PLEASE NOTE: The Overspill will be on holiday next week. So you’re unlikely to receive any emails/see any posts here.
(Why in bold red? Because I know some people will miss this. Next time I might bring back the <blink> tag for the message.)
A selection of 11 links for you. That’s the way it goes. I’m @charlesarthur on Twitter. Observations and links welcome.
The video-streaming service, which is owned by Google, said it would ban videos that promote either the construction or sale of firearms and their accessories. The new policy, developed with expert advice over the past four months, will go into effect next month, it said.
“While we’ve long prohibited the sale of firearms, we recently notified creators of updates we will be making around content promoting the sale or manufacture of firearms and their accessories, specifically, items like ammunition, gatling triggers, and drop-in auto sears,” YouTube said in a statement.
YouTube, which described the move as part of “regular changes” to policy, notified users in a Monday forum post. The company had previously banned videos showing how to make firearms discharge faster, a technique used by the gunman who killed 58 people in Las Vegas last fall.
The announcement comes days before planned student-led protests against gun violence on Saturday. It was met with frustration from gun rights advocates.
“Much like Facebook, YouTube now acts as a virtual public square,” the National Shooting Sports Foundation, a private group representing gun makers, said in a statement. “The exercise of what amounts to censorship, then, can legitimately be viewed as the stifling of commercial free speech, which has constitutional protection. Such actions also impinge on the Second Amendment.”
It’s not a stifling of commercial free speech (Google owns the platform; it gets to decide what’s on it) and it really doesn’t impinge on the Second Amendment. It’s not stopping anyone buying or owning a gun. Reason is a stranger to some.
link to this extract
Best Buy Co., the large consumer electronics retailer, plans to sever ties with Chinese phone maker Huawei amid U.S. government criticism of the phone maker, according to people familiar with the matter.
The U.S. retail giant will stop selling all Huawei phones, laptops, and smartwatches in the coming weeks, they said. In addition, Best Buy won’t sell phones under the Honor brand, a Huawei subsidiary that was supposed to help the Chinese phone maker sell in lower-cost smartphone markets globally, including in the U.S.
Best Buy follows U.S. mobile-phone carriers AT&T Inc. and Verizon Communications Inc. in distancing themselves from Huawei, which has come under scrutiny by U.S. officials concerned about whether the company is too closely affiliated with the Chinese government.
After the top two U.S. carriers decided not to go forward with Huawei devices, the company began selling its latest phone, the Mate 10 Pro, directly to consumers, through outlets like Best Buy and Amazon.com Inc. The device is still available on Best Buy’s website, but the retail giant won’t purchase new supply from Huawei and will stop selling the phone in the coming weeks, said the people, who asked not to be identified because the retailer’s action isn’t yet public.
Bannon oversaw Cambridge Analytica’s collection of Facebook data, says former employee • The Washington Post
The data and analyses that Cambridge Analytica generated in this time provided discoveries that would later form the emotionally charged core of Trump’s presidential platform, said Wylie, whose disclosures in news reports over the past several days have rocked both his onetime employer and Facebook.
“Trump wasn’t in our consciousness at that moment; this was well before he became a thing,” Wylie said. “He wasn’t a client or anything.”
The year before Trump announced his presidential bid, the data firm already had found a high level of alienation among young, white Americans with a conservative bent.
In focus groups arranged to test messages for the 2014 midterms, these voters responded to calls for building a new wall to block the entry of illegal immigrants, to reforms intended to “drain the swamp” of Washington’s entrenched political community and to thinly veiled forms of racism toward African Americans called “race realism,” he recounted.
The firm also tested views of Russian President Vladimir Putin.
“The only foreign thing we tested was Putin,” he said. “It turns out, there’s a lot of Americans who really like this idea of a really strong authoritarian leader and people were quite defensive in focus groups of Putin’s invasion of Crimea.”
