A new era in scientific discovery is upon us as humans are augmented by robotic computerized assistants to help them sort through the floods of data being created by modern scientific instruments. The increased compute power of the modern microprocessor teamed with breakthroughs in machine learning are the underlying foundations for this revolution.
The scientific transformation fueled by robotic/machine discovery will occur across a number of scientific drivers with life sciences leading the way. Life sciences will push the boundaries due to the explosion of new data in these domains. Take for example the gene-expression microarrays which make it possible to measure the rate at which cells expression the proteins for thousand of genes. The problem as noted by Michael Molla is "the amount of data this new technology produces is more than one can manually analyze."[1]
A great example of the potential for robotic discovery can be seen in a recent experiment done at Manchester University. Using advanced machine learning algorithms scientists were able to train a computer to look for new biological markers for diagnosing Alzheimer's disease. The computerized assistant was able to scan many more samples than a human assistant would have been able to and importantly was able to "learn" from each new marker found.[2]
Observational science will also be changed as breakthroughs in computer vision combined with automated observation allow scientist to have 24*7*365 surveillance of samples. For example, the Smart Vivarium project at the University of California San Diego, uses machine leaning and cameras to allow for the robotic monitoring of all of the mice used in animal experiments at UCSD. As Tony Yaskh at UCSD noted, "It allows you to look at subtle changes in behavior that we know are occurring but cannot study because of the labor intensive nature of long-term monitoring."[3]
Impacts:
The automation of scientific discovery through robotic data collection and analysis is necessity given the explosion of data being produced by modern science. The practical impact of this change will be to allow us to enhance the efficiency of scientific discovery.
However, this automation, will come at a cost which will be to decrease the flexibility of modern scientific experiments. Put simply, the "noise" eliminated by many robots is exactly the type of input that in the hands of a human observer may lead to a completely serendipitous discovery. [4]
The importance of serendipity to scientific breakthrough should not be underestimated. As Steve Shapin explains in his review of the book the Accidental Scientists, "Many scientists, including the Harvard physiologist Walter Cannon and,later, the British immunologist Peter Medawar, liked to emphasize how much of scientific discovery was unplanned and even accidental. One of Cannon's favorite examples of such serendipity is Luigi Galvani's observation of the twitching of dissected frogs' legs, hanging from a copper wire, when they accidentally touched an iron railing, leading to the discovery of "galvanism"; another is Hans Christian Ørsted's discovery of electromagnetism when he unintentionally brought a current-carrying wire parallel to a magnetic needle."[5]
Thus, automation will allow discovery when we know what we are looking for but it will likely decrease the serendipity that has fueled some of our most important breakthroughs.
Sources:
[1]"Using Machine Learning to Design and Interpret Gene-Expression Microarrays",Michael Molla, et al, AI Magazine, Volume 25, 2004, pgs 23-44.
[2]"Mark of Time", The Engineer, September 18, 2006
[3]"Big Brother Keeps an Eye on Lab Animals", New Scientist, Feb 28, 2004.
[4]"Humans Vs Robots: Who's Best in Space", WhyFiles, http://whyfiles.org/171manned_space/3.html
[5]"The Accidental Scientist", Steve Shapin, American Scientist Online,
http://www.americanscientist.org/template/BookReviewTypeDetail/assetid/34011
3d virtual reality adoption in the consumer marketplace, particularly the home, has been limited by three primary factors. First, the technology needed to display 3d virtual reality has been expensive and not integrated into consumer electronic devices. Second, the interfaces to achieve 3d have been cumbersome in specific those associated with active stereo display. Third, the bandwidth necessary to move highly accurate models has been limited to the home.These characteristics coupled with little "mass" exposure to 3d virtual reality immersion have effectively killed the home from being a virtual reality marketplace.
