Start Up No.1,043: Facebook’s AI death smarts, Amazon staff push for green scheme, how Russia trolls on vaccines, and more

Don’t want to alarm you, but scientists in China may have set us on this path. CC-licensed photo by James on Flickr.

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.

A selection of 12 links for you. Attack ships on fire off the shoulder of Orion. I’m @charlesarthur on Twitter. Observations and links welcome.

Facebook to use AI to stop telling users to say hi to dead friends • The Guardian

Alex Hern:


Facebook has promised to use artificial intelligence to stop suggesting users invite their dead friends to parties.

The site’s freshly emotionally intelligent AI is part of a rash of changes to how Facebook handles “memorialised” accounts – pages whose owner has been reported deceased, but that are kept on the social network in their memory.

Memorialisation of accounts allows for treasured images, videos and posts to be kept online, as well as providing a focal point for grieving friends and relatives to share memories.

But the feature has caused its fair share of pain: since the account is kept on the social network and treated similarly to any other Facebook user, it is used for the same algorithmic features as anything else. That means users have been sent recommendations to invite dead relatives to parties, suggestions to wish them a happy birthday, and more.

Facebook said that should be a thing of the past. “Once an account is memorialised, we use AI to help keep the profile from showing up in places that might cause distress, like recommending that person be invited to events or sending a birthday reminder to their friends,” wrote Facebook’s chief operating officer, Sheryl Sandberg, in a blogpost published on Tuesday. “We’re working to get better and faster at this.”


But after that was written late on Tuesday, Facebook tweaked Sandberg’s blogpost:


The post now says that AI will be used, not to keep memorialised accounts from showing up in algorithmic features, but to keep accounts that haven’t been memorialised from doing so. “If an account hasn’t yet been memorialized, we use AI to help keep it from showing up in places that might cause distress,” Sandberg is now quoted as saying. In other words, Facebook’s AI is being deployed to work out which of its users are dead, rather than waiting to be told.


Given the average mortality rate of 8.1 per thousand per year, and 2 billion users, about 16.2 million users die every year – that’s about 45,000 per day. That AI can’t come too soon.

unique link to this extract

Boeing 737 MAX crash and the rejection of ridiculous data • Philip Greenspun’s Weblog


“Boeing 737 Max: What went wrong?” (BBC) contains a plot showing the angle of attack data being fed to Boeing’s MCAS software. Less than one minute into the flight, the left sensor spikes to an absurd roughly 70-degree angle of attack. Given the weight of an airliner, the abruptness of the change was impossible due to inertia. But to have avoided killing everyone on board, the software would not have needed a “how fast is this changing?” capability. It would simply have needed a few extra characters in an IF statement.

Had the systems engineers and programmers checked Wikipedia, for example, (or maybe even their own web site) they would have learned that “The critical or stalling angle of attack is typically around 15° – 20° for many airfoils.” Beyond 25 degrees, therefore, it is either sensor error or the plane is stalling/spinning and something more than a slow trim is going to be required.

So, even without checking the left and right AOA sensors against each other (what previous and conventional stick pusher designs have done), all of the problems could potentially have been avoided…


As Greenspun points out, it would just be a little extra logic – if a reading is wildly impossible, then reality hasn’t shifted; the sensor is wrong. The logic presently tends to assume the sensor is right.
unique link to this extract

Chinese scientists have put human brain genes in monkeys—and yes, they may be smarter • MIT Technology Review

Antonio Regalado:


scientists in southern China report that they’ve tried to narrow the evolutionary gap, creating several transgenic macaque monkeys with extra copies of a human gene suspected of playing a role in shaping human intelligence.

“This was the first attempt to understand the evolution of human cognition using a transgenic monkey model,” says Bing Su, the geneticist at the Kunming Institute of Zoology who led the effort.

According to their findings, the modified monkeys did better on a memory test involving colors and block pictures, and their brains also took longer to develop—as those of human children do. There wasn’t a difference in brain size.

The experiments, described on March 27 in a Beijing journal, National Science Review, and first reported by Chinese media, remain far from pinpointing the secrets of the human mind or leading to an uprising of brainy primates.


Is the Statue of Liberty still in place? Just checking.
unique link to this extract

Public letter to Jeff Bezos and the Amazon Board of Directors • Medium

Amazon Employees for Climate Justice:


We, the undersigned 4,269* Amazon employees, ask that you adopt the climate plan shareholder resolution and release a company-wide climate plan that incorporates the principles outlined in this letter.

Amazon has the resources and scale to spark the world’s imagination and redefine what is possible and necessary to address the climate crisis. We believe this is a historic opportunity for Amazon to stand with employees and signal to the world that we’re ready to be a climate leader.
Climate change is an existential threat. The 2018 Intergovernmental Panel on Climate Change (IPCC) report predicts that a warming of 2° Celsius, which we’re currently on track to surpass, will threaten the lives of hundreds of millions of people and put thousands of species at risk of extinction.

