Start Up: Facebook’s woeful failures, unsentimental sentiment analysis, location-location-location?, and more

A self-driving Uber car: a pedestrian’s death means the safety debate is now serious. Photo by Eddie Codel on Flickr.

A selection of 15 links for you. Hands on the wheel, eyes on the road. I’m @charlesarthur on Twitter. Observations and links welcome.

Facebook security chief said to leave after clashes over disinformation • The New York Times

Nicole Perlroth, Sheera Frenkel and Scott Shane:


Facebook’s chief information security officer, Alex Stamos, will leave the company after internal disagreements over how the social network should deal with its role in spreading disinformation, according to current and former employees briefed on the matter.

Mr. Stamos had been a strong advocate inside the company for investigating and disclosing Russian activity on Facebook, often to the consternation of other top executives, including Sheryl Sandberg, the social network’s chief operating officer, according to the current and former employees, who asked not to be identified discussing internal matters.

After his day-to-day responsibilities were reassigned to others in December, Mr. Stamos said he would leave the company. He was persuaded to stay through August to oversee the transition of his duties because executives thought his departure would look bad, the current and former employees said. He has been overseeing the transfer of his security team to Facebook’s product and infrastructure divisions. His group, which once had 120 people, now has three, the current and former employees said.

Mr. Stamos would be the first high-ranking employee to leave Facebook since controversy erupted over disinformation on its site. His departure is a sign of heightened leadership tensions at the company.


Where is Zuckerberg in all this? One expects he’ll pop up in a day or two, or post some big screed on his Facebook page. The longer he waits, the more directions the story splinters into.
link to this extract

Facebook and the endless string of worst-case scenarios • TechCrunch

Josh Constine:


Here’s an incomplete list of the massive negative consequences and specific abuses that stem from Facebook’s idealistic product development process. [Thanks to user suggestions, we’ve added some more in an upate].


It’s a long list – Beacon revealing shopping, “engagement ranked feed” leading to fake news, “engagement”-priced ads leading to polarising ads, app platform driving game spam, privacy controls that encourage you to make stuff public, and plenty, plenty more.

Facebook’s shares lost nearly 7% of their value through the day.
link to this extract

Uber halts autonomous cars after 49-year-old pedestrian is killed in Arizona • The Washington Post

Faiz Siddiqui and Michael Laris:


The National Transportation Safety Board has opened an investigation into the crash, NTSB spokesman Eric Weiss said.

Uber issued a short statement.

“Our hearts go out to the victim’s family. We are fully cooperating with local authorities in their investigation of this incident,” a company spokeswoman said.

The vehicle was in autonomous mode at the time of the crash, though a driver was behind the wheel, Tempe police said in a statement. The crash occurred about 10 p.m. Sunday in the area of Curry Road and Mill Avenue, a busy intersection with multiple lanes in every direction.

Police said the vehicle was northbound on Curry Road when a woman, identified as 49-year-old Elaine Herzberg, crossing from the west side of street, was struck. She died at a hospital, the department said.

Missy Cummings, a robotics expert at Duke University who has been critical of the swift rollout of driverless technology across the country, said the computer-vision systems for self-driving cars are “deeply flawed” and can be “incredibly brittle,” particularly in unfamiliar circumstances.


Herzberg wasn’t on a “crosswalk” (UK lingo: pedestrian crossing) when she was hit. But that’s irrelevant. Cars are meant to yield to pedestrians. Lots more to be discovered about this, including how fast the car was going, how well-lit things were, what system it was using to detect obstacles, and more.

So the first self-driving car has killed a non-driver. Now the really hard questions begin. Who’s responsible – the person inside the car, or the authors of the software? How do you stop this happening again – or is there a level of pedestrian killing that is “acceptable”?
link to this extract

Why America can’t regulate bitcoin • Hacker Noon


Hearings on Bitcoin and its derivatives are being held in the USA on a regular basis, and invariably the expert witnesses fail to properly describe the actual processes going on. If they used the correct language and excluded all analogies, the only possible conclusion would be that America cannot regulate Bitcoin under its current legal system. The Constitution guarantees the inalienable rights of American citizens, and therefore Bitcoin is a protected form of publishing. The only way Bitcoin can be made regulable is if the Constitution is changed; and that does not mean adding a new Amendment, it means removing the First Amendment entirely. Inevitably the anti-Bitcoin protagonists will face a robust and ultimately successful legal challenge that will remove the possibility of any sort of “BitLicense” or interference from the CTFC, FinCEN or any other agency. It will also remove any possibility of interference at the State level. The consequence of adhering to the basic law of the United States will cause America to become the centre of all Bitcoin business for the entire world.

Let me explain why this is the case.


It isn’t the case, because he hasn’t recognised the crucial difference between “printing out how bitcoin works” (free to do) and “offering money and services for bitcoins created using the process that was printed out” (which the government will tax and regulate).