Warden was CTO at a company called Jetpac, which did some amazing deep learning stuff on Instagram photos and then on-device recognition of photo contents. Then Google bought Jetpac and now he’s shoulder-deep in machine learning stuff there:
In many real-world cases, the researcher won’t have made notes or remember exactly what she did, so even she won’t be able to reproduce the model. Even if she can, the frameworks the model code depend on can change over time, sometimes radically, so she’d need to also snapshot the whole system she was using to ensure that things work. I’ve found ML researchers to be incredibly generous with their time when I’ve contacted them for help reproducing model results, but it’s often months-long task even with assistance from the original author.
Why does this all matter? I’ve had several friends contact me about their struggles reproducing published models as baselines for their own papers. If they can’t get the same accuracy that the original authors did, how can they tell if their new approach is an improvement? It’s also clearly concerning to rely on models in production systems if you don’t have a way of rebuilding them to cope with changed requirements or platforms. At that point your model moves from being a high-interest credit card of technical debt to something more like what a loan-shark offers. It’s also stifling for research experimentation; since making changes to code or training data can be hard to roll back it’s a lot more risky to try different variations, just like coding without source control raises the cost of experimenting with changes.
It’s not all doom and gloom, there are some notable efforts around reproducibility happening in the community. One of my favorites is the TensorFlow Benchmarks project Toby Boyd’s leading. He’s made it his team’s mission not only to lay out exactly how to train some of the leading models from scratch with high training speed on a lot of different platforms, but also ensures that the models train to the expected accuracy. I’ve seen him sweat blood trying to get models up to that precision, since variations in any of the steps I listed above can affect the results and there’s no easy way to debug what the underlying cause is, even with help from the authors. It’s also a never-ending job, since changes in TensorFlow, in GPU drivers, or even datasets, can all hurt accuracy in subtle ways.
Alphabet Inc.’s Google will ask web publishers to obtain consent on its behalf to gather personal information on European users and target ads at them using Google’s systems, according to people familiar with the matter, part of a plan to comply with a coming data-privacy law in Europe.
Under the European Union’s forthcoming General Data Protection Regulation, which goes into effect on May 25, global companies will be required to obtain consent from European users to gather their personal information in many cases, and be more transparent about the data they collect and how it is used.
Companies found in violation of the sweeping regulation, known as GDPR, will face fines of up to 4% of their annual global revenue. Google is poised to announce its steps toward compliance for its ad-technology platforms as early as this week, the people familiar with the matter said.
The company will be gathering consent from users itself for data-usage on its own properties such as Google.com, Gmail and YouTube. But when it comes to third-party websites and apps that use Google’s ad technology to sell ads, the tech giant wants those publishers to be responsible for obtaining consent…
…It’s important for Google to get its GDPR strategy right. In January, Deutsche Bank analyst Lloyd Walmsley wrote in a research note that the GDPR could trim Google’s global revenue by 2 percentage points, should 30% of European users opt-out of some data sharing.
“GDPR is on the minds of most of us in the industry,” Sridhar Ramaswamy, Google’s senior vice president of ads and commerce, said Wednesday on stage at an ad-industry conference in London.
Google hasn’t yet briefed many publishers on its forthcoming plans. But people with some knowledge of Google’s plans said publishers might be concerned that by mandating through policy that publishers obtain consent on its behalf, Google is seeking different treatment from publishers’ other ad tech partners.
We’ve heard it can feel unexpected when your feed refreshes and automatically bumps you to the top. So today we’re testing a “New Posts” button that lets you choose when you want to refresh, rather than it happening automatically. Tap the button and you’ll be taken to new posts at the top of feed — don’t tap, and you’ll stay where you are. We hope this makes browsing Instagram much more enjoyable.
Based on your feedback, we’re also making changes to ensure that newer posts are more likely to appear first in feed. With these changes, your feed will feel more fresh, and you won’t miss the moments you care about. So if your best friend shares a selfie from her vacation in Australia, it will be waiting for you when you wake up.
What would make browsing Instagram much more enjoyable would be if posts appeared in reverse chronological order, newest at the top, always.