We’re already seeing devastating climate impacts: unprecedented flooding in India and Mozambique, dry water wells in Africa, coastal displacement in Asia, wildfires and floods in North America, and crop failure in Latin America. Vulnerable communities least responsible for the climate crisis are already paying the highest price.


A six-point manifesto, including transition away from fossil fuels, prioritise climate impact, and also stop donating to climate-denying legislators. Let’s hope this is heard.
unique link to this extract

How Russia sows confusion in the US vaccine debate • Foreign Policy

Katherine Kirk:


The existence of a Russian disinformation campaign that could make Americans hesitant to vaccinate their children highlights something important about the Kremlin’s information war on the United States. Moscow’s goal has never been to advantage Republicans or Democrats. Instead, it is after a far bigger prize: the exacerbation of Americans’ distrust of one another and, in turn, the erosion of their confidence in society and the U.S. government.

A recent study from David Broniatowski, a professor at George Washington University, and his co-authors found that thousands of Russian accounts used to spread disinformation had seized on anti-vaccine messaging.

After combing through nearly 2 million tweets recorded between 2014 and 2017, the researchers found that Russian troll accounts were significantly more likely to tweet about vaccination than general Twitter users. They had turned to vaccines as a wedge issue in an effort to ramp up social discord, erode trust in public health institutions, and exacerbate fear and division in the United States.

Three tweets from the study go a long way toward capturing the style of this disorienting campaign. In one round of keystrokes, a Russian-backed account lashed out: “#vaccines are a parent’s choice. Choice of a color of a little coffin #VaccinateUS.”In one round of keystrokes, a Russian-backed account lashed out: “#vaccines are a parent’s choice. Choice of a color of a little coffin #VaccinateUS.” Another went with: “Did you know there was a secret government database of #vaccine-damaged children? #VaccinateUS.”

Moving toward the opposite pole of the discussion, a Russian troll account tweeted: “Do you still treat your kids with leaves? No? And why don’t you #vaccinate them? Its medicine! #VaccinateUS.” The study suggested that by giving the illusion of a grassroots debate, complete with content pushing both for and against vaccination, Russia could better tap into the fears and divisions among Americans—and exploit them.


It’s easy to suggest that social media is the problem, but it’s credulity and lack of education that’s the real problem.
unique link to this extract

China spying: the internet’s underwater cables are next • Bloomberg

James Stavridis:


Just as the experts are justifiably concerned about the inclusion of espionage “back doors” in Huawei’s 5G technology, western intelligence professionals oppose the company’s engagement in the undersea version, which provides a much bigger bang for the buck because so much data rides on so few cables.

Naturally, Huawei denies any manipulation of the cable sets it is constructing, even though the US and other nations say it is obligated by Chinese law to hand over network data to the government. The US last year restricted federal agencies using from using its 5G equipment; Huawei responded with a lawsuit in federal court. Washington is pressuring its allies to follow its lead — the American ambassador to Germany warned that allowing Chinese companies into its 5G project would mean reduced security cooperation from the US — but this is an uphill battle. Most nations and companies feel that better cell phone service is worth the security risks.

A similar dynamic is playing out underwater. How can the US address the security of undersea cables? There is no way to stop Huawei from building them, or to keep private owners from contracting with Chinese firms on modernizing them, based purely on suspicions. Rather, the US must use its cyber- and intelligence-gathering capability to gather hard evidence of back doors and other security risks. This will be challenging — the Chinese firms are technologically sophisticated and entwined with a virtual police state.


The US and UK are itchy about this because they know they can do it: there’s a lot of tapping of cables at the point where they leave the ocean. What if the cables have the tapping built in?
unique link to this extract

Ikea and Sonos made the ultimate speaker lamp • The Verge

Chris Welch:


eventually, Ikea’s smart home app will integrate music controls for the speakers, allowing customers to build scenes that factor in the lamp, Ikea’s smart blinds, and any bulbs you own. (Sonos speakers can already be controlled entirely within Spotify, iHeartRadio, and other apps, so to see that extending to Ikea isn’t a surprise, given the significance of this partnership.) Who needs an alarm clock when your bedroom lamp can blast some music and switch on the light come morning?

The bookshelf speaker is the smallest speaker that Sonos has ever made (and it’s smaller than what you might expect based on its name), so I’m not very hopeful that it’s going to offer blow-away sound. Spence more or less confirmed that the bookshelf speaker won’t pack quite the same bass punch as the company’s regular lineup.

But he also said that Sonos’ TruePlay feature, which tunes a speaker’s audio output based on its position in a room, makes a big difference in improving sound quality for both the lamp and bookshelf products. [Sonos CEO Patrick] Spence expressed confidence that both speakers will sound better than anything else at their respective price points. I wish the bookshelf speaker had a rechargeable battery inside — it seems like a good size to take outside during a party — but that’s not the case. Either way, like most things Ikea, I think the goal here was to make something good enough for the masses.


