Start Up: TSB thefts continue, selfish economists, AirPlay 2!, spot the drowning child, and more

If this ticket wins, the neighbours are more likely to go bust. Photo by Sean MacEntee 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 14 links for you. Not available on ABC. I’m @charlesarthur on Twitter. Observations and links welcome.

TSB left man on hold as his wedding savings were stolen • BBC News

Jon Douglas:


Ben Alford from Weymouth in Dorset said it took more than four and a half hours to get through to TSB, by which time most of the money had gone. He is one of many affected by fraud who have struggled to contact the bank.

TSB says it has put in “additional resources” to support customers.

Ben called TSB after he noticed a £9,000 loan with another company had been taken out in his name without his knowledge. The money had been paid into the TSB joint bank account he shares with his girlfriend, Francesca Cuff.

Ben said a £1,000 overdraft had also been set up without their permission. He says he was logged into internet banking, and waiting for someone at TSB to answer his telephone call, when he noticed that money had begun to be stolen.

“There was initially £5,000 taken out of that account followed by another amount of £4,000, he told BBC Radio 4’s You & Yours programme. “Had they answered their fraud line promptly, none of this money would have been taken because it could have been stopped. I literally watched the money go out of our account”.


Thousands of people are suffering because TSB has not just screwed up the upgrade, but let its security down calamitously. It’s disgraceful.
link to this extract

Reproducibility in machine learning: why it matters and how to achieve it •

JEnnifer Villa and Yoav Zimmerman:


You’ve been handed your first project at your new job. The inference time on the existing ML model is too slow, so the team wants you to analyze the performance tradeoffs of a few different architectures. Can you shrink the network and still maintain acceptable accuracy?

The engineer who developed the original model is on leave for a few months, but not to worry, you’ve got the model source code and a pointer to the dataset. You’ve been told the model currently reports 30.3% error on the validation set and that the company isn’t willing to let that number creep above 33.0%.

You start by training a model from the existing architecture so you’ll have a baseline to compare against. After reading through the source, you launch your coworker’s training script and head home for the day, leaving it to run overnight.

The next day you return to a bizarre surprise: the model is reporting 52.8% validation error after 10,000 batches of training. Looking at the plot of your model’s validation error alongside that of your teammate leaves you scratching your head. How did the error rate increase before you even made any changes?


Via Pete Warden, who is one of Google’s people working on AI. A topic that one would imagine is close to his heart.
link to this extract

The rapid evolution of Homo Economicus: brief exposure to neoclassical assumptions increases self-interested behavior • Science Direct

John Ifcher and Homa Zarghamee:


Economics students have been shown to exhibit more selfishness than other students. Because the literature identifies the impact of long-term exposure to economics instruction (e.g., taking a course), it cannot isolate the specific course content responsible; nor can selection, peer effects, or other confounds be properly controlled for. In a laboratory experiment, we use a within- and across-subject design to identify the impact of brief, randomly-assigned economics lessons on behavior in the ultimatum game (UG), dictator game (DG), prisoner’s dilemma (PD), and public-goods game (PGG). We find that a brief lesson that includes the assumptions of self-interest and strategic considerations moves behavior toward traditional economic rationality in UG, PD, and DG. Despite entering the study with higher levels of selfishness than others, subjects with prior exposure to economics instruction have similar training effects.


link to this extract

Shutting down facebook in PNG is a reality • Papua New Guinea Post Courier

Benny Geteng:


Facebook users in the country can expect a month’s shutdown access to the site in PNG [Papua New Guinea] in order for the Communications and Information Technology Department to carry out research and analysis of its use.

Communications Minister Sam Basil said that the shutdown would enable the department and National Research Institute to conduct further research on how the social network was being used by users.

“The time will allow information to be collected to identify users that hide behind fake accounts, users that upload pornographic images, users that post false and misleading information on Facebook to be filtered and removed.

“This will allow genuine people with real identities to use the social network responsibly,” Mr Basil said.

The Minister said that the department could better analyse the positive impact it would have on the population during the month-long shutdown and weigh the impact of progress without or with its use.

