Waze might not be able to predict crashes ahead of time, but it’s good for saying they’ve happened. CC-licensed photo by 7-how-7 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 10 links for you. Think it through. I’m @charlesarthur on Twitter. Observations and links welcome.

Can a machine learn to write for the New Yorker? • The New Yorker

John Seabrook:

»

For several days, I had been trying to ignore the suggestions made by Smart Compose, a feature that Google introduced, in May, 2018, to the one and a half billion people who use Gmail—roughly a fifth of the human population. Smart Compose suggests endings to your sentences as you type them. Based on the words you’ve written, and on the words that millions of Gmail users followed those words with, “predictive text” guesses where your thoughts are likely to go and, to save you time, wraps up the sentence for you, appending the A.I.’s suggestion, in gray letters, to the words you’ve just produced. Hit Tab, and you’ve saved yourself as many as twenty keystrokes—and, in my case, composed a sentence with an A.I. for the first time.

Paul Lambert, who oversees Smart Compose for Google, told me that the idea for the product came in part from the writing of code—the language that software engineers use to program computers. Code contains long strings of identical sequences, so engineers rely on shortcuts, which they call “code completers.” Google thought that a similar technology could reduce the time spent writing e-mails for business users of its G Suite software, although it made the product available to the general public, too. A quarter of the average office worker’s day is now taken up with e-mail, according to a study by McKinsey. Smart Compose saves users altogether two billion keystrokes a week.

«

Long, but entertaining – and includes segments where the AI suggests the content. It’s pretty good. Worryingly good.
unique link to this extract

---

Cheap Android smartphones have a disturbing secret • Fast Company

Michael Grothaus:

»

Seventeen dollars for a smartphone sounds like a great deal, especially for people living in poverty who can barely afford rent.

But there’s a problem: low-cost smartphones are privacy nightmares.

According to an analysis by the advocacy group Privacy International, a $17 Android smartphone called MYA2 MyPhone, which was launched in December 2017, has a host of privacy problems that make its owner vulnerable to hackers and to data-hungry tech companies.

First, it comes with an outdated version of Android with known security vulnerabilities that can’t be updated or patched. The MYA2 also has apps that can’t be updated or deleted, and those apps contain multiple security and privacy flaws. One of those pre-installed apps that can’t be removed, Facebook Lite, gets default permission to track everywhere you go, upload all your contacts, and read your phone’s calendar. The fact that Facebook Lite can’t be removed is especially worrying because the app suffered a major privacy snafu earlier this year when hundreds of millions of Facebook Lite users had their passwords exposed. (Facebook did not respond to request for comment.)

Philippines-based MyPhone said the specs of the MYA2 limited it to shipping the phone with Android 6.0, and since then it says it has “lost access and support to update the apps we have pre-installed” with the device. Given that the MYA2 phone, like many low-cost Android smartphones, runs outdated versions of the Android OS and can’t be updated due to their hardware limitations, users of such phones are limited to relatively light privacy protections compared to what modern OSes, like Android 10, offer today.

The MYA2 is just one example of how cheap smartphones leak personal information, provide few if any privacy protections, and are incredibly easy to hack compared to their more expensive counterparts.

«

unique link to this extract

---

Waze data can help predict car crashes and cut response time • WIRED

Aarian Marshall:

»

In May, a team of medical researchers with UCLA and UC Irvine published a paper in the journal Jama Surgery suggesting that places in California might be able to use data from the crowdsourced traffic app Waze to cut emergency response times. (Waze has a four-year-old program that gives cities traffic data in exchange for real-time information about problems its users might want to avoid, like sudden road closures.) By comparing the data from the Google-owned service with crash data from the California Highway Patrol, the researchers concluded that Waze users notify the app of crashes an average of 2 minutes and 41 seconds before anyone alerts law enforcement.

That almost three minutes of lead time might not always be the difference between life and death, says Sean Young, a professor of medicine at UCLA and UCI who serves as executive director of the University of California Institute for Prediction Technology. But “if these methods can cut the response time down by between 20% to 60%, then it’s going to have the positive clinical impact,” he says. “It’s generally agreed upon that the faster you get into the emergency room, the better the clinical outcomes will be.”

Last year, the Transportation Department’s Volpe Center wrapped up its own analysis of six months of Waze and accident report data from Maryland, and found something similar: Its researchers could build a computer model from the crowdsourced info that closely followed the crashes reported to the police. In fact, the crowdsourced data had some advantages over the official crash tallies, because it caught crashes that weren’t major enough to be reported, but were major enough to cause serious traffic slowdowns. The government researchers wrote that the model could “offer an early indicator of crash risk,” identifying where crashes might happen before they do.

Now the DOT is funding additional research, this time with cities that might actually use the data.

