We are all caught up in the quickening whirl of technological change. As consumers we recognize that accelerating technology has had many benefits, such as quicker access to information, goods, and family and friends. Yet this change can also create many new social problems—from an aging population to employees displaced by machines to new kinds of weapons of mass destruction. As a result, one of our most urgent tasks is to adapt governance to the speed of our age.
This is why President Obama’s BRAIN Initiative (Brain Research through Advancing Innovative Neurotechnologies) is a welcome and worthy program. It addresses three problems that were either created by technological advances or are a side effect of them. First, it is attempting to speed research into neurodegenerative diseases like Alzheimer’s and Parkinson’s that afflict the growing ranks of the elderly. Second, it will accelerate progress in artificial intelligence (AI) and so bolster the security of the United States by allowing us to wield ever-smarter weapons against new kinds of terrorism. Third, boosting AI can help society manage emerging threats from all accelerating technologies—including AI itself.
The brain is the most complex organ known in the universe. Understanding the brain is thus far more daunting a task than putting a man on the moon or sequencing the genome—the most famous 20th-century government science initiatives to which BRAIN can be compared. But the initiative is also a more important political program than these predecessors because it will ultimately help social policy keep pace with the dizzying rate of technological change.
Of course, research into the brain would be proceeding with or without additional government support. But BRAIN’s additional $200 million in public and private funds will create new tools for research and better coordination among researchers. Through advanced imaging, scientists can already get a general picture of the brain. A recent example is Brainbow, which uses florescent proteins to show the position of different neurons. Nevertheless, we currently lack mechanisms to represent more dynamically hundreds of thousands of neurons firing in rapid sequence. Such tools would help us understand more fully how patterns of neuron activity mediate our representations of the outside world, thereby showing how physical processes translate to intelligent activity and emotional affects.
More precise mapping of our mental mechanisms is a step toward resolving our growing “Tithonus problem.” In Greek myth, the demigoddess of the dawn sought immortality for her mortal lover Tithonus, but forgot to ask the greater gods to give him eternal youth. Tithonus then became progressively more decrepit until the gods took mercy and turned him into a grasshopper. We face a similar dilemma. Life expectancy climbs by two months every year, but devastating neurodegenerative diseases are disabling more people as we live longer. These afflictions do not kill quickly, but they rob the aged of their ability to flourish independently, and they drain billions of dollars a year in Medicare and other medical costs.
Many drug companies have aimed to find cures for such diseases. But economic theory supports a role for government as well. Private companies cannot capture all the benefits of basic science for themselves, and so as a result such science is underfunded. Some basic research, such as the discovery of natural processes, cannot be patented at all, as the Supreme Court reminded us this last term in the Myriad case. Companies do not even capture all the value of patentable research, because the consumer surplus of innovative products is so enormous. For instance, individuals have benefitted far more from the computer revolution than is reflected in the sales and profits of the companies associated with it.
Fortunately, funding basic research does not require the government to decide which companies to subsidize. Republicans are right to worry about the risk of corruption in mixing government with industry, or the inefficiencies caused when it tries to pick “winners and losers”, but those concerns don’t supply any reasons not to fund the BRAIN initiative, which largely supports non-profit inquiry.
The case for the BRAIN initiative would be strong if the only payoff were public health. But as great as those potential benefits are they likely pale in comparison to the benefits to national security and good government. Learning more about the brain expedites the development of AI, which is essential both to the defending the nation and navigating the rapids of accelerating technological change.
AI is already improving. Its most recent public triumph came in the form of Watson—the machine that roundly beat two of the best Jeopardy champions of all time. This victory required the computer to disentangle humor, recognize puns, and resolve the inherent ambiguities of natural language. Watson represents substantial progress over Deep Blue, the machine that beat world chess champion Garry Kasparov in 1997. Because of its multiple, sophisticated algorithms and vast store of relatively unstructured data, this program succeeded in a less precisely rule-governed world that much more closely resembles the chaotic one we inhabit. Spinoffs from Watson will soon perform far more important tasks than winning at game shows, such as aiding in medical diagnosis.
Nevertheless, there is room for much more improvement in AI. The BRAIN initiative promises further progress by allowing machine intelligence to take advantage of new discoveries about the functioning of the human brain.. Artificial intelligence should become more natural, whenever doing so improves performance.