This is a step towards that; maybe if enough people mash that button, then they’ll move to a time-based timeline.
link to this extract
Above I have included a brightened frame from 3 seconds into the video. It is the first frame in which the white running shoes of the victim are visible in the dashcam video. They only appear then because she is previously in darkness, crossing at a poorly lit spot, and the headlamps finally illuminate her. Impact occurs at about 4.4 seconds (if the time on the video is right.)
She is crossing, we now see, at exactly this spot where two storm drains are found in the curb. It is opposite the paved path in the median which is marked by the signs telling pedestrians not to cross at this location. She is walking at a moderate pace.
The road is empty of other cars. Here are the big issues:
• On this empty road, the LIDAR is very capable of detecting her. If it was operating, there is no way that it did not detect her 3 to 4 seconds before the impact, if not earlier. She would have come into range just over 5 seconds before impact.
• On the dash-cam style video, we only see her 1.5 seconds before impact. However, the human eye and quality cameras have a much better dynamic range than this video, and should have also been able to see her even before 5 seconds. From just the dash-cam video, no human could brake in time with just 1.5 seconds warning. The best humans react in just under a second, many take 1.5 to 2.5 seconds.
• The human safety driver did not see her because she was not looking at the road. She seems to spend most of the time before the accident looking down to her right, in a style that suggests looking at a phone.
• While a basic radar which filters out objects which are not moving towards the car would not necessarily see her, a more advanced radar also should have detected her and her bicycle (though triggered no braking) as soon as she entered the lane to the left, probably 4 seconds before impact at least. Braking could trigger 2 seconds before, in theory enough time.)
To be clear, while the car had the right-of-way and the victim was clearly unwise to cross there, especially without checking regularly in the direction of traffic, this is a situation where any properly operating robocar following “good practices,” let alone “best practices,” should have avoided the accident regardless of pedestrian error.
The videos (external view, interior view) are alarming, and disturbing. The lighting is terrible – though it’s hard to tell what a (driving) human would have seen; our eyes adapt to darkness in ways that cameras don’t.
But the LIDAR failure is astonishing. Google has described early self-driving tests where the SDC stopped in a forest because it detected a deer at the side of the road. This fatality could be due to LIDAR failure. But if that can happen without alarms going off, it’s just as bad, if not worse, than anything else.
link to this extract
Sure, Ikea’s ubiquitous instruction manuals look so simple and friendly, but translating the schematics from the page into real life can be challenging.
A designer named Adam Pickard has shown us a better way. He imagined that Ikea’s instructions were rendered in augmented reality–much like the company allows you to preview a couch in your living room today with its AR app.
Using 3D modeling and a bit of post-production trickery, he created a concept called AssembleAR. It’s a high fidelity vision for an app that could place Ikea’s wireframe build instructions right onto your living room floor. After scanning the barcode on the box, you could literally lay the step-by-step models right next to your actual built project.
In principle, this AR effect shouldn’t be all that much of an improvement over good old paper instructions. But in rendered reality, the little nuances, like animated bolts and screws twisting into place, seem like they could do wonders to eliminate those half-guess moments that seem so intrinsic to building a piece of furniture on your own.
Except it’s not definitely the manual of the future, is it? Not a great headline. But a nice use of AR.
link to this extract
Before former spy Sergei Skripal and his daughter Yulia collapsed on a park bench in Salisbury on 4 March, the only other person confirmed to suffer the effects of novichok was a young Soviet chemical weapons scientist.
“Circles appeared before my eyes: red and orange. A ringing in my ears, I caught my breath. And a sense of fear: like something was about to happen,” Andrei Zheleznyakov told the now-defunct newspaper Novoye Vremya, describing the 1987 weapons lab incident that exposed him to a nerve agent that would eventually kill him. “I sat down on a chair and told the guys: ‘It’s got me.’”
By 1992, when the interview was published, the nerve agent had gutted Zheleznyakov’s central nervous system. Less than a year later he was dead, after battling cirrhosis, toxic hepatitis, nerve damage and epilepsy.
But by deciding to go public, he joined those blowing the whistle on a chemical weapons programme that was still charging forward years after George Bush and Mikhail Gorbachev signed the 1990 US–Soviet Chemical Weapons Accord in which each pledged to halt the production of chemical weapons.
