact distributions—and should therefore be taken with a grain of salt. In other words, the pretext of the study is not to invalidate the findings of the IPCC or to cast doubt on the link between climate change and species. Beale is quick to point out that his study did find a significant relationship between the climate and a third of species—and that he is concerned his findings could be misused as evidence that there is no link between climate change and species extinctions.

Other scientists have been critical of the IPCC for seemingly lending too much credence to models’ predictive abilities. While several existing models, especially the so-called “coupled” models (which consider the atmosphere, oceans, land surface, sea ice, and other physical characteristics in conjunction to project future climate trends), have become advanced enough to yield valuable insights on current and past climate patterns—almost matching the accuracy of conventional atmospheric observations—most fall woefully short when it comes to answering the most important question: What will our future climate look like?

One big problem, some argue, is that many current models suffer from oft-debilitating inconsistencies—in their representations of observed changes in global mean surface temperature or in their range of sensitivities, for example—that could significantly diminish their capability to reduce uncertainties in Earth’s climate dynamics and, thus, to predict future changes. As a result, they suggest that international organizations like the IPCC, which have a lot of clout in the scientific and political communities, tamp down some of their expectations—lest they invest too much credibility in models that could very well turn out to be wrong.

If there’s one issue on which most scientists—modelers included—agree, it’s that climate modelers need more: more research funding, more powerful computers to run their “petascale” models (which can make a whopping 1,000,000,000,000,000 calculations per second), and more time to meet ever-rising expectations. With the IPCC shifting its focus to examining the community and state-level impacts of various climate change scenarios (so as to impart more actionable information for policymakers) for its 2013 report, the pressure is on climate modelers to redouble their efforts to come up with more powerful and accurate global models. The bottom line is that we should by no means abandon modeling, but do need to help improve it.

Already premier research institutions like the Breckenridge, Colorado-based National Center for Atmospheric Research, the originator of one of the most widely used coupled models in the United States, is falling behind on meeting an October 1 deadline to update it. Even legislators who tend to only focus on the short-term—in other words, most of them—should see the wisdom in supporting work that could also lead to better hurricane research, an outcome that would yield immediate, and very tangible, benefits.

Perhaps Science Progress contributing editor Chris Mooney put it best in framing his argument for more research funding by tying together hurricanes, climate change and what he calls the “next” New Orleans:

“Which inevitably brings us to contemplating the future—one in which we will be even more exposed to hurricane risks. While it remains hard to predict precisely what global warming will do to hurricanes, we know that it will raise sea levels, and probably intensify storms on average, not to mention increasing their rainfall rates. No matter how you slice it, then, global warming worsens hurricanes—and, accordingly, hurricane-related insurance costs—which makes it a more-than-legitimate topic to invoke in the context of this year’s hurricane threats and landfalls.”

The impacts of climate change are predictable, but the task will be difficult without the additional resources to build more accurate models and adapt to an altered planet.

Jeremy Jacquot is a graduate student in marine environmental biology at the University of Southern California and is a contributing writer for VentureBeat, DeSmogBlog and TreeHugger.

Already, doing scientific research or engineering without computers is inconceivable. (Would you design a computer chip without a computer, or write a scientific paper without using a word processor, email, and Google?) However, humans currently do the majority of all thinking, with computers mostly “providing support.” This will inevitably change, as computing power allows machines to work more creatively and autonomously.

At some point it will be as hard to think without a computer as it is to build a building without physical tools.

It is hard to conceive of the magnitude of the disruption this will cause. The industrial revolution caused large-scale displacement as human physical labor was replaced or augmented by machine labor. Vast numbers of jobs were destroyed, and only later replaced with new, different, jobs. Agricultural and manufacturing jobs were replaced with white collar and service jobs. The coming informational revolution will be even more dramatic. It will happen far more rapidly than the industrial revolution, and will displace human intellectual labor, rather than human physical labor.

Given the magnitude of the potential disruption of the informational revolution, it is surprising that it has not generated more attention. It will have far more impact than global warming, and will occur far more rapidly. It will affect far more people than terrorism. In part, the lack of attention to the informational revolution is due to the difficulty visualizing exponential growth. With computers doubling in power every one or two years (depending on whose analysis you believe), in 20 years computers will be a thousand to a million times faster than they are now.

A current iPhone, with 16 gigabytes of memory, is superior to a desktop Mac of 10 years ago, and is far faster than a $10 million Cray supercomputer of the 1970s. More importantly, it has access to the web: billions of web pages and a rapidly increasing fraction of scientific publications, all indexed on a million-CPU distributed computer—a machine with a million billion bytes of memory capable of executing a million billion operations per second. A computer like the one Google uses for search is roughly similar to the human brain in speed and capacity, but is not (yet?) close in software sophistication. The question is what a computer can do in two decades when it is thousands of times larger and faster than a human brain.

