AmiMoJo shares a report from Ars Technica: The afternoon commute of Reddit user Beastpilot takes him past a stretch of Seattle-area freeway with a carpool lane exit on the left. Last year, in early April, the Tesla driver noticed that Autopilot on his Model X would sometimes pull to the left as the car approached the lane divider -- seemingly treating the space between the diverging lanes as a lane of its own. This was particularly alarming, because just days earlier, Tesla owner Walter Huang had died in a fiery crash after Autopilot steered his Model X into a concrete lane divider in a very similar junction in Mountain View, California.

Beastpilot made several attempts to notify Tesla of the problem but says he never got a response. Weeks later, Tesla pushed out an update that seemed to fix the problem. Then in October, it happened again. Weeks later, the problem resolved itself. This week, he posted dashcam footage showing the same thing happening a third time -- this time with a recently acquired Model 3. "The behavior of the system changes dramatically between software updates," Beastpilot told Ars. "Human nature is, 'if something's worked 100 times before, it's gonna work the 101st time.'" That can lull people into a false sense of security, with potentially deadly consequences.

The Electronic Frontier Foundation has published a new report calling for a "fiber for all" plan to combat the broadband access crisis in the United States. Government data and independent analysis show we are falling behind the rest of the developed world in this area, and "the U.S. is the only country that believes having no plan will solve this issue," writes Ernesto Falcon from the EFF. "We are the only country to completely abandon federal oversight of an uncompetitive, highly concentrated market that sells critical services to all people, yet we expect widely available, affordable, ultra-fast services. But if you live in a low-income neighborhood or in a rural market today, you know very well this is not working and the status quo is going to cement in your local broadband options to either one choice or no choice." From the report: Very small ISPs and local governments with limited budgets are at the frontline of deploying fiber to the home to fix these problems, but policymakers from the federal, state, and local level need to step up and lead. At least 19 states still have laws that prohibit local governments from deploying community broadband projects. Worst yet, both AT&T and Verizon are actively asking the FCC to make it even harder for small private ISPs to deploy fiber, so that the big incumbents can raise prices and suppress competition, a proposal EFF has urged the FCC to reject.

This is why we need to push our elected officials and regulators for a fiber-for-all-people plan to ensure everyone can obtain the next generation of broadband access. Otherwise, the next generation of applications and services won't be usable in most of the United States. They will be built instead for markets with better, faster, cheaper, and more accessible broadband. This dire outcome was the central thesis to a recently published book by Professor Susan Crawford (appropriately named Fiber) and EFF agrees with its findings. If American policymakers do not remedy the failings in the US market and actively pursue ways to drive fiber deployment with the goal of universal coverage, then a staggering number of Americans will miss out on the latest innovations that will occur on the Internet because it will be inaccessible or too expensive. As a result, we will see a worsening of the digital divide as advances in virtual reality, cloud computing, gaming, education, and things we have not invented yet are going to carry a monopoly price tag for a majority of us -- or just not be accessible here. This does not have to be so, but it requires federal, state, and local governments to get to work on policies that promote fiber infrastructure to all people. Most of the talk lately has been about 5G networks, but the less-spoken truth about these networks is that they need dense fiber networks to make them work. "One estimate on the amount of fiber investment that needs to occur is as much as $150 billion -- including fiber to the home deployments -- in the near future, and we are far below that level of commitment to fiber," the report says.

An anonymous reader quotes a report from ExtremeTech: MIT researchers have reportedly developed an algorithm that can accelerate [neural networks] by up to 200x. The NAS (Neural Architecture Search, in this context) algorithm they developed "can directly learn specialized convolutional neural networks (CNNs) for target hardware platforms -- when run on a massive image dataset -- in only 200 GPU hours," MIT News reports. This is a massive improvement over the 48,000 hours Google reported taking to develop a state-of-the-art NAS algorithm for image classification. The goal of the researchers is to democratize AI by allowing researchers to experiment with various aspects of CNN design without needing enormous GPU arrays to do the front-end work. If finding state of the art approaches requires 48,000 GPU arrays, precious few people, even at large institutions, will ever have the opportunity to try.