Really not sure what I make of the idea of a lamp speaker.
unique link to this extract

Black hole pictured for first time — in spectacular detail • Nature

Davide Castelvecchi:


The EHT ran another observing campaign in 2018 — the analysis of those data is still in the works — but cancelled a planned observation campaign this year because of security issues near one of its most important sites, the 50-metre LMT Large Millimeter Telescope in Puebla, Mexico. They plan to continue to do observations once a year starting in 2020.

The collaboration is now looking for funding to establish a foothold in Africa, which would fill in a major gap in the network. The plan is to relocate a 15-metre dish — a decommissioned Swedish telescope — from Chile to the Gamsberg Table Mountain in Namibia. For now, the network has already secured two major additions: a dish in Greenland and an array in the French Alps.

An expanded EHT network could provide detail on what happens inside the voids — “how the world behaves inside black holes, and if it is as we expected it to be”, says David Sánchez Argüelles, a physicist at the Large Millimeter Telescope.

“It was a great sense of relief to see this, but also surprise,” says Doeleman of the results. “You know what I was really expecting to see? A blob. To see this ring is probably the best outcome that we could have had.”


A planet-sized telescope. Visualising a black hole. Amazing.
unique link to this extract

Mysterious hackers hid their ‘Swiss Army’ spyware for five years • WIRED

Andy Greenberg:


Kaspersky says it first detected the TajMahal spyware framework last fall, on only a single victim’s network: The embassy of a Central Asian country whose nationality and location Kaspersky declines to name. But given the software’s sophistication, Shulmin says TajMahal has likely been deployed elsewhere. “It seems highly unlikely that such a huge investment would be undertaken for only one victim,” he writes. “This suggests that there are either further victims not yet identified, or additional versions of this malware in the wild, or possibly both.”

Those initial findings may indicate a very cautious and discreet state-sponsored intelligence-gathering operation, says Jake Williams, a former member of the National Security Agency’s elite Tailored Access Operations hacking group. “The extensibility of it requires a large developer team,” Williams notes. He points out also that the ability to avoid detection and the single known victim suggest extreme care in targeting, stealth, and operation security. “There’s all kinds of stuff here that screams opsec and very regimented tasking.”

Shulmin says Kaspersky hasn’t yet been able to connect TajMahal, named for a file the spyware uses to move stolen data off a victim’s machine, to any known hacker groups with the usual methods of code-matching, shared infrastructure, or familiar techniques. Its Central Asian target doesn’t exactly provide any easy clues as to the hackers’ identities either, given the vagueness of that description and the countries with sophisticated hacker teams with Central Asian interests, including China, Iran, Russia and the US. Nor has Kaspersky determined how the hackers behind TajMahal gain initial access to a victim network. But they do note that the group plants an initial backdoor program on machines, which the hackers labelled Tokyo.


“Central Asia” implies somewhere in the ambit of Russia and China to me. Could be US, could be Israel, could be China, could be Russia.
unique link to this extract

Does Donald Trump have dementia? We need to know – psychologist • USA Today

John Gartner is a former assistant professor at Johns Hopkins University School of Medicine:


In Alzheimer’s, as language skills deteriorate, we see two types of tell-tale speech disorders, or paraphasias.

“Semantic paraphasia” involves choosing the incorrect words. For instance, after Attorney General William Barr released a letter on the Mueller report, Trump said: “I hope they now go and take a  look at the oranges, the oranges of that investigation, the beginnings of that investigation.”

“Phonemic paraphasia”, which is linked to the moderate to severe stages of Alzheimer’s, is described as “the substitution of a word with a nonword that preserves at least half of the segments and/or number of syllables of the intended word.” For example, Israeli Prime Minister Benjamin Netanyahu becomes “Betanyahu,” big league becomes “bigly,” anonymous becomes “enenamas” or “anenomynous,” renovation becomes “renoversh,” missiles become “mishiz,” space capsule becomes “capsicle,” midterm elections become “midtowm” and “midturn” elections, and Christmas becomes “Chrissus.”

Trump’s speech patterns appear even more disordered when you go beyond the sound bite and look at a whole speech. He careens from one thought to the next in a parade of non sequiturs, frequently interrupting himself in the middle of a sentence to veer into another free association. When commentators described his two-hour  speech at the Conservative Political Action Conference last month as “unhinged,” they were referring in large part to this quality.


