Rise of the A.I. thought police
Julie Beal — We know we’re being surveilled in the matrix: AI Law, empowered by algorithms, feasts on Twittered hate crimes and the like to try to predict crime. We are leaving virtual trails of data which are used to feed simulation models, for predictive analytics – but we can still opt out, throw away our phones, disconnect. There is still some control over what they take from us. But AI Thought Police wants more – to climb into our minds, understand our physical make-up, really get to know us.
We are all under suspicion, but AI needs to know which ones to focus on. So the US army is developing methods to covertly identify and track people who plan to do ‘something bad’. Hidden sensors will be used to detect AI’s version of ‘adversarial intent’ by reading and cataloguing our emotions and health.
A report called ‘Remote Detection of Covert Tactical Adversarial Intent of Individuals in Asymmetric Operations’ was authored by the US Army Research Laboratory in 2010; it details the requirements of researchers wishing to gain funding from the US Federal Government by developing techniques to hone in on individuals in crowds, to detect antagonistic attitudes among the “clutter” of innocents. The prime directive to protect national security, counter ‘insurgency’, and generally ‘keep the peace’, however, means the technology that is developed will spread beyond airports and be used for wider civilian applications, such as “crowd control and in antidrug, anticrime, and immigration enforcement.” In fact, applications in the civilian economy are said to be plentiful, and also include “border security, and ensuring the security of government and private personnel and property”.
The report points out that fusion of information of different types is getting better, such as combining data obtained with a laser with photographic or video images. This technique can help reveal “seemingly hidden patterns”; but the army wants the data to be detected from at least 3m away, and preferably up to 50m away, for use in “asymmetric defense scenarios”.
The report is basically a solicitation for research and development initiatives which can unobtrusively monitor physiological, biological, and physical reactions, and other behavioural signatures (facial expressions, eye movements, gestures, heart rate, perspiration, anomalous behaviour, etc.) in order to deduce thoughts and emotions. From this they hope to be able to detect adversarial intent, but the subjects mustn’t know they are being studied. Several trials are noted to have attempted to do this, and include:
- Future Attribute Screening Technology (FAST) – Department of Homeland Security, DHS)
- Violent Intent Modeling and Simulation (VIMS) – DHS
- Detection of Intent through Perception of Biomotion Signatures, and Visualization of Belief Systems - U.S. Army Research Laboratory Human Research and Engineering Directorate, ARL/HRED
- Remote and Passive ID of Electrodermal Response – Night Vision and Electron Sensors Directorate, NVESD
- Behavioral Signatures, and Human MASINT – U.S. Air Force Research Laboratory, AFRL
- Hostile Intent – U.S. Naval Research Laboratory, NRL
- Computational Modeling of Adversary Attitudes and Behaviors – U.S. Air Force Office of Scientific Research, AFOSR, George Mason University
- Dynamic, Adaptive Techniques for Adversary Behavior Modeling – AFOSR, University of Maryland, U MD
- Human, Social, Cultural, and Behavioral Modeling – U.S. Army Research Office, ARO, Carnegie Mellon University, U MD
- Tools for Recognizing Unconscious Signals of Trustworthiness Program (TRUST) – Intelligence Advanced Research Projects Activity, IARPA
The main directive of the army report is to develop “theoretically justified quantitative predictive principles (models) and their implementation in tractable analytical and computational procedures.” In other words, it’s all about crunching numbers using algorithms and really is very ambitious, because it will go beyond mere physiological signs such as fear or stress, to try to pinpoint cognitive intent, even from people who show no outwards signs. This is also the trend in law enforcement, just like in the Minority Report, where predictive analytics are used to make sense of surveillance data – we are empowering machines to be the Thought Police.
Researchers are expected to look to knowledge gained from:
- cognitive science
- sociocultural anthropology, and
- information science.