Algorithms produced by the new NAS were, on average, 1.8x faster than the CNNs tested on a mobile device with similar accuracy. The new algorithm leveraged techniques like path level binarization, which stores just one path at a time to reduce memory consumption by an order of magnitude. MIT doesn't actually link out to specific research reports, but from a bit of Google sleuthing, the referenced articles appear to be here and here -- two different research reports from an overlapping group of researchers. The teams focused on pruning entire potential paths for CNNs to use, evaluating each in turn. Lower probability paths are successively pruned away, leaving the final, best-case path. The new model incorporated other improvements as well. Architectures were checked against hardware platforms for latency when evaluated. In some cases, their model predicted superior performance for platforms that had been dismissed as inefficient. For example, 7x7 filters for image classification are typically not used, because they're quite computationally expensive -- but the research team found that these actually worked well for GPUs.

Devin Nunes, R-Calif., escalated the feud between conservatives and Twitter earlier this week with a lawsuit accusing the company of defamation and negligence -- two different allegations, one of which poses a more serious question for the social media platform and technology companies in general. Nunes is claiming that Twitter negligently violated its terms of service when it allowed people onto its online "premises" to say false or disparaging things about him. He is seeking $250 million in damages due to "pain, insult, embarrassment, humiliation, emotional distress and mental suffering, and injury to [Nunes'] personal and professional reputations" brought on by what Twitter users said about him. From a report: Defamation is an interesting legal matter to discuss, at least in theory, but suing for defamation is seldom profitable in reality. Negligence may not sound as exciting as defamation, but this theory of liability quietly drives most successful civil litigation. Relatively easy to prove, it generally requires that the defendant show conduct that came up short of what can be expected, and that this shortcoming caused the plaintiff's damages. [...] The primary reason that technology companies are not sued into oblivion is the existence of the Communications Decency Act, or CDA, and in particular Section 230, which states that providers of an interactive computer service shall not be treated as the publisher or speaker of any information provided by another information content provider. Ordinarily, a lawsuit like this is properly filed against the Twitter user or account (like "Devin Nunes' Mom") and not Twitter itself.

Section 230 and the CDA have become the targets of growing backlash against the idea that technology companies should not be held responsible for what is published on their platforms. Technology companies have voluntarily taken steps to moderate some content, such as extremism, conspiracy theories and fake news, but most personal insults and parodies are still allowed to flourish. Section 230, however, isn't necessarily bulletproof. At least one federal court has stressed that the statute does not "create a lawless no-man's-land on the internet." That provides some basis for Nunes' claim that Twitter has been negligent in keeping its platform from being used to spread damaging statements about him. But a negligence claim against Twitter may still be precluded by the CDA. The test is whether the cause of action requires the court to treat Twitter as the publisher or speaker of content provided by another. In the meantime, one of the Twitter parody accounts that is mocking Nunes -- Devin Nunes' Cow (@DevinCow) -- has gained a lot of attention, with its followers count jumping from about 1200 followers last week to more than 615,000 followers -- and in doing so, surpassed the number of followers Devin Nunes has (about 399k).

A group of researchers claims to have built a prototype for an "online polygraph" that uses machine learning to detect deception from text alone. But as a few machine learning academics point out, what these researchers have actually demonstrated is the inherent danger of overblown machine learning claims. From a report: When Wired showed the study to a few academics and machine learning experts, they responded with deep skepticism. Not only does the study not necessarily serve as the basis of any kind of reliable truth-telling algorithm, it makes potentially dangerous claims: A text-based "online polygraph" that's faulty, they warn, could have far worse social and ethical implications if adopted than leaving those determinations up to human judgment.

"It's an eye-catching result. But when we're dealing with humans, we have to be extra careful, especially when the implications of whether someone's lying could lead to conviction, censorship, the loss of a job," says Jevin West, a professor at the Information School at the University of Washington and a noted critic of machine learning hype. "When people think the technology has these abilities, the implications are bigger than a study."

An anonymous reader shares a report: Tez. Trendalyzer. Panoramio. Timeful. Bump! SlickLogin. BufferBox. The names sound like a mix of mid-2000s blogs and startups you'd see onstage at TechCrunch Disrupt!. In fact, they are just some of the many, many products that Google has acquired or created -- then killed.