Mr Basil said that his Ministry was trying to enforce the Cyber Crime Act which was legislated in 2016.

“The Act has already been passed, so what I’m trying to do is to ensure the law is enforced accordingly where perpetrators can be identified and charged accordingly. We cannot allow the abuse of Facebook to continue in the country.”


PNG population: about 8 million. Facebook users there: about 600-700,000.
link to this extract

Keeping up with the Joneses: neighbors of lottery winners are more likely to go bankrupt • Bloomberg

Peter Coy:


As if you needed proof that trying to keep up with the Joneses isn’t a good idea, here it is: close neighbors of lottery winners in Canada tended to spend more on conspicuous goods, put more money into speculative investments such as stocks, borrow more money—and eventually declare bankruptcy.

“The larger the dollar magnitude of a lottery prize of one individual in a very small neighborhood, the more subsequent bankruptcies there will be from other individuals in that neighborhood,” says the latest version of a working paper from the Federal Reserve Bank of Philadelphia by Sumit Agarwal of Georgetown University, Vyacheslav Mikhed of the Philadelphia Fed, and Barry Scholnick of the University of Alberta. It’s titled: “Does the Relative Income of Peers Cause Financial Distress? Evidence from Lottery Winners and Neighboring Bankruptcies”…

…A telltale sign was that they raised spending on things that everyone in the neighborhood could see, such as cars, but not on indoor items like furniture.

The new version adds some important insights, co-author Mikhed explained in an email. One is that neighbors who filed for bankruptcy tended to have more of their assets in high-risk investments such as stocks vs. low-risk ones like insurance and cash. That’s consistent with the theory that they were hoping to make a killing in the market and even things up with the lottery winner.


Keep the neighbourhood safe from lottery winners!

link to this extract

China prosecutes 98 over alleged $2bn OneCoin pyramid scheme • CoinDesk

Wolfie Zhao:


the legal process launched in September 2017 and has been conducted in three phases that have seen 98 people prosecuted for allegedly deceiving investors across over 20 provinces in China. A number of those have already been sentenced with up to four years in prison and/or fines ranging from 10,000–5 million yuan ($1,565–$783,000).

The prosecutor said that the scheme involved up to 2 million victims, while the amount of capital received from investors totals as much as 15 billion yuan (around $2 billion). Nearly 1.7 billion yuan ($266 million) has been recovered, the report states.

As previously reported by CoinDesk, the OneCoin scheme, which was founded by an individual called Ruja Ignatova, has been scrutinized by police in a number of countries over suspicions that it is fraudulent.

Promoters in Italy have been fined millions of euros, while authorities in India also moved to arrest suspects associated with OneCoin in April of last year and subsequently brought charges against Ignatova in July.


Pyramid schemes never die, they just look for new formats to exist in.
link to this extract

Morgan Stanley: Apple’s App Store clobbers Google Play • Philip Elmer‑DeWitt

Analyst Katy Huberty put together a presentation about “The Emerging Power of Apple Services”. The telling graphics are these two, I think:


That widening delta between the App Store and Google Play is not what had been expected. Possibly it understates advertising revenue because those figures are hard to extract, but most of the revenue will come from games, and those can be easily estimated. (Note too that Google hasn’t said much about Google Play revenue.)

But it’s clear that iOS customers are really valuable. Android has conquered the world in terms of penetration; Apple has conquered it in terms of getting wallets open.
link to this extract

Microsoft is now more valuable than Alphabet — by about $10bn • CNBC

Jordan Novet:


When Google first passed Microsoft in terms of stock market value six years ago, it looked like the companies were headed in opposite directions.

But over the past 12 months, Microsoft has surged 40%, more than five times Alphabet’s gain, and has again become the more valuable of the two. As of mid-day Tuesday, Microsoft was worth $749bn and Alphabet’s market capitalization stood at $739bn.

Microsoft’s latest rally has been sparked by growth in its cloud-computing business, which is bigger than Google’s though it still trails Amazon Web Services. In March, Microsoft reorganized its Windows and Devices Group and moved its engineering resources into other units, including one focusing on cloud and artificial intelligence.

