USA: Intelligence Advanced Research Projects Activity (IARPA) is trying to figure out how to predict the future. Humans and machines can be better together.

IARPA is like “the national science foundation for national intelligence […] we’re scientists and engineers who help support spies,” explains IARPA Director Jason Matheny. Think Q Branch from the James Bond series. So how can humans accurately forecast events? Matheny, at the monthly Cipher Brief Georgetown Salon series, said the answer is to co-ordinate more efficiently with machines. In hybrid forecasting competitions run by IARPA, the best teams are the joint human-machine ones, he said. This is because humans and machines augment one another. Humans – who are incapable of collecting and storing as much information as machines – can use machines’ mass data collection capabilities to better their analyses. For example, IARPA’s Open Source Indicators (OSI) program, which seeks to anticipate or detect events by combining all types of publicly available data, was “the first program to detect Ebola in West Africa due to local news reporting of a suspicious hemorrhagic fever [and commercial overhead imagery in which] you could also see increased crowding at local clinics [and] increases in funerals.” Still, the human analysis component is vital because there is no algorithm a machine can use for predicting human behavior. “Whether we’ll ever get to the point where we can create the human society in a super computer and run it out ten years and see what happens, I’m pretty pessimistic,” Matheny remarked.

In addition to working on quantum computing, other super computing, and machine learning, IARPA also tries to answer the question: What makes people wise observers of the world? In an IARPA program called ACE (Aggregative Contingent Estimation), Matheny described a scene in which two different prediction markets competed head-to-head for the title of most accurate fortune-teller. One was a completely unclassified external prediction market, that is, amateur analysts playing around with open source information and hypothesizing outcomes for geopolitical events. The other was inside the intelligence community, with professional intel analysts reviewing both unclassified and classified information to make educated guesses on the future. There were “many cases where the uncleared folks, working only with open source, were substantially more accurate than intel analysts,” according to Matheny. This begs the question, if members of the public who are skilled at observing the world and interacting with machines can engage in behavior historically reserved for the intelligence and national security communities, what else can they do… for good, or for bad?

Bioterrorism is “one of the biggest risks that we’ll face over the next few decades,” said Matheny, adding that “somebody who is technically sophisticated but isn’t part of a nation-state program developing an engineered pathogen […] can cause a pandemic.” How? Individuals can get their hands on open source genome information and the material for creating certain pathogens. In addition, biology has the unique characteristic of self-replication. Perhaps the most troubling part is that this is a situation Matheny thinks we are not prepared for. He explained, “Our national security system has been designed and structured in such a way to focus on nation-state threats, and more recently, on fairly organized groups that do have a signature,” and not on the single biologist or chemist or computer scientist who wants to do indiscriminate harm to society. As biotechnology and artificial intelligence develop, so does the capability – and threat – of the super-empowered individual. Ron Howard’s Inferno is a great example. In the movie, a wealthy individual develops a virus that he wants to unleash on the world, in order to ultimately save humankind from over-population (a not-that-unrealistic possibility, says Matheny). Yet although the collaboration between lone actors and technology can release highly destructive forces, human-machine pairs that work on, for example, geopolitical forecasting for the intel community have the potential to enhance national security and improve the world.


No comments :