While Google is notorious for eliminating underperforming products -- because even though these products often don't cost much for ongoing operations, they can pose a serious legal liability for the company -- it's rare to hear them spoken of after they've been shuttered. In fact, Killed By Google is the first website to memorialize them all in one place. Created by front-end developer Cody Ogden, the site features a tombstone and epitaph for each product the company has killed since it originated.

AT&T's "5G Evolution" cellular service isn't just controversial and arguably misleading, it's also slower than the 4G speeds offered by rivals T-Mobile and Verizon, according to a new report today from OpenSignal. From a report: Over a one-month period spanning January 28 through February 26, OpenSignal compared the average performance of "5G E capable" phones and "all others" on AT&T's network with similarly equipped devices on T-Mobile's, Verizon's, and Sprint's networks -- a fair test in that all four of the major U.S. carriers have deployed pre-5G, late-stage 4G technologies across the country. Only Sprint's network fell behind AT&T's performance, though that's no surprise, as the fourth-place carrier's network has lagged behind its rivals in performance for years.

Up to 95% of all reported trading in bitcoin is artificially created by unregulated exchanges, according to a new study [PDF], raising fresh doubts about the nascent market following a steep decline in prices over the past year. From a report: Fraudulent trading volume has dogged cryptocurrency trading for years, but the extent of the market manipulation has been difficult to determine. Bitwise Asset Management said its analysis of trading activity at 81 exchanges over four days in March indicates that the actual market for bitcoin is far smaller than previously thought. The San Francisco-based company submitted its research to the U.S. Securities and Exchange Commission with an application to launch a bitcoin-based exchange-traded fund.

The study, made public Thursday, is an attempt to alleviate the agency's longstanding concerns that a bitcoin ETF would leave investors exposed to fraud and market manipulation. Bitwise's fund, if approved, would be based upon the 5% of trading it considers legitimate, said Matthew Hougan, Bitwise's head of global research. That volume comes from 10 regulated exchanges that can verify that their trading data and customers are real. This slice of the market, he said, is well regulated, transparent and efficient. "I hope everyone sees there is a real market for bitcoin," he said.

An anonymous reader quotes a report from The Guardian: Facebook employees were aware of concerns about "improper data-gathering practices" by Cambridge Analytica months before the Guardian first reported, in December 2015, that the political consultancy had obtained data on millions from an academic. The concerns appeared in a court filing by the attorney general for Washington DC and were subsequently confirmed by Facebook. The new information "could suggest that Facebook has consistently mislead [sic]" British lawmakers "about what it knew and when about Cambridge Analytica," tweeted Damian Collins, the chair of the House of Commons digital culture media and sport select committee (DCMS) in response to the court filing.

In a statement, a company spokesperson said: "Facebook absolutely did not mislead anyone about this timeline." After publication of this article, the spokesperson acknowledged that Facebook employees heard rumors of data scraping by Cambridge Analytica in September 2015. The spokesperson said that this was a "different incident" from Cambridge Analytica's acquisition of a trove of data about as many as 87 million users that has been widely reported on for the past year. "In September 2015 employees heard speculation that Cambridge Analytica was scraping data, something that is unfortunately common for any internet service," the spokesperson said. "In December 2015, we first learned through media reports that Kogan sold data to Cambridge Analytica, and we took action. Those were two different things." The filing raised questions about when Facebook first learned about the misuse of personal data by Cambridge Analytica, the now defunct political consultancy.

The recent Boeing 737 MAX crashes involving an Ethiopian Airlines flight and a Lion Air flight may have been a result of two missing safety features that Boeing charged airlines extra for (Warning: source may be paywalled; alternative source). The New York Times reports that many low-cost carriers like Indonesia's Lion Air opted not to buy them so they could save money, even though some of these systems are fundamental to the plane's operations. "Now, in the wake of the two deadly crashes involving the same jet model, Boeing will make one of those safety features standard as part of a fix to get the planes in the air again," the report says. From the report: It is not yet known what caused the crashes of Ethiopian Airlines Flight 302 on March 10 and Lion Air Flight 610 five months earlier, both after erratic takeoffs. But investigators are looking at whether a new software system added to avoid stalls in Boeing's 737 Max series may have been partly to blame. Faulty data from sensors on the Lion Air plane may have caused the system, known as MCAS, to malfunction, authorities investigating that crash suspect.