The latest knowledge from each of these disciplines, of detectable signs of ‘(mal)intent’, is to be pooled together with the best innovations in remote bio-sensing techniques. The proposed list of characteristics to track and measure is:
- Posture rigidity
- Heartbeat waveform
- Heart rate
- Breath rate, volume approximation, patterns, anomalies
- Wheezing, coughing, gasping
- Blood pressure trends: waveform shape and transit time
- Pulse-wave velocity giving a beat-by-beat approximation of blood pressure
- Movement: fidgeting, remaining still, shaking, shivering, having spasms
- Body stiffness, muscle tension, resonant frequency of body movement
- Voice stress analysis and voice onset timing
- Gastrointestinal distress, bowel sounds
- Reluctance to engage socially: distance from others, response to attempts to engage verbally
- Observation tendencies of subject, eye-glancing, head turning, situational awareness
- People whose actions are coordinated or who are actively avoiding each other
- Exposure to bomb-making materials/chemicals
- Hyperthermia from stress (generally expressed in the face, palms of the hands, and soles of the feet)
- Gait as indicators of stiffness (stress) or carrying a load or wearing protective clothing
- Breath biochemistry
- Microbiological organisms on skin or clothing
It is also suggested that as well as taking such measurements from people in a relaxed state, little things should be put in their way to see how they respond. This “active elicitation” of information should be subtle, so as not to arouse suspicion and skew the results, because it involves presenting a stimulus to the subject and analysing the behavioural response, i.e. seeing how the person reacts to a certain noise, picture, person, etc. – measuring physiological and emotive responses from a distance. These little incidents designed to elicit a response are called ‘perturbations’, and can be physical or psychological. The examples given include putting an obstacle in the path of the subject so as to cause a change in their way of moving, or flashing a picture of the subject onto a TV screen, “to let them know that they are being watched”.
The report also stresses the need for the behavioural metrics obtained to be translated into something that can be readily understood by the user; in other words, the computer has to decide what constitutes intent, in order to alert the person using the system.
The main obstacle to developing such a system is the testing stage – how to verify that what the machine classes as ‘fear’ or ‘excitement’ is correct. Who’s to say what the truth is about the person’s original intent? It is quite simply impossible to assess the success of any system designed to calculate ‘mal-intent’, for two reasons: 1) the designers’ belief system dictates what kind of ‘intent’ is undesirable, and 2) there is no ground truth available against which to test the results.
Research suggestions include contrived scenarios and analysis of battlefield video footage, but the best hope would seem to lie in the study of people playing online games – their physiological readings can be analysed for both personal and universal anomalies. These readings are highly valuable for neuromarketers looking to fine-tune their targeted advertising.
In fact, the gamification of life is creating the ideal setting for profilers of all kinds because the trend in online gaming and overall human-machine interfaces is the connection of body and mind to the Internet – the human responses are tracked, catalogued and analysed, because they are worth a lot of money. Individual gaming style can also be used as a behavioural biometric. Relevant data can also be gleaned from employees who track themselves: over two-thirds of companies around the world are running ‘wellness programs’, and self-tracking tools such assmart apps for mobile phones are becoming the norm.
Consumers can even use neural interfaces at home now, with the ‘fun and funky’ Emotiv EPOC, a headset which reads users’ brainwaves (EEG), enabling them to power software with their minds.
But some people will not volunteer such personal information, so they’re the ones that need working on. There are plenty of ways to remotely, and covertly, probe for evidence; Laser Doppler vibrometry, for instance, ‘interrogates’ the body to analyse “heartbeats, breaths, body resonances, muscle stiffness, voice, and voice stress.” Other suggested methods include:
- E-Field: Passive free-space electrodes (capacitively coupled) can detect electrical activity of the heart, brain, muscles, and hidden electronic devices.
- Radar: Active radar can detect and track human gait, heartbeats, breaths, radar cross section (RCS), and arm/leg movements.
- Ladar: Active laser radar produces three-dimensional (3-D) imagery related to “hostile” stance, 3-D posture, gait dynamics, facial recognition, carrying of backpacks, and unnatural frame proportions.
- Gas chromatography: Active collection of exhaled gas, such as through a suction tube near a microphone or portal, can provide a sample for trace gas analysis, chemical emission, odor, genetic material, and biochemistry assessment.
Bizarrely, even the micro-organisms which live in and on our bodies can be analysed to reveal things about us. These ‘prokaryotes’ undergo genetic mutation in a short-time scale, and the type/rate of these mutations can indicate where someone has been recently, what they’ve eaten, and what they’ve been exposed to – even what they have been doing. Patterns in the genetic mutation of these prokaryotic microorganisms can be identified for an individual, so these can be checked to see if they are consistent with someone’s “purported history, identity, and activities”.
This was two years ago, when the range of techniques represented novel innovations which were yet to be tested and developed, as all of them present difficulties; however, the proposal is that, just like the multi-modal approach to biometric authentication, these techniques should be fused, i.e. several methods should be used together – the more evidence that can be provided from surveillance, the better. Just in case.
