You have something personal to discuss with a friend. You arrange to meet for a coffee and a chat in a public restaurant. It’s noisy, and you have a good heart to heart in the belief that nobody else can hear what you’re saying. Or you leave a message on an answer phone or via a phone app and think that it’s just between yourself and the recipient.
Fast forward to the near future, and your innermost thoughts could be soon be recorded and stamped as government property. On Nov. 17, the U.S. announced that it intended to develop a new project to be called Automatic Speech recognition in Reverberant Environments, a.k.a ASpIRE. The new initiative is being directed by the Office of the Director of National Intelligence (ODNI), and the Intelligence Advanced Research Projects Agency (IARPA), who are planning to build on old speech recognition software to create a program that can scan all recorded speech across the globe.
ASpIRE is going to be a dedicated project solely for the purpose of intelligence gathering; it will be able to use recorded data collated from a variety of different sources including YouTube videos and chatter recorded by microphones placed in public areas. Speech recognition is a notoriously fallible science with many pitfalls, many of which were outlined in in a 2013 paper titled “What’s Wrong With Speech Recognition.” Researcher and author of the paper, Nelson Morgan, describes the process as “the science of recovering words from an acoustic signal meant to convey those words to a human listener.”
Despite the success of programs such as "Dragon Naturally Speaking," the usefulness of currently available speech recognition programs is actually fairly limited to a relatively small number of situations. Expanding the capabilities of such programs has been surprisingly difficult and has therefore restricted their use in intelligence operations, hence the dedicated focus for the ASpIRE initiative.
The US military has been developing speech recognition technology since World War II, and since then it has been built on a statistical methodology known as "Hidden Markov Modeling." This concept of probability is based on the principle that future states depend only on the present state and not on the sequence of events that preceded it; however, Nelson Morgan argues that this principle could now be holding the field back.
“In short,” Morgan wrote, “the speech recognition field has developed a collection of small-scale solutions to very constrained speech problems, and these solutions fail in the world at large. Their failure modes are acute but unpredictable and non-intuitive, thus leaving the technology defective in broad applications and difficult to manage even in well-behaved environments. In short, this technology is badly broken.”
The ASpIRE program intends to adopt a lateral approach to the issue, explained Mary Harper, ASpIRE Program Manager at IARPA: "The ASpIRE challenge is aimed at identifying entirely new approaches to speech recognition that will do away with the need for extensive – and expensive – training data to achieve results."
New artificial intelligence applications such as the Iphone’s Siri could pave the way for new directions in the field, said Nelson in his paper:
“There is an existing significant example of speech recognition that actually works well in many adverse conditions, namely, the recognition performed by the human ear and brain," he wrote.
"Methods for analyzing functional brain activity have become more sophisticated in recent years, so there are new opportunities for the development of models that better track the desirable properties of human speech perception."
Enabling speech recognition to operate in a variety of different environments could expand its potential to collect speech data for intelligence purposes, and then there would be no shortage of tools available to provide that data. Every single mobile device with increasingly popular voice-activation software could be utilised, including iPhones, the new Google Moto X superphone and the Apple Watch.
But would this type of data collection be legal?
The defeat of the U.S.A. Freedom Act now means that the National Security Agency can continue to collect this type of data from cell phone users, and whether they are speaking in public or in private, a judge may rule you don’t have a reasonable expectation of privacy. This could be the least of our problems, however: the very air we breathe could soon give away our secrets.
It is predicted that in the not-too-distant future, listening devices will be so small that they can be dispersed in the air in a form known as "Smart Dust." In a development that appears to jump straight out of science fiction, scientists have created a microphone that is just one molecule of dibenzoterrylene, so small it is invisible to the naked eye yet is still a highly effective listening device.
"Smart Dust" was a term used by Berkeley professor Kristofer Pister to describe these tiny electronic bundles of power, sensors, computing and communications electronics, which are so small and cheap to produce that they can be scattered like a handful of sand. Yet, despite their diminutive proportions, their technological capabilities are huge: they are able to sense their environment, perform basic data processing and communicate with each other.
Their application within the military is obvious: remote sensors have already been successfully utilised to track enemy troops, and the untraceable nature of Smart Dust expands this potential still further.U.S. Air Force Major Scott A. Dickson believes that future micro-electromechnical systems or MEMS will “sense a wide array of information with the processing and communication capabilities to act as independent or networked sensors. Fused together into a network of nanosized particles distributed over the battlefield capable of measuring, collecting, and sending information, Smart Dust will transform persistent surveillance for the warfighter [sic].”
Their potential for the intelligence services is now the intriguing and rather worrying question: these minute microphones have the capability to record conversations anywhere, anytime, without detection, and combined with the future ASpIRE project, almost every conversation on earth could be vulnerable to official eavesdroppers.
In practical terms there are still issues to overcome with the technology: providing remote power to such tiny devices is difficult as they are too small for effective solar devices even if sunlight is available in their location. But they require very little power and the technology is being improved all the time, as are the new forms of speech recognition software. It is possible to build antennas for the microphones from graphene that is only a few atoms thick.
Even without the aid of Smart Dust, there is no shortage of data already available for use with ASpIRE as, in addition to voice recordings collated via mobile phones, conversations are being routinely taken in commercial exchanges when customers call companies, and also in law enforcement where recordings of all interviews are taken and stored. True "privacy" is fast becoming a diminishing concept across the globe, though we are told that the actions of intelligence agencies are essentially taken in the interests of national security rather than population control. This type of technology falling into the wrong hands is a major consideration, however, and secrecy may soon become an out-dated commodity. We are so used to living as separate entities, but could this be a positive development as we take physical steps towards a true "collective consciousness?"
In the future, the answers to most questions may be found blowing in the wind…