Both Microsoft and Alphabet beat analysts’ expectations in the first quarter.

Google went public in 2004 and spent the next eight years closing the gap with Microsoft, which debuted on the stock market in 1986. Even after Google first passed Microsoft in 2012, the companies flip-flopped several times over the next few years.


The growing confidence in Microsoft is all down to Nadella tearing it away from its past obsessions – mobile and, most recently, the fixation on Windows as the centre of everything. (There’s a good recent episode of the Exponent podcast with Ben Thompson and James Allworth on this.) Google’s growing, but slower. Where’s its second act?
link to this extract

iOS 11.4 brings stereo pairs and multi-room audio with AirPlay 2 • Apple


HomePod, the breakthrough wireless speaker from Apple, now delivers an even more immersive listening experience throughout the home with support for HomePod stereo pairs and a new multi-room audio system in iOS 11.4. This free software update introduces the most advanced, easy to use, wireless multi-room audio system using AirPlay 2 to play music in any room from any room, move music from one room to another or play the same song everywhere using an iOS device, HomePod, Apple TV or by asking Siri. HomePod is available in the US, UK and Australia and arrives in Canada, France and Germany starting June 18.


So AirPlay 2 – the long-awaited, better-than-v1 flavour – arrives. Now my question is: will the tvOS update that comes with it allow you to set a HomePod as the default output for an Apple TV?

I ask because the HomePod makes a great output speaker for the Apple TV – far better than the reedy speakers of most flat-screen TVs – but although you can set the HomePod as an output, as soon as the Apple TV goes to sleep it forgets about the HomePod, and you have to tangle through the settings to get to the Audio/Video outputs again, and once more set the HomePod as output. It’s as boring to do as it is to read.

Defaults matter; being able to default to this would be huge.
link to this extract

Trump’s right-hand troll • The Atlantic

McKay Coppins speaks to, and profiles, Stephen Miller, the thirtysomething who writes many of Trump’s speeches and has been a right-wing outrageist for decades:


When president Trump needs to learn about an issue, he likes to stage his own cable-news-style shout-fests in the Oval Office. In lieu of primped pundits, he has to make due with White House staffers, but the basic concept is the same: two people with conflicting points of view whacking away at each other as forcefully—and entertainingly—as possible. Trump seems to process information best in this format, according to people who have worked in the administration. Often, when the debate lacks a voice for a position the president wants to hear articulated, he will call Miller into the room and have him make the case.

Miller “can play both sides for the sake of the argument,” Gidley told me. “He can come in and play the staunch conservative or the Democrat, because he understands both.” What’s more, he often wins. “You can pull a debate-club argument out of a hat and Stephen can argue it convincingly,” a former administration official said. “It’s not that he knows everything in the world—it’s that he understands Trump. He’s been dealing with him a long time, and he understands how he inputs information.”

Miller told me that while there is sometimes a need for a devil’s advocate, he spends most of his time pushing for positions that he believes in. Indeed, a review of his record thus far leaves little doubt about the agenda he’s trying to advance, from more aggressive law enforcement to a conservative-nationalist economic policy. Notably, he’s emerged as one of the most strident immigration restrictionists in an administration known for such draconian measures as forcibly separating children from their parents at the border.

But Miller’s work in the White House has also borne the same trollish hallmark that defined his campus activism.


The article doesn’t get to the heart of whether Miller has a cohesive political theory. But maybe he doesn’t need to. He just likes provoking.
link to this extract

What it’s like when Elon Musk’s Twitter mob comes after you • Daily Beast

Erin Biba:


look, you don’t have to take my word for it. Maybe a bunch of men calling me a cunt doesn’t strike you as harassment. The thing is, many, many other female journalists have experienced the same pile-on from MuskBros every time they tweet criticism of him. Shannon Stirone, a freelance journalist who covers space for publications like Popular Science, Wired, and The Atlantic, told me: “Sadly there is a pattern to what happens after criticizing Elon. There is a reason I don’t do it very often because I don’t enjoy dealing with the backlash from the army of men who come out to defend him. I’ve gotten replies calling me a ‘stupid bitch’ and names along the same vein. They are so deeply angry and instead of using their words they lash out in the only way they seem to know how which is to be abusive and demeaning.