The jet's software system takes readings from one of two vanelike devices called angle of attack sensors that determine how much the plane's nose is pointing up or down relative to oncoming air. When MCAS detects that the plane is pointing up at a dangerous angle, it can automatically push down the nose of the plane in an effort to prevent the plane from stalling. Boeing's optional safety features, in part, could have helped the pilots detect any erroneous readings. One of the optional upgrades, the angle of attack indicator, displays the readings of the two sensors. The other, called a disagree light, is activated if those sensors are at odds with one another. The angle of attack indicator will remain an option that airlines can buy. Neither feature was mandated by the Federal Aviation Administration. All 737 Max jets have been grounded. "Boeing will soon update the MCAS software, and will also make the disagree light standard on all new 737 Max planes," the report adds, citing a person familiar with the changes. "Boeing started moving on the software fix and the equipment change before the crash in Ethiopia."

Slashdot reader Futurepower(R) adds to the story: The FBI has joined the criminal investigation into the certification of the Boeing 737 MAX, lending its considerable resources to an inquiry already being conducted by U.S. Department of Transportation agents, according to people familiar with the matter. "The federal grand jury investigation, based in Washington, D.C., is looking into the certification process that approved the safety of the new Boeing plane, two of which have crashed since October.

The Homeland Security Department has issued an alert Thursday describing two types of computer-hacking vulnerabilities in 16 different models of Medtronic implantable defibrillators sold around the world, including some still on the market today. The vulnerability also affects bedside monitors that read data from the devices in patients' homes and in-office programming computers used by doctors. From the report: Medtronic recommends that patients only use bedside monitors obtained from a doctor or from Medtronic directly, and to keep it plugged in so it can receive software updates, and that they maintain "good physical control" over the monitor. Implantable defibrillators are complex, battery-run computers implanted in patients' upper chests to monitor the heart and send electric pulses or high-voltage shocks to prevent sudden cardiac death and treat abnormal heart beats. The vulnerabilities announced Thursday do not affect Medtronic pacemakers.

The more serious of the two is a vulnerability that could allow improper access to data sent between a defibrillator and an external device like an at-home monitor. The system doesn't use formal authentication or authorization protections, which means an attacker with short-range access to the device could inject or modify data and change device settings, the advisory says. A second vulnerability allows an attacker to read sensitive data streaming out of the device, which could include the patient's name and past health data stored on their device. The system does not use data encryption, the advisory says. (Deploying encryption in medical devices is tricky because is increases computational complexity and therefore uses the battery faster.) The FDA isn't expected to issue a recall as the vulnerabilities are expected to be patched via a future software update.

Since 2016, Microsoft has been working with the University of Washington to develop the first device to automatically encode digital information into DNA and back to bits again. "So far, DNA storage has been carried out by hand in the lab," reports MIT Technology Review. But now Microsoft and researchers at the University of Washington "say they created a machine that converts electronic bits to DNA and back without a person involved." From the report: The gadget, made from about $10,000 in parts, uses glass bottles of chemicals to build DNA strands, and a tiny sequencing machine from Oxford Nanopore to read them out again. According to a publication on March 21 in the journal Nature Scientific Reports, the team was able to store and retrieve just a single word -- "hello" -- or five bytes of data. What's more, the process took 21 hours, mostly because of the slow chemical reactions involved in writing DNA. While the team considered that a success for their prototype, a commercially useful DNA storage system would have to store data millions of times faster.

An anonymous reader quotes a report from Motherboard: Microsoft this week was the latest to highlight the U.S. government's terrible broadband mapping in a filing with the FCC, first spotted by journalist Wendy Davis. In it, Microsoft accuses the FCC of over-stating actual broadband availability and urges the agency to do better. "The Commission's broadband availability data, which underpins FCC Form 477 and the Commission's annual Section 706 report, appears to overstate the extent to which broadband is actually available throughout the nation," Microsoft said in the filing. "For example, in some areas the Commission's broadband availability data suggests that ISPs have reported significant broadband availability (25 Mbps down/3 Mbps up) while Microsoft's usage data indicates that only a small percentage of consumers actually access the Internet at broadband speeds in those areas," Microsoft said.