Then again he thinks climate change is a fraud (he’s absolutely certain that “anthropogenic global warming theory has been falsified”) and that gun control is stupid.
link to this extract

The ridiculousness of sentiment analysis • Diginomica

Dennis Howlett got an email (three times!) from a PR company certain that it had got some amazing sentiment analysis on peoples’ opinions on social media about United Airlines suffocating a pet:


It was accompanied by the dross, breathtakingly insightful, no sh-t Sherlock commentary that:


This latest incident could be another massive blow to United’s reputation unless upper management takes control of the storm on social in an effective way. In this instance, United will need to do more than just apologize — they will need to provide solutions and reassure their wide customer base, and to do so intelligently, they must be mindful of the overwhelming responses they’re already receiving about the incident. Leveraging social listening during a crisis can help any brand gauge the right response, and hopefully, make a comeback.


And your point is…..????

I don’t know if these people have noticed, but following the United Breaks Guitars fiasco, the company barely missed a beat in reporting earnings.

The most recent incident, while wildly more egregious than the earlier one, will have almost zero effect on United. Unless…a few large corporate specifiers put ethics to the front of their choice parameters and say enough is enough.

The likelihood of that happening is almost zero because, like the other major U.S. airlines, United operates what are near monopoly hubs that act as choke points for others. You wanna go to the Bay Area from Chicago, Houston, Frankfurt or Denver? UA is pretty much your only realistic choice. In short, the U.S. airline majors operate as a set of cartels, ostensibly in competition, but in reality, having ‘safe’ harbors into and out of which they are the mob bosses owners with very little to lose.


link to this extract

Open Location Code and what3words • Medium

David Piesse:


Open Location Code is a Google (Zurich) based code library for giving a code to anywhere in the world. It is entirely based off of latitude and longitude tying it perfectly with GPS; and is indiscriminate with regards land, sea or population. It is a tiny amount of code for this algorithm, but has a few features that make it the best way to locate anything from a latitude & longitude.

Firstly it has scale. The code is built up of pairs of characters (e.g. 9C) that drill you into the next level of detail.

The first pair relates to a 20 by 20 degree square around the world (18 x 9 of them).

The second pair breaks this down to 1 degrees squares (20 by 20) within this 20 square degree area.

The third and fourth pair do the same again with 0.05 and 0.0025 degrees. That makes up roughly 10.3 billion areas.

Beyond that you can further refine up to 3 more characters, all of which are a 4×5 grid within the previous. At 10 characters (excluding the +) you have a area roughly 10m wide, at 11 characters it is only 3.4m x 2.7 ~ to w3w.

So to get to this area ( 6GCRMQPX+9GG) but people can’t remember that.

Cool — that is where w3w wins hands down. They have something that almost any literate people can understand.

However you can convert a OLC code into parts and potentially assign large levels of them a single word. Of the 162 top level areas only ~69 have any substantial population so we could assign only a few dozen words to over half the globe.

Think atlantic.something.something for the whole of the Atlantic ocean; or australia.something.something for Australia. This gives a high level drill down as to where you are. Large swathes of the globe can be covered with a few words; which both frees up word combinations, but also shortens codes eventually.


what3words is a neat concept but as he points out, it has some failings. His idea seems good; the trouble is making it work.
link to this extract

South Koreans reportedly not that enthusiastic about the Galaxy S9 • SamMobile

Adnan F.:


A local news outlet quotes mobile retailers in its report who have witnessed that consumers aren’t that interested in the Galaxy S9 and the Galaxy S9+ since they’re not that different from last year’s models.

Another retailer pointed out that it’s actually the Galaxy A8 (2018) that’s Samsung’s best-selling smartphone in South Korea currently. The handset’s dual front-facing camera is said to be one of the reasons why it’s immensely popular with the late-teens and 20s demographic which likes to take a lot of selfies.

Granted that the Galaxy S9 and Galaxy S9+ are faster and feature better cameras but the design is largely the same and so is the display size. The new flagship looks and feels quite similar to its predecessor and that’s why it might be a difficult sell to those who already own the Galaxy S8…

…Customers who are sold on the new features like Super Slow-motion video, variable aperture camera, AR Emoji, etc will certainly go ahead and purchase the Galaxy S9. However, those who view them as mere novelties or can’t justify purchasing an expensive smartphone for only these features will likely wait for Samsung’s 10th anniversary Galaxy smartphone due next year.


