December, 2017

Report Examines Use of Tasers on Jail Inmates

Recently, Reuters – as part of an investigative series – examined the use of Tasers by corrections officers on inmates. The report found 104 inmate deaths where Tasers were involved occurring since 2000. Of those, only two of the inmates were armed, 18 were actively involved in a physical struggle that could result in serious injury to guards or other inmates, and the large majority were either handcuffed or otherwise immobilized when the Tasers were deployed.
 
Twenty-seven states (including Michigan) allow corrections officers to be armed with Tasers, but most of the states only allow them for special units involved in high-risk roles; the federal prisons do not use them. Local and county jails, which hold about 728,000 people and often have a high inmate to guard ratio (sometimes even 80 inmates to one guard), more commonly use Tasers, especially, as noted by ACLU senior staff counsel Eric Balaban, because jails often are “high-intake, high-turnover facilities,” where people may “be intoxicated on alcohol or drugs [or are] in mental health crisis.”

The unwarranted and excessive use can be costly to the local governments. Lawsuits were filed in almost 70% of the 104 inmate deaths; settlements or payouts were awarded in 93% of those lawsuits. A series of lawsuits brought by 40 current and former inmates in San Bernardino, California, and which included allegations of Tasers being used just to inflict pain or torture, were settled in 2017 for $2.8 million. One inmate complained that he was tased five times a day over a two-month period – in “surprise attacks” by deputies – as he delivered meals to other inmates.

Some facilities are phasing out Taser use. For example, in the two jails in Franklin County, Ohio, Tasers were used on average once per week between 2008 and 2010. After lawsuits and U.S. Department of Justice intervention in 2011, the Taser use declined to one per month in 2011, one per quarter in 2014, three deployments in 2015, and one use of a Taser in 2016.

Sources:  Peter Eisler, Jason Szep, and Charles Levinson, Shock Tactics; Part 6, “Inmate deaths reveal ‘torturous’ use of Tasers,” reuters.com, December 6, 2017:
https://www.reuters.com/investigates/special-report/usa-taser-jails/

Nvidia AI Creates Realistic Fake Images and Video

A research team at Nvidia, led by Ming-Yu Liu, used unsupervised learning and generative modeling to allow an AI program to better create imagined scenes in images and video. The program can change video content it receives to, for example, create images of people, change a day scene to a night scene, or change a summer day to a winter day. As one reporter recently stated, it can “imitate reality to a startling degree.”

Bernard Marr, writing for Forbes, noted that supervised learning algorithms are most common and are widespread; the outcomes from data entered into an algorithm are known. With unsupervised learning, the outcomes are not known; the AI utilizes its own logic operations to reach an outcome.

Recently, at a Nvidia blog, Kimberly Powell suggested some potential benefits of the new technology, including that “deep learning experts can apply the technique across domains. For self-driving cars alone, training data could be captured once and then simulated across a variety of virtual conditions: sunny, cloudy, snowy, rainy, nighttime, etc.” However, others caution that humans may become immersed in a “hyper-reality” close to real life, where one’s consciousness cannot distinguish between what is real and what is simulated, and that it could lead to the creation of fabricated video-evidence that could be disbursed through social networking and media sites.

Tristan Green in a recent article noted, “It no longer matters whether AI can fool everyone; the important thing to understand at this juncture is that it’s become good enough to fool some of us.”

Sources:  Kimberley Powell, “NVIDIA Researchers Showcase Major Advances in Deep Learning at NIPS.” blogs.nvidia.com, December 3, 2017:
https://blogs.nvidia.com/blog/2017/12/03/nvidia-research-nips/
Phoebe Weston, “Can YOU tell which is fake? AI creates a 'false reality' that is almost identical to the real world.” Dailymail.co.uk, December 6, 2017:
http://www.dailymail.co.uk/sciencetech/article-5151117/AI-create-life-like-false-reality.html
Tristan Greene, “Nvidia’s new AI creates disturbingly convincing fake videos,” thenextweb.com, December 4, 2017:
https://thenextweb.com/artificial-intelligence/2017/12/04/nvidias-new-ai-creates-disturbingly-convincing-fake-videos/
Tristan Greene, “Nvidia’s new AI creates ‘people’ out of thin air,” thnextweb.com, November 1, 2017:
https://thenextweb.com/artificial-intelligence/2017/11/01/nvidias-new-ai-creates-people-out-of-thin-air/
Bernard Marr, “Supervised v Unsupervised Machine Learning – What’s The Difference,” forbes.com, March 17, 2017”
https://www.forbes.com/sites/bernardmarr/2017/03/16/supervised-v-unsupervised-machine-learning-whats-the-difference/#42b1a7d6485d

by Neil Leithauser
Associate Editor