The intelligence establishment's dream supercomputer will make Raytheon's RIOT program seem like child's play

The US intelligence community is obsessed with data. The NSA wants it all, and is prepared to keep it for as long as 100 years. The National Counter Terrorism Center (NCTC) last year told us it’s now dipping into or collating every bit of information we give to federal government agencies under one roof, to mine it for 'suspicious' information that may be linked to terrorism. State and local law enforcement, with help from the Department of Homeland Security, have established so-called intelligence "fusion centers" in most states nationwide -- little spy centers of their own, where they can view surveillance camera feeds and access intelligence databasesUnited States surveillance drones at home and abroad collect impossibly enormous quantities of data. Satellites do, too

The government has largely gotten away with blowing up the parts of the Bill of Rights that should limit its ability to collect and store in intelligence databases massive amounts of information on all of us, absent any suspicion of wrongdoing or probable cause to believe we are involved in some kind of criminal activity. But it hasn’t quite figured out what to do with all that data. Not yet.

Enter technologies like the Raytheon RIOT social media data mining system. The company boasts in a freakishly frank video about how it can track a fictional person as he goes about his daily life, simply using the information he posts to public facing social networks. The RIOT system is an example of a technology that collates and regroups large amounts of data to make information useful to human analysts. It’s essentially an information management system, like a 21st century index card operation with brightly colored maps and drop down menus.

But researchers at the far out intelligence office IARPA want access to information management tools that go well beyond this kind of indexing and reorganizing. Raytheon RIOT is surely useful to government agencies that want to track activist groups or people they suspect are involved with drug activity, but the inputs it uses are largely flat: GPS coordinates, written text, associations drawn from friend networks documented in plain text. All of that is powerful enough on its own. After all, historically data-hungry regimes have managed their spying operations entirely on paper.

Still, the most secretive and powerful intelligence agencies want something that will do for the totality of the data they vacuum up what Raytheon RIOT does for social networking information. And that’s a lot more complicated -- and potentially a lot more dangerous for civil liberties.

In a Federal Business Opportunities advertisement dated January 14, 2013, the Intelligence Advanced Research Projects Activity (IARPA) offered up a challenge to companies like Raytheon and the innumerable other military and surveillance contractors that work on data management solutions. IARPA says it wants investment in “high-risk, high-payoff research that has the potential to provide our nation with an overwhelming intelligence advantage over future adversaries. This research is parsed among three Offices: Smart Collection, Incisive Analysis, and Safe & Secure Operations.” The program aims to find new ways to collect yet more information to add to the massive pile of data government agencies already control, and maybe more importantly, come up with ways of deciphering, comprehending, and organizing it so that it’s actually useful to government agents.

Among the “topics of interest” that the Office of the Director of National Intelligence (ODNI), which houses IARPA, wants researchers to explore are:

  • Methods for developing understanding of how knowledge and ideas are transmitted and change within groups, organizations, and cultures;
  • Methods for analysis of social, cultural, and linguistic data; 
  • Multidisciplinary approaches to assessing linguistic data sets;
  • Methods for measuring and improving human judgment and human reasoning; 
  • Methods for extraction and representation of the information in the non-textual contents of documents, including figures, diagrams, and tables;
  • Methods for understanding and managing massive, dynamic data; 
  • Analysis of massive, unreliable, and diverse data;
  • Methods for assessments of relevancy and reliability of new data; 
  • Methods for understanding the process of analysis and potential impacts of technology;
  • Multidisciplinary approaches to processing noisy audio and speech; 
  • Development of novel top-down models of visual perception and visual cognition; 
  • Methods for analysis of significant societal events; 
  • Methods for estimation and communication of uncertainty and risk; 
  • Novel approaches for mobile augmented reality applied to analysis and collection;
  • Methods for topological data analysis and inferences of high-dimensional structures from low-dimensional representations;
  • Methods for the study of algorithms stated in terms of geometry (computational geometry);
  • Methods for geolocation of text and social media;
  • Methods to make machine learning more useful and automatic;
  • Methods to construct and evaluate speech recognition systems in languages without a formalized orthography; and,
  • Methods and approaches to quantifiable representations of uncertainty simultaneously accounting for multiple types of uncertainty.

The goals laid out above are the true cutting edge of information collection and analysis research. In short, the ODNI wants to be able to teach computers how to think more like humans, to pick up sarcasm and irony, to understand when those devices are deployed in foreign languages, to learn how to integrate disparate sets of data from wildly different sources, and ultimately to automatically make sense of it all without any human interaction whatsoever. It sounds like IARPA wants a computer that can understand the world and all its little details in a manner most of us can't even conceive of.

So what?

As my colleague Jay Stanley eloquently explains, Raytheon RIOT is a worrisome technology because it raises the real concern "that government agencies will blunderingly use these techniques to tag, target and watchlist people coughed up by programs [like it], or to target them for further invasions of privacy based on incorrect inferences. The chilling effects of such activities, while perhaps gradual, would be tremendous." 

Chilling as RIOT may be, in some ways the knowledge driving the creation of this kind of information management system is far from groundbreaking. The tool may well be able to track individuals and groups as we go about our daily lives. But can it predict political revolutions using open source and classified intelligence information combined with wiretap intercepts, all without human interference? Can it teach itself how to think through previously unknown problems while simultaneously translating millions of hard-to-hear phone calls into English, making note of the cultural signifiers and metaphors in those conversations, and transcribing them? Can it do all of that while making a map of the movements those people were making as they talked on those phones, and connect all of that data to economic indicators as they fluctuate on the international market, or to medical surveillance trends of communicable disease? 

It sounds crazy, but prediction of the future is what the government is after. In order to predict the future, the government needs to know literally everything about what's happening in the world, down to the smallest detail. After all, had the Tunisian government received a tip about Mohamed Al Bouazzizi's plan to self-immolate in front of a government building, and had it stopped him, the Arab Spring -- a thorn in the side of the US intelligence and defense establishment -- may never have happened as it did.

You may be thinking that this kind of omniscience is impossible, or a waste of time and energy. Maybe. But it's exactly what the US intelligence agencies are after. And they appear to think that super computers will help them get there. Super computers that are a million times more incisive than we could ever hope to become, adept at not only collecting and analyzing all conceivable data in the world but also teaching themselves how to think through unforeseen problems on the go.

It may never happen. Plenty of IARPA projects go nowhere. But this one isn’t going away any time soon, and developments in nanotechnology, robotics and supercomputing make it all the more likely that at least some -- if not all -- of these goals will someday be met. 

And when they are, they’ll make Raytheon RIOT, as serious a threat to our liberty as it has the potential to become, look like absolute child’s play. 

Sign up for email alerts // Contact us

Theme by Danetsoft and Danang Probo Sayekti inspired by Maksimer