Would have thought the target market is those who own something older than the S8, really. The TV adverts for the S9 in the UK are really good. But the top-end market is static, at best.
link to this extract

A shakeout is underway among VR startups • The Information

Matt Pressberg and Tom Dotan:


A shakeout is underway in the VR startup world. An early flow of investments by consumer marketers and film studios, enthusiastic about experimenting with VR for marketing, has dried up. Instead, studios and venture capitalists are focusing on arcade-style VR installations at malls and movie theaters. These typically charge $30 for a half hour experience tied to a movie—like the Star Wars VR experience near Disneyland, where people wearing headsets can roam amongst Stormtroopers. That’s not a business best suited to smaller VR studios.

The shift reflects the continuing search for a viable business model tied to VR content, which can cost north of $1m a minute to produce. Sales for consumer VR headsets also have been slower than expected, limiting the market for VR games and videos.

“The early phase of VR is over,” says Vince Pizzica, an executive vice president of  Technicolor, which integrates technology for VR content production. Making short entertainment for VR has “no business case,” he says. Aside from VR arcades, education and architecture are both markets with lots of potential, he says.

It’s too early to say whether “location-based entertainment,” as the venues are known, can generate a return for content producers and venue owners. Aside from the high cost of making the entertainment, consumer demand remains uncertain. Mr. Pizzica estimated that the arcade-style model has two or three years to prove it can make money.


Arcades were the first success story for VR. Maybe they’re going to be the second as well. But the signs aren’t promising.
link to this extract

Demand for augmented reality/virtual reality headsets expected to rebound in 2018 • IDC


Worldwide shipments for augmented reality (AR) and virtual reality (VR) headsets will grow to 68.9m units in 2022 with a five-year compound annual growth rate (CAGR) of 52.5%, according to the latest forecast from the International Data Corporation (IDC) Worldwide Quarterly Augmented and Virtual Reality Headset Tracker. Despite the weakness the market experienced in 2017, IDC anticipates a return to growth in 2018 with total combined AR/VR volumes reaching 12.4m units, marking a year-over-year increase of 48.5% as new vendors, new use cases, and new business models emerge.

The worldwide AR/VR headset market retreated in 2017 primarily due to a decline in shipments of screenless VR viewers. Previous champions of this form factor stopped bundling these headsets with smartphones and consumers have shown little interest in purchasing such headsets separately. While the screenless VR category is waning, Lenovo’s successful fourth quarter launch of the Jedi Challenges Mirage headset—a screenless viewer for AR—showed the form factor may still have legs if paired with the right content. Other new product launches during the quarter included the first Windows Mixed Reality VR tethered headsets with entries from Acer, ASUS, Dell, Fujitsu, HP, Lenovo, and Samsung.

“There has been a maturation of content and delivery as top-tier content providers enter the AR and VR space,” said Jitesh Ubrani senior research analyst for IDC Mobile Device Trackers. “Meanwhile, on the hardware side, numerous vendors are experimenting with new financing options and different revenue models to make the headsets, along with the accompanying hardware and software, more accessible to consumers and enterprises alike.”


Not mentioned anywhere in this release: what the actual uses, demand and markets are that will drive this growth. I just don’t see it.
link to this extract

Be ready to pay a lot for Vive Pro’s higher-res virtual reality • Ars Technica

Kyle Orland:


HTC’s higher-resolution Vive Pro, first announced back in January, is setting new records for the price of a mass-market virtual reality headset. In pre-orders starting today ahead of planned April 5 shipments, customers will have to shell out $799 for the improved Vive Pro headset, a price that does not include any controllers or Lighthouse tracking base stations.

While the original Vive also cost $799 when it launched nearly two years ago, that package included two controllers and the two tracking stations necessary for un-occluded, room-scale VR. Existing HTC Vive owners will be able to reuse those accessories if and when they upgrade to the Vive Pro headset. New users, however, will currently have to purchase them à la carte (an HTC representative tells Ars that pricing for a separate “full kit” Vive Pro package will be announced soon).

HTC currently sells Vive controllers for $130 each and tracking base stations for $135 each. That means new Vive Pro customers will have to pay $1,330 for a higher-fidelity version of the same basic hardware included in the package for the original Vive (which is being reduced to $499 today, from the $599 price it has held since last April).

While the Vive and Vive Pro both technically work with other controllers (including some competing motion-tracked options), the vast majority of Vive-compatible VR software is designed to work with the hand-tracking “wand” controllers that have been a standard part of the Vive package since launch. At least one tracking base station is required to follow the Vive headset as a user moves their head, though two are recommended for more “room-scale” applications.


I think I’d argue with that phrase “mass-market virtual reality headset”. There is no mass market for VR headsets at present – and there’s still no obvious sign of what would spark wide adoption. (This isn’t, of course, good news for HTC. But these days, pretty much nothing is.)
link to this extract

Apple likely to buy up to 270 million smartphone panels in 2018 • Digitimes

Rebecca Kuo and Steve Shen:


Apple’s purchases of OLED panels in 2018 will reach 110-130m units, including 70-80m 5.9in units for the current iPhone X and an upgraded version of the same size. The remaining will be 40-50n 6.5in OLED panels for the production of a less expensive model, said the sources.