Despite Zheleznyakov’s role in creating a binary of a nerve agent believed to be more potent than the deadly VX nerve agent, he remains a hero to some.
“He gave all the information – I couldn’t do that at the time,” said Vil Mirzayanov, a chemical weapons scientist put on trial in Russia for first revealing the existence of the novichok programme, speaking to the Guardian at his home in Princeton, New Jersey. “He was not afraid because he knew his days were numbered.”
Zheleznyakov was never prosecuted, but he could not outrun the poison. He lost the ability to concentrate, Mirzayanov said, and eventually isolated himself.
He died in 1993 of a brain seizure while eating dinner, divorced and childless, largely disgruntled at the perceived indifference shown him by his superiors and journalists.
Russian officials continue to deny ever having such a programme.
Novichoks (it’s a class) are binary agents – you mix two relatively harmless substances together.
link to this extract
For years, executives at France-based Ledger have boasted their specialized hardware for storing cryptocurrencies is so securely designed that resellers or others in the supply chain can’t tamper with the devices without it being painfully obvious to end users. The reason: “cryptographic attestation” that uses unforgeable digital signatures to ensure that only authorized code runs on the hardware wallet.
“There is absolutely no way that an attacker could replace the firmware and make it pass attestation without knowing the Ledger private key,” officials said in 2015. Earlier this year, Ledger’s CTO said attestation was so foolproof that it was safe to buy his company’s devices on eBay.
On Tuesday, a 15-year-old from the UK proved these claims wrong. In a post published to his personal blog, Saleem Rashid demonstrated proof-of-concept code that had allowed him to backdoor the Ledger Nano S, a $100 hardware wallet that company marketers have said has sold by the millions. The stealth backdoor Rashid developed is a minuscule 300-bytes long and causes the device to generate pre-determined wallet addresses and recovery passwords known to the attacker. The attacker could then enter those passwords into a new Ledger hardware wallet to recover the private keys the old backdoored device stores for those addresses.
Using the same approach, attackers could perform a variety of other nefarious actions, including changing wallet destinations and amounts for payments so that, for instance, an intended $25 payment to an Ars Technica wallet would be changed to a $2,500 payment to a wallet belonging to the backdoor developer.
link to this extract
“The situation is catastrophic,” said Benoit Fontaine, a conservation biologist at France’s National Museum of Natural History and co-author of one of the studies.
“Our countryside is in the process of becoming a veritable desert,” he said in a communique released by the National Centre for Scientific Research (CNRS), which also contributed to the findings.
The common white throat, the ortolan bunting, the Eurasian skylark and other once-ubiquitous species have all fallen off by at least a third, according a detailed, annual census initiated at the start of the century.
A migratory song bird, the meadow pipit, has declined by nearly 70%.
The museum described the pace and extent of the wipe-out as “a level approaching an ecological catastrophe”.
The primary culprit, researchers speculate, is the intensive use of pesticides on vast tracts of monoculture crops, especially wheat and corn.
The problem is not that birds are being poisoned, but that the insects on which they depend for food have disappeared.
“There are hardly any insects left, that’s the number one problem,” said Vincent Bretagnolle, a CNRS ecologist at the Centre for Biological Studies in Chize.
Recent research, he noted, has uncovered similar trends across Europe, estimating that flying insects have declined by 80%, and bird populations has dropped by more than 400m in 30 years.
Despite a government plan to cut pesticide use in half by 2020, sales in France have climbed steadily, reaching more than 75,000 tonnes of active ingredient in 2014, according to European Union figures.
“What is really alarming, is that all the birds in an agricultural setting are declining at the same speed, even ’generalist’ birds,” which also thrive in other settings such as wooded areas, said Bretagnolle.
This has been going on silently for years: older readers might remember how car windscreens and radiator grilles would be covered in dead insects after long journeys in the past. Now? Hardly anything. It’s not because insects are getting better at dodging cars.
I hope this doesn’t turn out to be the most significant story I ever link to.
link to this extract
Errata, corrigenda and ai no corrida: none notified
You can sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.