«

It’s not quite predict car crashes; more “identify where they’re likely to happen”.
unique link to this extract

---

Using Spotify data to predict what songs will be hits • Tech Xplore

Ingrid Fadelli:

»

According to the researchers [who published a preprint on ArXiv of a system which used four different machine learning models to look at patterns of hits and non-hits, and draw conclusions], if record labels were to use any of these models to predict what songs will be more successful, they would probably choose a model with a high precision rate than one with a high accuracy rate. This is because a model that attains high precision assumes less risk, as it is less likely to predict that a non-successful song will become a hit.

“Record labels have limited resources,” Middlebrook said. “If they pour these resources into a song that the model predicts will be a hit and that song never becomes one, then the label may lose lots of money. So if a record label wants to take a little more risk with the possibility of releasing more hit records, they might choose to use our random forest model. On the other hand, if a record label wants to take on less risk while still releasing some hits, they should use our SVM model.”

Middlebrook and Sheik found that predicting a billboard hit based on features of a song’s audio is, in fact, possible. In their future research, the researchers plan to investigate other factors that might contribute to song success, such as social media presence, artist experience, and label influence.

“We can imagine a world where record labels who are constantly seeking new talent are inundated with mix-tapes and demos from the “next hot artists,”” Sheik said. “People only have so much time to listen to music with human ears, so “artificial ears,” such as our algorithms, can enable record labels to train a model for the type of sound they seek and greatly reduce the number of songs they themselves have to consider.”

«

Is the problem at record labels really that they don’t have enough time to listen to the music?
unique link to this extract

---

Inside Google Stadia • WIRED UK

Stephen Armstrong:

»

For all Stadia’s promises, there remains one big question: can it succeed? And what will it mean for the gaming industry if it does?

“This is definitely the kind of power move that only a large tech company could make,” says David Farrell, lecturer in computer games at Glasgow Caledonian University. We meet in a pub in Edinburgh, south of Scotland’s gaming hub Dundee, where the companies behind Lemmings, Grand Theft Auto, Crackdown and Minecraft were all originally based. In 2018, Edinburgh-based Cloudgine, which developed real-time cloud gaming technology, was bought by Fortnite creators Epic Games to help move its Unreal game engine into the cloud.

“Cloud gaming is the future – although when it comes to the next generation of consoles, Google’s offering isn’t the most exciting thing around, and it’s not clear how long it’ll take to get there,” he says. “In the long term, Google isn’t really trying to be Xbox; they’re trying to be the platform on which everyone else builds their cloud gaming. EA is using Google as its streaming provider rather than developing its own streaming tech – so essentially, they’re offering their ‘Netflix of gaming’ on the back of Google technology. Unless Google comes up with some killer app games, it’s just building the pipes for cloud gaming to run through.”

George Jijiashvili, senior analyst at tech research giant Ovum, has reservations about the technology, especially when it comes to latency and lag. “Most of what Google is promising is possible and deliverable, but there are three or four pain points that will take a few years to be ironed out,” he says. “The biggest one is networks – they can open up new data centres closer to hubs, but most of the networks users are receiving are low quality, and were put in place to transfer voice or small packets of data.”

Majid Bakar insists Google has developed a solution to this. “Our platform and infrastructure allows for techniques that create additional time buffers,” he says. “We can generate frames in less time than it takes consoles or PCs, and with our machine learning experience we have built models to help with the prediction and generation of content faster. This counteracts the impact of network distribution time.”

«

As Farrell says: it’s really about the games. You can have as many data centres as you like, but without the games it’s nothing.
unique link to this extract

---

Most deepfakes are used for creating non-consensual porn, not fake news • VICE

Joseph Cox:

»

While media, politicians, and technologists panic over the risk of deepfakes impacting elections, a new study has found that the vast, vast majority of deepfakes are pornographic in nature. On top of that, to the surprise of absolutely no one, all of the pornographic deepfakes analyzed in the study exclusively targeted women.

The news acts as a reminder that although in the future political actors may adopt deepfakes for the purposes of disinformation, at the moment their use is squarely in their original, designed purpose: to target and harass women.

“[A] key trend we identified is the prominence of non-consensual deepfake pornography, which accounted for 96% of the total deepfake videos online,” the study, titled The State of Deepfakes and authored by cybersecurity company Deeptrace , reads.

«

This misses the point, though. The problem isn’t how many. It only takes one deepfake video going viral and being believed by a significant number of people to make a difference. It only takes a couple being shared in closed Facebook groups to make a small difference. This is a danger at the margins, not in the main field.
unique link to this extract

---

Donald Trump tax return history: a history of presidents providing tax returns • Esquire

Kevin Kruse:

»

On November 17, 1973, the president [Richard Nixon] sought to reestablish his credibility in the fantasy-friendly confines of Disney World. In a televised Q&A session with 400 newspaper editors, he hoped to convince the nation of his honesty and integrity. He only made things worse.