Of course, raw computing power alone is not enough to make computers smart; they also need much more sophisticated software. In some regards, software has become much more powerful in recent years. In a few hours, a good programmer can write a program that checks the availability of tables at local restaurants, displays the results on map, and then sends the information to your cell phone. Not long ago, such a complex program would have taken years to write. But in other ways, software is still primitive. There are no good generic tools for building models of how people make decisions, and then executing the complex decision-making processes we go through everyday. Such tools will come.

How will we recognize that the informational revolution is upon us? The first manifestation will probably be the widespread use of virtual assistants. To some degree, this is already happening. Secretaries have been largely replaced by word processors and calendaring systems. Travel agents have been replaced by Web systems. In the coming decade, human-computer interfaces will become mostly voice driven, rather than based on typing. (This article was dictated to Dragon systems speech recognition software, but still required major editing.)

Virtual assistants will shift from helping retrieve information (think of Google) to making suggestions and decisions.

Your computer can already suggest to you a book or a restaurant you might like. But with advances in software sophistication and raw computing power, it will suggest more complex ideas like a whole paragraph to put in your letter. It will help manage your calendar, screen candidates and draft e-mails, evaluate ideas, and draft proposals. But the most disruptive advances will happen when computers outpace us at basic problem-solving tasks, eliminating the need to spend time and cognitive effort on what we now consider complex tasks.

In their more advanced forms, computers could apprehend much broader sets of information and its consequences, offering an approach to solving a problem. They could construct a lucrative business deal or design a novel research methodology. At some point it will be as hard to think without a computer as it is to build a building without physical tools.

Given the nature of exponential growth, the speed with which your computer will go from being a dumb device, to being an idiot savant, to being smarter than you will come as a shock. We should start thinking about it now.

Lyle Ungar is an associate professor of Computer and Information Science at the University of Pennsylvania.

Conference attendees also called for funds to keep top computer programmers in climate modeling, as many are finding it more difficult to resist the financial rewards and job security that companies like Google can provide. Jeffrey Sachs, Director of the Earth Institute at Columbia University, told summit attendees that if climatologists hope to secure federal funding, they must be able to answer key questions about climate change that politicians would find useful in policy making. As conference chair Jagadish Shukla put it, “If we just ask for enhanced understanding, then we have very little chance of getting the necessary funding.”

Chris Mooney recently explained how Cyclone Nargis demonstrated the need to understand and predict meteorological phenomena around the world—better modeling can protect settlements and save lives. Policy makers can provide scientists with the necessary tools.

Jonah Lehrer at the Frontal Cortex argues that we need more science critics and an open public atmosphere for critiquing science. His suggestion to science bloggers: Don’t post anonymously.

Eric Berger over at SciGuy discovered that the Federal government spends 200 times more on bioterrorism preparedness than on hurricane research. This discrepancy is even more significant, he suggests, because bioterrorism might happen while hurricane disasters will happen.

Jacob Goldstein at the Wall Street Journal Health blog covers several stories on the growing number of parents refusing to vaccinate their children over fears that the injections may be linked to autism or neurological disorders, despite the fact that no solid evidence exists suggesting vaccines pose any such danger.

The Chronicle’s Wired Campus covered the news that Intel and Microsoft have teamed to open research centers at top universities to enlist them in a new initiative to harness the power of parallel computing for the next generation of computing systems. It’s worth noting the long-haul five-year commitment to the research.

The basic problem in life science modeling runs like this: researchers working in the life sciences conduct their research with an exponentially increasing amount of data, to the point that biology and environmental science work can start to look like computer science.

For example, ecologists studying watersheds need to have data on water flow, temperature, and acidity; rainfall; the shape of streambeds; the levels of pollutants in the water—and they need to gather, store, and manipulate data gathered at different times over long periods at geographically dispersed points by different groups using potentially different equipment. It’s a tall order, but with an appropriate set of models, a scientist can predict the effect of various land and water management decisions, isolate the causes of species loss, and determine the source and levels of pollutants—all of which directly impacts major policy choices for regions across the country.

The solution to the problem? Collaborations between computer scientists and life science researchers. The specialists adept at manipulating databases, designing sensors, and translating information between formats can build the scientific workflows and modeling systems that allow the life scientists to comprehend the complex natural systems. Several such eco-modeling demonstration projects were on display at TechFest. The “E-Science: Science in the Cloud” booth demonstrated watershed monitoring projects. The “Trident” workbench provided a set of research tools for oceanography. And the “Science for the 21st Century” booth offered an array of resources that allow non-programmers to harness database technology for studying climate models, forest ecology, and genetic material.

These types of collaborations between information technology experts and biologists fueled the race to sequence the human genome and drive current work in DNA mapping, personal gene sequencing, and synthetic biology. But in order to continue advances in the field of bioinformatics and fuel more such collaborations in biological and environmental sciences, there have to be enough computer scientists to work on the collaborative research.

But new numbers indicate that enrollment in computer science programs in 2007 was half what is was in 2000. The numbers seem to be stabilizing, with an increase between 2006 and 2007, but the issue may be relevant when Microsoft Chairman Bill Gates testifies before the House Science and Technology Subcommittee on Technology and Innovation next Wednesday.