“It is as though they’ve invested their own identity as males into Elon and his work that when anyone (especially women) dares to say anything that isn’t ‘praise for Elon’ it’s only a matter of minutes before the nasty messages come flowing in,” Stirone said. “That ‘bro’ culture is aggressive and deeply misogynistic. It’s exhausting and painful to watch my female colleagues get threats and hurtful messages sent to them all because we called him out.”

Mika McKinnon, a geophysicist and freelance science journalist who writes for Gizmodo, Racked, New Scientist, and others, has said she has stopped tagging Elon, SpaceX, or Tesla in any of her tweets in an attempt to protect herself from the onslaught of abuse.

“The cost of joining a larger conversation is too high. I’m good at handling barrages of hate mail—I was working for Gawker during Gamergate—but it takes energy and it’s easy to miss opportunities when I need to heavily filter my email and social media mentions,” McKinnon said. “This is the only person and company I deliberately avoid tagging out of a desire to not get swamped. It makes me sad that engaging in conversation is so painful, and it took me too long to realize it wasn’t worth the cost.”


The price of being female and having an opinion, especially about a man in the public eye, on Twitter seems calamitously high. The price of being an anonymous male and being rude about a woman in the public eye seems calamitously low. Biba received hate on Twitter, on Instagram, via email.

The “MuskBros” go after male writers too, but the implied threat is lower. The problem is the cultish behaviour, which we see again and again.
link to this extract

Apple’s Star project could be an ARM-based touchscreen hybrid with LTE • 9to5Mac

Guilherme Rambo:


Apple is now working on a new device, codenamed Star. With an interesting model name N84, it could be the first Mac with an ARM processor, or the first iOS notebook…or something completely different.

Macs have been using Intel processors since 2006 and Apple mobile devices have been using Apple-designed processors since 2010. It was recently reported that Apple was going to move Macs to their own processors by 2020.

We have been following information about the Star project for a few months, with sources in the supply chain. It is currently in prototype stage, with prototypes being manufactured by Pegatron, Apple’s partner in China which also manufactures other Apple iOS devices.  A small number of units have been shipped to Cupertino for testing by Apple employees. These prototypes have been in production since at least January 2018.

There’s not much information on what the device could possibly be, but we do know that it has a touch screen, a sim card slot, GPS, compass, is water resistant and it also runs EFI. EFI (Extensible Firmware Interface) is the boot system used by Macs, which leads us to believe that the Star project could potentially be the first ARM-based Mac, with a ship date as soon as 2020.


Also: tweet from Longhorn, a hardware hacker, saying it’s part of a “new device family” which runs an “iOS derivative”. And Digitimes saying Pegatron (which makes laptops) is “likely to get” the order; Pegatron wouldn’t comment.

But then with a bucket of ice-cold water, Mark Gurman “is told” (doesn’t say by whom) that it’s the low-cost LCD-screen iPhone for this year which looks like the iPhone X.

So, pick your rumour.
link to this extract

Smartphone AI: separating hype and reality • CCS Insight Research

Geoff Blaber:


With artificial intelligence firmly at the peak of the hype curve, the industry must be collectively conscious that technologies deliver tangible benefits rather than an empty claim of intelligence. This should be easy given that artificial intelligence isn’t a new phenomenon. What is new is the way solutions are being marketed expressly under the banner of artificial intelligence.

The advent of dedicated accelerators for artificial intelligence workloads is a mixed blessing. Even defining these is difficult because of architectural similarities to digital signal processors (DSPs). Artificial intelligence is becoming pervasive in smartphones, spanning everything from power management to predictive user interface, natural language processing, object detection, facial recognition… the list is endless. For these tasks to be entirely efficient, it’s not realistic that they run exclusively on the CPU or even the graphics processing unit (GPU). Equally, developers need to have the tools to fully maximize the resources available.