Of course, it is impossible to predict all the synergies created by melding the insights of neuroscience to computer-based technology. But already neuroscience researchers, like Georgia Institute of Technology Professor Steven Potter, observe that humans deploy lots of sophisticated sensors that help the brain gather information from the world. Research also shows that the brain has different feedback loops to improve its output. Further study of these processes will refine the neural nets that AI is already using to improve learning and pattern recognition.
More research will also likely illuminate the most central difference between machine intelligence and the natural kind. As Professor Potter notes, machine intelligence is as a rule indifferent to the physical substrate on which it actuates: the logic of Boolean algebra can be programmed on a variety of hardware. But the brain does not appear to have this neat separation. The properties of cells seem bound up in the output of natural intelligence. Understanding how this process works may in the long run allow for more efficient artificial intelligence by creating physical substrates conducive to faster processing.
More powerful AI sustains a more potent national defense. From ancient times nations have always innovated to gain military advantage. Today smart weapons are the cutting edge of military competition. Advances in AI can make weapons even smarter. It is not surprising therefore that the Defense Advanced Research Projects Agency is receiving almost half of the new federal money from the BRAIN initiative.
It is hard to overstate the extent to which the progress in robotics, itself propelled by AI, is transforming the United States military. Under President Obama, the campaign against al-Qaeda is being conducted primarily by drone strikes. It has been said that the F-35 now being built will be the American military’s last manned fighter plane, and that the last American fighter pilot has already been born. Already in development are robots that perform as snipers and others as paramedics. In twenty or twenty five years from now war will likely be fought largely though robots.
But perhaps most important of all, better AI can improve long-term governance. Precisely because technology is moving so rapidly, government needs to get better at recognizing the nature of the new social challenges that technology itself creates. Energy intensive machines began the process of injecting greenhouse gases into the atmosphere at the time of industrialization. Yet almost no one recognized this development or offered solutions until recently in large part because we did not have the necessary tools to model climate change.
Improving AI will aid three important tasks in addressing social policy in fast changing circumstances. First, it will help gather and organize the ever-burgeoning amount of socially relevant data. Already the military is having trouble analyzing all the information that it is getting from its drones. As accelerating technology advances in areas like nanotechnology, biotechnology, and robotics, our capacity for assessing the benefits and dangers of new products will be sorely tested. Democracies prosper if they can use all information available to make the best decisions possible. AI is an indispensable tool for organizing salient information.
Second, AI provides assistance in analyzing the data it organizes. In fact, it is expected that computers will be able to independently generate testable hypotheses by 2020, aiding human researchers in testing the full range of explanations of the causes of social behavior and the effects of government policy. This capability should allow more hypotheses to emerge from data rather than being imposed on the data.
A third benefit of greater information processing is an increased capacity to predict natural catastrophes and either prevent them or avoid their worst consequences. The more sophisticated the simulations and modeling of earthquakes, weather and asteroids, the better such measures are likely to be. More accurate assessments of the risks of catastrophes will help society decide how to direct its limited resources to focus on the most serious ones
The final advantage of BRAIN is deeper comprehension of the potential dangers of AI itself. In his well-known article, “The Future Does Not Need Us”, Bill Joy, the former chief technologist for Sun Microsystems, acknowledged that artificial intelligence may well realize its holy grail: a machine that surpasses humans in a general intelligence. But he fears the coming of strong AI. For Joy, man resembles the sorcerer’s apprentice—too weak and ignorant to master the master machines that will take over.
Joy argues that the way to avoid being mastered by our machines is to stop research into AI altogether. But given that greater machine intelligence is at the heart of military might, such relinquishment is impossible in a world of competing sovereigns. The better course is to ensure that AI remains friendly to humans. What better way is there to do this than by illuminating the dark parts of our own minds with the light of science?
Man is both nature’s master of tools that can shape the world as well as its master of the information needed to understand it. But danger arises from the gap between our capacity to transform the world and our capacity to understand the effects of those transformations. Today’s technology is radically enlarging our ability to remake the world. The BRAIN initiative can help us comprehend the world of relentless change that we are creating.