Meanwhile, there’s been some progress . . . the Future Attribute Screening Technology program has moved beyond its testing phase; a Privacy Impact Assessment Update was formulated in December, 2011, which reveals the FAST program is still on track, and has in fact been revised to include a new type of research called ‘Passive Methods for Precision Behavioral Screening’: the combined program is now known as ‘FAST/Passive’, and will attempt to screen individuals covertly by measuring (passive) responses to tactile, visual and audio cues, rather than just using cues such as questions which require an active response. It is hoped this can be achieved at a distance, by using “non-intrusive sensors” to detect malintent, as per the requirements detailed in the 2010 report. Critics argue the number of false-positives created by the sensors make the system ludicrously unworkable. In addition to this, much can be deduced about our characters from the sensor readings, and will be of enormous value to neuromarketers. Yet research presses on.DARPA, for instance, is figuring out how to glean cognitive biometrics from people without their knowledge or consent, by getting them to ‘play games’ which mimic ordinary human computer interactions, so the ‘game’ is imperceptible, and they don’t even realise their responses are being tested. The method will “incorporate the principles of adaptive learning, behavior modification and game theory”.
Another system able to fit the remit of the US Military solicitation report (but yet to make the headlines) is the new ‘Millimeter-Wave Remote Biometric Identification and Tracking (mmWRBIT) System for Security Applications’. Designed by Argonne National Laboratory, in collaboration with Northwestern University, the RBIT is portable, and is able to “remotely identify and persistently track a subject while recording his or her heartbeat, respiration, and movement”. It can be used to scan large crowds and zoom in on suspects; it can then be aimed at the chest to track heart rate and breathing rate, and movement, whilst the pan and tilt camera maintains a lock on the suspect’s face for facial profiling. The cameras work under even harsh conditions, whilst the sensors can take measurements from tens of meters away, even through walls.
Although it has ostensibly been designed to ‘find terrorists’, the researchers suggest several other applications for the technology:
It can be used for biometrics security, battlefield triage, to search for vital signs after disasters, to monitor a patient’s heart condition and movement, to combat identity theft, for on‐line monitoring of operating machinery, and to monitor vehicles to avoid collisions.
The RBIT can also be used for “condition monitoring of operating machinery, and non-destructive evaluation”, so the technology could become pervasive, especially since it can be used to detect chemicals, including covert monitoring of nuclear facilities from miles away. This is all the more significant in the light of market research by the Homeland Security Research Corporation (HSRC); business is booming for anti-terrorist technologies and markets such as “covert walk-through biometric identification corridors”, and “fused standoff video surveillance and biometrics”, as detailed in a comprehensive and forward-looking report, entitled, ‘Standoff Person-Borne and Vehicle-Borne Explosives & Weapon Detection: Technologies and Markets’ (2012-2016). The HSRC has also released a report on the latest advances in people screening, and expected market trends. Such identity profiling could either improve, or be said to justify, any attempt to judge intent – an all round picture could always be said to be needed. A web applet called VIBES (Visualization of Belief Systems), for instance, can be used to study the effects of propaganda on “the dissemination of beliefs through a social system”, by analysing “social, behavioral, and cultural characteristics, with state-of-the-art dynamic network analysis and visualization.”
Another development of mm wave (mmW) technology is also said to help establish intent: Raytheon has developed a series of non-lethal tools which can be aimed at an adversary, in order to “establish and confirm the intent of aggressive individuals without increasing exposure to liability.”
The company’s Silent Guardian series constitutes a range of sizes for the ‘Active Denial System’ (ADS): anyone hit by the mm wave beam experiences an horrific burning sensation, which, after just 2 or 3 seconds, is too much for anyone to bear. It is said to be ‘safe’, because it is (supposed to be) a momentary sensation which makes the target move away, and hence they are actively denied from entering the forbidden area. The burning sensation is caused by a 1 mm radio wave which penetrates the top 1/64th of an inch of skin, and heats it up, just like a microwave. It is said to be a safe and well-tested technique, which the US Air Force claims has only one tenth of a 1% chance of causing a lasting injury, and that this would only be minor, such as a blister or rash.
It can be used in maritime situations, such as deterring pirates, and for crowd control, in prisons and during protests – and even for controlling hungry hordes rushing for food aid.
The active denial system has been developed for nearly twenty years now, and it relies on a water-cooled gyrotron, developed by CPI Inc., based in California. The gyrotron creates the 95GHz (mmW) radio frequency beam which inflicts intense pain. Many people have heard of VMADS(vehicle mounted ADS) for military situations, and some have heard of the more recent developments where ADS is being deployed in prisons in the US. This is the result of Raytheon’s latest addition to its Advanced Security and Directed Energy Systems product line – a much smaller version of the ADS Silent Guardian range has been designed so as to not require any certification as a weapon system. According to Jane’s,
The final, even smaller, member of the family has a 25m range and breaks with the naming convention as it is called the Assault Intervention Device. This was developed under an August 2010 contract from the US Department of Justice, which was looking for a non-lethal system able to break up riots in California prisons that had become too dangerous for prison officers.