Nixon grew increasingly angry and agitated at the podium when the Orlando press conference turned to questions about his finances. Reporters had been hounding him for weeks, asking how he could afford two separate private homes on his relatively meager presidential salary and whether he’d benefitted personally from administration dealings. There had even been rumors that the President of the United States was being bankrolled in some way by the eccentric billionaire Howard Hughes.

Grabbing the podium with both hands and bobbing nervously on his feet, Nixon tried to dispel the rumors and shore up his credibility:

»

Let me just say this, and I want to say this to the television audience: I made my mistakes, but in all of my years of public life, I have never profited, never profited from public service—I have earned every cent. And in all of my years of public life, I have never obstructed justice. And I think, too, that I could say that in my years of public life, that I welcome this kind of examination, because people have got to know whether or not their President is a crook. Well, I am not a crook. I have earned everything I have got.

«

«

Well, it turned out his tax returns hadn’t been totally on the up-and-up. And then there was the little matter of impeachment. Trump’s been told to hand over his tax returns. I’m looking forward to November 17.
unique link to this extract

---

Facebook to pay $40m in proposed settlement in video metrics suit • Hollywood Reporter

Eriq Gardner:

»

On Friday, several advertising agencies revealed the details of a proposed settlement with Facebook that would end a class action alleging the social media giant overstated the average time its users spent watching video.

According to a brief in support of the settlement, Facebook would pay $40 million to resolve claims. Much of that would go to those who purchased ad time in videos, though $12 million — or 30% of the settlement fund — is earmarked for plaintiffs’ attorneys.

The suit accused Facebook of acknowledging miscalculations in metrics upon press reports, but still not taking responsibility for the breadth of the problem. “The average viewership metrics were not inflated by only 60%-80%; they were inflated by some 150 to 900%,” stated an amended complaint.

Faced with claims of violating unfair competition law, breaching contract and committing fraud, Facebook contested advertisers’ injuries, questioning whether they really relied on these metrics in deciding to purchase ad time. In early rounds in the litigation, Facebook was successful in getting the judge to pare the claims, though until a settlement was announced, several of the claims including fraud were still live. Even after agreeing to pay $40m for settlement, Facebook maintains the suit is “without merit.”

«

“900%” inflation is tenfold. Is Facebook really suggesting that advertisers wouldn’t look at something claiming they’ll watch for 100 seconds when it’s really 10 seconds, and not be persuaded? Or 10 seconds vs 1 second? You only have to ask to know how crazy that defence is.
unique link to this extract

---

Apple increases production of iPhone 11: sources • Nikkei Asian Review

Cheng Ting-Fang, Lauly Li, and Kensaku Ihara:

»

Apple has told suppliers to increase their production of its latest iPhone 11 range by up to 10%, or 8 million units, the Nikkei Asian Review has learned, following better-than-expected demand worldwide for its new cut-price handset.

The increase in orders appears to validate Apple CEO Tim Cook’s new strategy of enticing budget-conscious consumers with cheaper models amid the weakening world economy. The order boost of between 7 million and 8 million units is equivalent to total annual phone shipments this year by Google, a rising iPhone rival in Apple’s home US market.

“This autumn is so far much busier than we expected,” one source with direct knowledge of the situation said. “Previously, Apple was quite conservative about placing orders,” which were less than for last year’s new iPhone. “After the increase, prepared production volume for the iPhone 11 series will be higher compared to last year,” the source said.

«

So there’s downward pressure on pricing as the phone market becomes saturated and people don’t need the tippy-top specs because there’s very little difference as the improvement in capabilities becomes harder to discern. Neat burn on Google, though.
unique link to this extract

---

Mike Postle: why is this the point where he started winning at poker? • YouTube

If you read the lead item in yesterday’s posting, you’ll know there’s a discussion about how Mike Postle is able to win while playing a “high variance” poker style. If you’re interested in more, then via David Chu, here’s a link to a video (whose title is different from mine – I’m not suggesting Postle cheats!) which points to a peculiar breakpoint at which Postle stops losing and starts winning.

It’s to do with his phone, though what I find astonishing about what’s going on is that all the players have their phones with them and are fiddling with them all the time. How do you stop people cheating, or using some kind of card-counting, or whatever, in that situation?

The Postle allegation, though, seems to be about a much more sophisticated method of knowing what others are doing. If he’s doing it, he’s well beyond card-counting.
unique link to this extract

---

Errata, corrigenda and ai no corrida: none notified