Thursday, February 27, 2020

Curating the Internet: Science and technology digest for February 27, 2020

Harvard researchers find that free open source software (FOSS) comes with a cost; Small levels of daily alcohol consumption may increase longevity, but binge drinking may reduce it; A recently declassified report reveals that the NSA phone meta-data monitoring program is expensive and unproductive; An argument that researchers need to study deceptive practices that emerge naturally in artificial intelligence (AI) systems; and a Steem essay describing a recent search for dark matter in the Earth's core.






Fresh and Informative Content Daily: Welcome to my little corner of the blockchain

Links and micro-summaries from my 1000+ daily headlines. I filter them so you don't have to.


First posted on my Steem blog: SteemIt, SteemPeak*, StemGeeks.

  1. The Hidden Vulnerabilities of Open Source Software - Free and open source software (FOSS) has become a dominant player in the technology space during the last decade, with 80-90% of the typical application being comprised of FOSS components. However, according to Frank Nagle and Jenny Hoffman, this comes with a downside. Working with the Linux Foundation, the pair uncovered a number of troubling trends and reported them in ‘Vulnerabilities in the Core,’ a Preliminary Report and Census II of Open Source Software. The problems that they reported included (i) inconsistent naming conventions that can interfere with proper maintenance, security, and transparency; (ii) hosting FOSS packages under personal developer accounts; and (iii) Discontinuation of development efforts that can lead to long-lasting security vulnerabilities. The article suggests that this state of affairs has emerged because FOSS moved from the domain of hobbyists to robust production use during the last 20 years, and it suggests that commercial organizations should begin taking action by conducting internal reviews and staying engaged with the developer communities. The research effort is continuing, and interested organizations can sign up to participate, here.

  2. Alcohol consumption in later life and reaching longevity: the Netherlands Cohort Study - Researchers obtained a statistically significant finding that the highest likelihood of reaching 90 years of old among people who consume 5-15 grams of alcohol per day (1/6-1/2 oz.) among both men and women, and a non-significant finding that binge drinking leads to earlier death. The study involved 5,479 participants from the Netherlands. -h/t Daniel Lemire

  3. N.S.A. Phone Program Cost $100 Million, but Produced Only Two Unique Leads - A newly declassified report shows that NSA metadata analysis cost $100 million between 2015 and 2019, but yielded only two unique pieces of information that the NSA didn't already have by other means (like subpoenas to telcos), and only a single significant investigation emerged from it. The outcome of that investigation has not been disclosed. The program was first authorized under the Freedom Act after the 9/11 attacks in 2001, and was most recently reauthorized under the Freedom Act of 2015. It is up for potential renewal again on March 15, 2020. Throughout the program's existence, it's most significant accomplishment "was leading the F.B.I. to scrutinize a San Diego man who turned out to have donated several thousand dollars to the Shabab, the Islamist group in Somalia." The program also has a history of inadvertent collection of records for which it had no authority to collect, which led to a full database purge in 2018. -h/t Bruce Schneier




  4. AI Deception: When Your Artificial Intelligence Learns to Lie - In this article, Heather Roff argues that current thinking about artificial intelligence (AI) misses an important facet of the phenomenon. In particular, she says there is much research into "adversarial attacks, or techniques for fooling AI systems, and there is also much research into detection efforts when some people use AI to fool other people (like deepfake technologies). However, she says there has been little research into the ways that AI can learn to be deceptive on its own, without being steered that way by humans. In traditional definitions, she says that deception involves an intent, which requires a theory of mind, so she redefines deception to work in the context of AI. A definition she suggests is, "hiding resources or information, or providing false information to achieve some goal", and she recognizes two types of deception: acts of omission and acts of commission. To begin preparing for deception by AI, she says that we start by recognizing that it already happens, and then to analyze the circumstances under which it occurs and the level of risk that it entails. From there, she says that we can begin to engineer defenses against AI deception. Stated as the article's subtitle, the main thrust is that, "We need to understand the kinds of deception an AI agent may learn on its own before we can start proposing technological defenses".

  5. Steem @kralizec: Superconductive Magnets Scaned The Earths Core In Search For Dark Matter - In this post, the author describes recent efforts at discovering dark matter in the Earth's core. According to theoretical physicists, the universe should be teeming with a poorly understood substance (for lack of a better word) that is known as dark matter. However, to date, no one has been able to observe it. Most efforts at discovery have focused on searching through space, but this article describes a team of researchers who have developed a technique to search for compact dark objects (CDO) inside the earth through the use of gravitometers, which can detect changes in gravitational forces. The researchers postulated that dark matter inside the Earth should orbit the core with a period of about 55 minutes. Unfortunately, the search was unsuccessful, but even a negative result means that something has been learned. (A 10% beneficiary setting has been applied to this post for @kralizec.)




In order to help bring Steem's content to a new audience, if you think this post was informative, please consider sharing it through your other social media accounts.


This post will also appear on my pre-Steem blog, chescosteve.blogspot.com.




And to help make Steem the go-to place for timely information on diverse topics, I invite you to discuss any of these links in the comments and/or your own response post.

Beneficiaries




About this series



Sharing a link does not imply endorsement or agreement, and I receive no incentives for sharing from any of the content creators.



Follow on steem: remlaps-lite, remlaps
If you are not on Steem yet, you can follow through RSS: remlaps-lite, remlaps.




Thanks to SteemRSS from philipkoon, doriitamar, and torrey.blog for the Steem RSS feeds!

No comments:

Post a Comment