This is highly reminiscent of the early days of the smartphone CPU core wars. Adding more cores created little impact beyond marketing hype until developers began writing to those cores to create multithreaded apps.

The approach taken by Qualcomm is noteworthy as it contrasts with that of Apple, HiSilicon and MediaTek, all of which are positioning a single, dedicated accelerator for artificial intelligence. Instead, Qualcomm is emphasizing its heterogeneous approach that comprises its Hexagon DSP, Adreno GPU and Kryo CPU. The Qualcomm AI Engine consists of these cores alongside software frameworks and tools to accelerate artificial intelligence app development using the platform.


The idea that AI-on-your-phone would be the “next big thing” is, I’m happy to point out, what I forecast in my TedX talk in Hilversum back in November 2015. (I was explaining how “selfies” became so big and peaked in 2014.)
link to this extract

Spot the drowning child • Lifeguard Rescue

For those who didn’t get enough of “drowning doesn’t look like drowning”, or who just missed it, here’s a lifeguard training video where you have to spot the drowning child. (As embedded below.)

And – bonus! – a Hacker News discussion on the topic from 2010, which points out that trying to rescue someone who is drowning can be incredibly dangerous to you. Suggestion: take a long stick.

(Because it’s summer, and people are going to be on unfamiliar beaches for holidays soon…)
link to this extract

Errata, corrigenda and ai no corrida: none notified

4 thoughts on “Start Up: TSB thefts continue, selfish economists, AirPlay 2!, spot the drowning child, and more

  1. I don’t get how parents don’t go crazy with worry. A toddler can drown in a dog bowl because they don’t have the arm strength to lift themselves up. I once babysat a sick nephew with instructions to “make sure he keeps breathing”… spent the night by the crib… And I failed all those “spot the drowning child” tests (there are more after the linked one).

    Standard pro-Android whining: 2 things
    1- It’s funny how now that the TOTAL of sales puts the Android ecosystem ahead of iOS, the discussion is moving to PER-DEVICE sales. Apart from that goalpost move tickling my paranoid streak, I’m not sure it’s pertinent: dev costs are mostly fixed, so what counts to devs is total sales, not per-device ?
    2- I’m not sure how they came to that per-device ratio. Half of Android’s app revenue is in China so not on the PlayStore. They didn’t count thatrevenue, but did they count Chinese devices when they divided the app revenue ?

    • On the Android point: dev costs aren’t that fixed. There are many more Android devices to test for, which has led to a combinatorial explosion problem which has been noted in the past. The per-device metric means that if you are as popular (ie on as many devices) on iOS as on Android, you’ll make more money on iOS and your cost of development will be lower. Just because there are more Android devices, that doesn’t mean you’ll get more downloads, especially on a paid app.
      I doubt they counted Chinese AOSP devices in the device stat. Google gives its figure for Android outside China and the growth in eg India is known.

      • Most estimates I’ve seen say developing for Android is 30% more expensive than for iOS. So not an “explosion”, and reasonnable when you realize you’ll be supporting 10x more installed devices, and 300x more models. I’d guess many of the old horror stories came from devs struggling to adapt to a new platform. Devs certainly don’t test on all devices, part of Google’s certification is “make sure AppStore apps work”, so it’s on the OEM to not mess apps up. I keep a handful of marginal devices on hand (I’d be hard-pressed to tell you their brand… Cube, Teclast, and… a mysterious H3088), I haven’t had an issue with any app.

        You do get a lot more downloads on Android (Google Play alone is 2x the downloads of Apple’s AppStore, add China to that so 3-4X in the end), what you don’t get is money from store payments. I’m not sure about ad revenue, it’s probably not utterly negligible, though it is rarely mentioned. I’ve seen several anecdotes about devs surprised they made as much form Android as from iOS, but from different sources (up-front vs IAP and ads). Not data, but I’ve never seen data about that.

        A lot of Android figures floating around don’t reliably discriminate between PlayStore and China. I’ve seen the confusion more often than not, so I like to double-check.

  2. I set HomePod as my audio output on my Apple TV (3rd generation) in March and haven’t had to do it again since. It must be a TVOS thing…?

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.