Apple will also take up a total of 60-70m 6.1in all-screen LTPS panels for another new iPhone model to be released in fall 2018, as well as 60-70m LTPS panels in 4-, 4.7- and 5.5in sizes for the production of iPhone 8 devices, iPhone SE and iPhone 7, indicated the sources.

Samsung Display will continue to serve as the primary supplier of OLED panels to Apple, according to IHS Markit. However, LG Display will strive to become the second supplier in the second half of 2018, aiming to secure orders for the new 6.5in OLED model to be launched this fall.

On the other hand, Japan Display, Sharp and LG Display will be the main suppliers of LTPS panels to Apple in 2018, said the sources.


So the expectation – if this is right – is that the larger “X” model will sell 40-50m units in a single quarter (given that Apple releases close to calendar Q4)? Or that those purchases will carry it across into the first quarter of 2019 as well, which would make a lot more sense.
link to this extract

Apple is secretly developing its own screens for the first time • Bloomberg

Mark Gurman:


Apple is designing and producing its own device displays for the first time, using a secret manufacturing facility near its California headquarters to make small numbers of the screens for testing purposes, according to people familiar with the situation.

The technology giant is making a significant investment in the development of next-generation MicroLED screens, say the people, who requested anonymity to discuss internal planning. MicroLED screens use different light-emitting compounds than the current OLED displays and promise to make future gadgets slimmer, brighter and less power-hungry.

The screens are far more difficult to produce than OLED displays, and the company almost killed the project a year or so ago, the people say. Engineers have since been making progress and the technology is now at an advanced stage, they say, though consumers will probably have to wait a few years before seeing the results.

The ambitious undertaking is the latest example of Apple bringing the design of key components in-house. The company has designed chips powering its mobile devices for several years. Its move into displays has the long-term potential to hurt a range of suppliers, from screen makers like Samsung Electronics Co., Japan Display Inc., Sharp Corp. and LG Display Co. to companies like Synaptics Inc. that produce chip-screen interfaces. It may also hurt Universal Display Corp., a leading developer of OLED technology.


Back in October 2017 I linked to a Digitimes story which said


Through acquiring US-based LuxVue Technology, Apple has acquired patented micro LED technology, especially that for mass transfer. In February 2017, Apple acquired a patent of fingerprint recogniton on micro LED panels via LuxVue, signaling Apple’s continued R&D of micro LED technology. But some reports have claimed that Apple, after LuxVue encountered bottlenecks in mass transfer, has withdrawn some of its technological staff working at a micro LED lab in northern Taiwan.


Seems like they may have got past the bottlenecks. But don’t hold your breath. Even so, it’s a classic example of Apple seeking to control the technologies that it depends on, from fingerprints to facial recognition to processors to modems. Only surprising that screens should have taken so long, really.

link to this extract

11 different brands of AA batteries, tested in identical flashlights. [OC] : dataisbeautiful • Reddit

The graphic seems to tell the obvious story – but, as commenters then point out, if you really want to know what you’re getting for your money, you need to look at the price per hour. That, it turns out, doesn’t go in line with the graph.
link to this extract

How has the average Hollywood movie crew changed? • Stephen Follows

Follows works in the film industry and (if you didn’t know) in his spare time does some wonderful analyses of public data from it:


In the past, I’ve looked at how big a movie crew can get, for both UK films and Hollywood movies. But I was recently asked by a reader how the composition of such crews has changed over time. Which departments are getting larger? Which jobs are on the rise and which are waning?

To answer this, I looked at the credits of the top 200 US-grossing movies of each of the past 20 years (1997-2016), giving me a dataset of 4,000 movies.


You can probably guess the department that’s grown biggest and fastest (clue: it’s not stunt actors), but some of the others are surprising.
link to this extract

Amazon Alexa meets music composed by AI in DeepMusic • RAIN News

After yesterday’s request for a sample of that Amazon Alexa AI-generated music, reader Alex Barredo points us to this, by Anna Washenko:


The AI compositions are generated from a collection of audio samples and a neural network. None of the music has received post-production editing by a human. If you listen on an Echo Show or Echo Spot speaker, you’ll also see artwork created by AI.

Given the number of services working to aid with the speed and ease of Alexa skill creation, it’s likely that we’ll be seeing a wave of innovative and creative applications of the voice technology. AI-made music is likely just the start of how people will think to take advantage of smart speakers.

Here’s what it sounds like:

Possibly not Grammy caliber, but interesting.


I can see endless possibilities for Muzak and Spotify playlists in this.
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.

1 thought on “Start Up: Facebook’s woeful failures, unsentimental sentiment analysis, location-location-location?, and more

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.