He’s always – or at least, in our public experience of him – caromed around from topic to topic, but it seems to be getting worse. He’s 72, and the pressures of the office are enormous, even allowing for time spent tweeting in bed. Don’t forget that Reagan certainly had Alzheimer’s in his second term. (Thanks Nic for the link.)
unique link to this extract

Video of Apple’s W.A.L.T. In action – the 1993-edition iPhone • Sonny Dickson

Sonny Dickson:


There has been a lot said about Apple’s development of the iPhone, and the history and inspiration behind the device – such as this notable 1983 concept of a “Telephone Mac.” One of the most notable examples of this is Apple’s lesser known desk phone known as the W.A.L.T. (Wizzy Active Lifestyle Telephone).

The W.A.L.T., which was announced at MacWorld 1983, was never released to the public, and only a very small handful of prototypes were ever constructed for the device. One of the few known samples of this was sold on eBay for $8,000 back in 2012. It was even prototyped in both “classic Mac” color and a somewhat more “business looking” dark gray color.

While there are some key details known about the device, such as that it was designed in partnership with BellSouth and offered advanced features for its time, such as online banking access, a full touchscreen, fax and caller ID support, a built in address book, and even the ability to customize ringtones (it even featured Newton-like full handwriting support), a functioning version of the device has never, to my knowledge, been seen by “virgin eyes.”

We’ve obtained exclusive video footage showing the device in working order. The Mac System 6 powered device is fascinating to see in function, and although it seems primitive by today’s standards, it was certainly an impressive feat for 1993. The videos show many features of the device, including the handwriting recognition and address book, as well as the hardware controls.


It really is primitive by today’s standards, but also shows that people don’t limit their dreams to what can be effectively done today.
unique link to this extract

Ampere: SVOD originals struggle to survive beyond two seasons • Broadband TV News


VoD services posted the lowest number of seasons with cancellations coming after an average of 2.1 seasons. Netflix accounts for 68% of VOD cancellations in the USA. 12 of Netflix’s 13 cancellations since September 2018 occurred at three seasons or lower, including four Marvel shows, as well as Originals comedies All About the Washingtons, The Good Cop and Friends from College. Over half of Netflix’s cancellations have been in the sci-fi genre. No streaming series was renewed or cancelled beyond a seventh season.

The Cable Networks account for nearly half of cancellations – 12 of their 29 cancelled titles were Comedy. Cancellations were spread out, occurring over 21 different networks, with no channel making more than three cancellations (AMC and Comedy Central).

The Broadcast Networks accounted for just 22% of cancellations and had by far the highest average cancellation season [seasons before cancellation] at 6.5. Long-running series reaching their conclusion include Supernatural (The CW) and Criminal Minds (CBS) at season 15, and Last Call with Carson Daly (NBC) ending after its 17th season. With half of broadcast networks’ cancellations occurring at season two or below, if a show leaps this hurdle, the potential to run and run is close to infinite.

Fred Black, Analyst at Ampere Analysis says: “The VOD services seem determined to drive subscriber growth through a continuous pipeline of new content, but this comes at the cost of missing out on long-running franchises like NBC’s Law & Order that keep customers coming back year after year, reducing churn.”


The more I think about this, the more I think it’s poor showrunning by the SVODs. Netflix pays hugely to have Friends and The Office US and other long-running series. Unless the calculation is that the actors in a long-running series could hold it to ransom, with little marginal benefit in subscriber numbers.
unique link to this extract

Errata, corrigenda and ai no corrida: none notified

3 thoughts on “Start Up No.1,043: Facebook’s AI death smarts, Amazon staff push for green scheme, how Russia trolls on vaccines, and more

  1. “Once an account is memorialised, we use AI to help keep the profile from showing up in places that might cause distress, like recommending that person be invited to events or sending a birthday reminder to their friends”.

    Going to add this to the ‘low bar for AI’ list. Along with the YouTube commend from a few weeks ago: “Engineers also immediately ‘hashed’ the video, meaning that artificial intelligence software would be able to recognize uploads of carbon copies”.

  2. I find a great way of not being recommended that I invite a dead person to a birthday party is to avoid Facebook like the plague – hopefully advances in NI (Natural Intelligence) will lead more and more people to a similar conclusion 😉

    • I’ve never really been on FB ( I’ve got a years-old ghost acxount I used for 1 month) bit the pressure to join keeps increasing for me and everyone around me, both both IM and blog about 50/50.

      I mostly blame Google. They never managed to
      1- make a convincing Android messaging service à la iMessage. I tried and tried, bit I’ve had to keep the family on Skype for various reasons (reliable, multi-platform, multi-device, OK features and UI; main issues are startup time, new contact identification, and lack of widgets). That’s 3rd-party, so a much harder sell to anyone I can’t simply strong-arm. Most of my contacts are on whatsapp or straight FBMess.
      2- adding a blog feature to messaging seems obvious. I don’t think they ever did that. (I think Skype just did, but nobody cares).

      Until Android has a good enough default, FB will keep dominating. Google’s impotence boggles the mind

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.