Similar criticism has long plagued the agency. The FCC's broadband data is received via the form 477 data collected from ISPs. But ISPs have a vested interest in over-stating broadband availability to obscure the sector's competition problems, and the FCC historically hasn't worked very hard to independently verify whether this data is truly accurate. The FCC's methodology has long been criticized as well. As it currently stands, the agency declares an entire ZIP code as "served" with broadband if just one home in an entire census block has it. In its filing, Microsoft "suggested that the Commission's ongoing effort to more accurately measure broadband could be improved by drawing on the FCC's subscription data, along with other broadband data sets from third-parties such as Microsoft, to complement survey data submitted under the current rules."

Tesla is suing a former engineer at the company, claiming he copied the source code for its Autopilot technology before joining a Chinese self-driving car startup in January. Reuters reports: The engineer, Guangzhi Cao, copied more than 300,000 files related to Autopilot source code as he prepared to join China's Xiaopeng Motors Technology Company Ltd, the Silicon Valley carmaker said in the lawsuit filed in a California court. Separately, Tesla lawyers on Wednesday filed a lawsuit against four former employees and U.S. self-driving car startup Zoox Inc, alleging the employees stole proprietary information and trade secrets for developing warehousing, logistics and inventory control operations. The Verge reported on the lawsuit filed against Cao: Tesla says that last year, Cao started uploading "complete copies of Tesla's Autopilot-related source code" to his iCloud account. The company claims he ultimately moved more than 300,000 files and directories related to Autopilot. After accepting a job with XPeng at the end of last year, Tesla says Cao deleted 120,000 files off his work computer and disconnected his personal iCloud account, and then "repeatedly logged into Tesla's secure networks" to clear his browser history before his last day with the company. Tesla also claims Cao recruited another Autopilot employee to XPeng in February. Tesla claims that it gives XPeng "unfettered access" to Autopilot: "Absent immediate relief, Tesla believes Cao and his new employer, [XPeng], will continue to have unfettered access to Tesla's marquee technology, the product of more than five years' work and over hundreds of millions of dollars of investment, which they have no legal right to possess," the company's lawyers write.

chicksdaddy writes: The grandson of Theo Brown, a legendary engineer and inventor for John Deere who patented, among other things, the manure spreader is calling out the company his grandfather served for decades for its opposition to right to repair legislation being considered in Illinois. In an opinion piece published by The Security Ledger entitled "My Grandfather's John Deere would support Our Right to Repair," Willie Cade notes that his grandfather, Theophilus Brown is credited with 158 patents, some 70% of them for Deere & Co., including the manure spreader in 1915. His grandfather used to travel the country to meet with Deere customers and see his creations at work in the field. His hope, Cade said, was to help the company's customers be more efficient and improve their lives with his inventions.

In contrast, Cade said the John Deere of the 21st Century engages in a very different kind of business model: imposing needless costs on their customers. An example of this kind of rent seeking is using software locks and other barriers to repair -- such as refusing to sell replacement parts -- in order to force customers to use authorized John Deere technicians to do repairs at considerably higher cost and hassle. "It undermines what my grandfather was all about," he writes. Cade, who founded the Electronics Reuse Conference, is supporting right to repair legislation that is being considered in Illinois and opposed by John Deere and the industry groups it backs. "Farmers who can't repair farm equipment and a wide spectrum of Americans who can't repair their smartphones are pushing back in states across the country."

An anonymous reader quotes a report from Ars Technica: Cable lobbyists don't want to be called cable lobbyists anymore. The nation's top two cable industry lobby groups have both dropped the word "cable" from their names. But the lobby groups' core mission -- the fight against regulation of cable networks -- remains unchanged. The National Cable & Telecommunications Association (NCTA) got things started in 2016 when it renamed itself NCTA-The Internet & Television Association, keeping the initialism but dropping the words it stood for. The group was also known as the National Cable Television Association between 1968 and 2001. The American Cable Association (ACA) is the nation's other major cable lobby. While NCTA represents the biggest companies like Comcast and Charter, the ACA represents small and mid-size cable operators. Today, the ACA announced that it is now called America's Communications Association or "ACA Connects," though the ACA's website still uses the americancable.org domain name.