In its prison configuration the system is remotely controlled, enabling officers to ‘hose’ the beam across an unruly group from behind cover.
The SG-R50 can be fired at people up to 50m away, with a beam “roughly the shape and size of a human head”. The full range of Silent Guardian products was showcased at the 2012 SOFEX trade show; Mike Booen, vice president of Raytheon’s Advanced Security and Directed Energy Systems product line, understands the potential market for the technology, and has commented,
Customers worldwide are interested in these systems to protect their high-value sites and critical assets.
Currently, ADS is employed by the US Marine Corps, and is set to be used for crowd control more generally, including ’perimeter and entrance control operations’.
Several sites mention that the mmW technology will soon be available as a portable ‘tool’, and a Department of Homeland Security document details the development of the prototype ‘Millimeter Wave Stun-Gun’ by researchers at the College of Judea & Samaria (CJS), under the direction of Dr. Moshe Einat, in a program administered by the US Joint Non-Lethal Weapons Directorate. The U.S. Marine Corps and police are also said to be trying to develop such a ‘Man-Portable Active Denial System’.
Although the mmW ADS has been given approval by the Directed Energy Bioeffects Division of the US Air Force Research Laboratory, the mildest concerns are that the effects of the technology could be more intensive on sweaty skin, which could exacerbate the burning, and there are no experiments which prove it would not damage the eyes, or its effect on women, children, and the sick. It also seems to be affected by hot and humid conditions. But one of the biggest threats of ADS, especially the ‘stun gun’, is its potential to be used as a weapon of torture. The fact that short bursts leave no mark means that officials will be able to repeatedly fire the device, delivering incredible pain in an undetectable way. All of the media reports say it forces people to move out of the way, and so it would seem that test situations have escape routes built in as a matter of routine, as the weapon is designed to be used in open spaces. The chances of it being used on individuals who cannot escape the beam should be enough to prevent its development, despite assurances of ‘built-in safety mechanisms’.
It gets a lot more serious than that, though – testimonies by civilians in Iraq indicate this weapon has been used in the most dreadful lethal manner. Bearing witness to what appears to have been a ‘burning beam’ attack, Majid Ai Ghalzali describes in a short film on YouTube the bodies he saw in 2003 in Iraq, after they’d been attacked by the US military; the heads and bodies had been burnt away, but not their bodies or cars, by an unexplained ‘beam’. No bullets or shrapnel were found in or near the bodies. The military cleaned up the site afterwards – even replacing the earth. Donald Rumsfeld and General Richard Meyers squirm and struggle to answer a question posed at a press conference about the new microwave weaponry. A powerful version of the weapon is shown in the video destroying missiles in the sky. Marc Garlasco from Human Rights Watch also features in the film, and warns of the illegality of mmW weapons:
We have to be very careful, because in international law – it’s very clear – that devices created solely for the creation of pain, can eventually lead to torture and are therefore illegal.
An online petition has been mounted by the Medical Whistleblower Advocacy Network against the use of ADS (or the ‘Assault Intervention Device’, which it is now being called) in prisons, citing overdose and misuse by prison officers as prime concerns. The group notes the position of the ACLU, which considers use of the weapon to be a violation of the Eighth Amendment, basic human rights norms, and international law.
Last but not least is the old ‘Problem: Reaction: Solution’ routine – the ‘answer’ to the outrage caused by TSA scanners at airports, which show images of the body, and the numerous patdowns, is the introduction of new mmW imagers, which only reveal the location of anomalies, such as hidden weapons. These have been installed around the world, with new scanners announced for theUK: the Automatic Target Detection machines are produced by ProVision. Guidelines published by the International Commission on Non-Ionizing Radiation Protection state,
… mm wave body scanners operate in pulse modes. The power levels employed by these mm wave body scanners are low but can generate power densities up to 1.0 kW m_2 for a pulsed field averaged over the pulse width. The resulting human exposures are about a tenth of currently recommended guidelines for the general public.
This year’s SPIE Security and Defence Conference featured a joint session on Terahertz sensing with the Millimetre Wave and Terahertz Sensors and Technology Conference, considering various aspects of the technology, including biometrics for security screening, and techniques for longer range surveillance. Terahertz sensing, in the range between 300 GHz to 10 THz (1 millimeter to 30 micrometers), is the latest buzz, and is said to be “attracting increased interest from military and security fields”. Time will tell what novel applications the Terahertz bring.
*This article originally appeared on Activist Post