"The new name reflects a leading position for the association in the fast-growing telecommunications industry, where technology is rapidly changing how information is provided to and used by consumers," the cable lobby said. "It's all about the communications and connections our members provide," said cable lobbyist Matthew Polka, who is CEO of the ACA. The "ACA Connects" moniker "explains what our association and members really do," Polka continued. "We connect, communicate, build relationships and work together with all, and that will never change."

The battle between PewDiePie, currently the most subscribed channel on YouTube, and T-Series, an Indian music label, continues to have strange repercussions. In recent months, as T-Series closes in on the gap to beat PewDiePie for the crown of the most subscribers on YouTube, alleged supporters of PewDiePie, in an unusual show of love, have hacked Chromecasts and printers to persuade victims to subscribe to PewDiePie's channel. Now ZDNet reports about a second strain of ransomware that is linked to PewDiePie. From the report: A second one appeared in January, and this was actually a fully functional ransomware strain. Called PewCrypt, this ransomware was coded in Java, and it encrypted users' files in the "proper" way, with a method of recovering files at a later date. The catch --you couldn't buy a decryption key, but instead, victims had to wait until PewDiePie gained over 100 million followers before being allowed to decrypt any of the encrypted files. At the time of writing, PewDiePie had around 90 million fans, meaning any victim would be in for a long wait before they could regain access to any of their files. Making matters worse, if T-Series got to 100 million subscribers before PewDiePie, then PewCrypt would delete the user's encryption key for good, leaving users without a way to recover their data.

While the ransomware was put together as a joke, sadly, it did infect a few users, ZDNet has learned. Its author eventually realized the world of trouble he'd get into if any of those victims filed complaints with authorities, and released the ransomware's source code on GitHub, along with a command-line-based decryption tool.

HMD Global, the Finnish company that sublicensed the Nokia smartphone brand from Microsoft, is under investigation in Finland for collecting and sending some phone owners' information to a server located in China. From a report: In a statement to Finnish newspaper Helsingin Sanomat, the company blamed the data collection on a coding mistake during which an "activation package" was accidentally included in some phones' firmware. HMD Global said that only a single batch of Nokia 7 Plus devices were impacted and included this package. The data collection was exposed today in an investigation published by Norwegian broadcaster NRK, which learned of it from a user's tip. According to NRK, affected Nokia phones collected user data every time the devices were turned on, unlocked, or the screen was revived from a sleep state. Collected data included the phone's GPS coordinates, network information, phone serial number, and SIM card number.

US companies are going to keep hiring foreign tech workers, even as the Trump administration makes doing so more difficult. For a number of US companies that means expanding their operations in Canada, where hiring foreign nationals is much easier. From a report: Demand for international workers remained high this year, according to a new Envoy Global survey of more than 400 US hiring professionals, who represent big and small US companies and have all had experience hiring foreign employees. Some 80 percent of employers expect their foreign worker headcount to either increase or stay the same in 2019, according to Envoy, which helps US companies navigate immigration laws. That tracks with US government immigration data, which shows a growing number of applicants for high-skilled tech visas, known as H-1Bs, despite stricter policies toward immigration. H-1B recipients are all backed by US companies that say they are in need of specialized labor that isn't readily available in the US -- which, in practice, includes a lot of tech workers. Major US tech companies, including Google, Facebook, and Amazon, have all been advocating for quicker and more generous high-skilled immigration policies. To do so they've increased lobbying spending on immigration.

Hundreds of millions of Facebook users had their account passwords stored in plain text and searchable by thousands of Facebook employees -- in some cases going back to 2012, KrebsOnSecurity reported Thursday. From the report: Facebook says an ongoing investigation has so far found no indication that employees have abused access to this data. Facebook is probing the causes of a series of security failures in which employees built applications that logged unencrypted password data for Facebook users and stored it in plain text on internal company servers. That's according to a senior Facebook employee who is familiar with the investigation and who spoke on condition of anonymity because they were not authorized to speak to the press. The Facebook source said the investigation so far indicates between 200 million and 600 million Facebook users may have had their account passwords stored in plain text and searchable by more than 20,000 Facebook employees. The source said Facebook is still trying to determine how many passwords were exposed and for how long, but so far the inquiry has uncovered archives with plain text user passwords in them dating back to 2012. Facebook has responded.