Start Up: Google gets lobbying, Samsung rejigs, iPhone X reviewed, Congress quizzes Facebook, and more

Soon we won’t need human celebrities – AI will be able to generate photos of entirely fake ones. Weird, huh? Photo by Thomas_H_photo on Flickr.

A selection of 10 links for you. Usable straight from the fridge! I’m @charlesarthur on Twitter. Observations and links welcome.

Google’s dominance in Washington faces a reckoning • WSJ

John D. McKinnon and Brody Mullins:


Even before Mr. Trump was sworn in Jan. 20, Google faced its first tough policy challenge of the new era. During the transition, opponents of the company began pushing to install a Google adversary, Utah Attorney General Sean Reyes, as the new chairman of the Federal Trade Commission, which enforces antitrust laws.

In 2016, Mr. Reyes had called for the FTC to reopen a closed antitrust investigation into Google—a major threat to the firm, which controls more than 80% of the business for internet search by some measures. Earlier this year, European officials imposed a groundbreaking $2.7 billion antitrust fine on Google for unfairly steering web searchers to Google’s own shopping platforms.

Earlier, Mr. Reyes had joined with a half-dozen other state attorneys general in battling with Google over what some viewed as the company’s facilitation of internet ills such as online sex trafficking. The company says it has made extensive efforts to combat such harms.

Google responded to the threat of potentially unfriendly policies from Mr. Reyes by engaging a squadron of GOP lobbyists to press the incoming Trump administration not to name him to the position, and instead pick another candidate who was viewed by some as more Google-friendly, according to several people familiar with the matter. The lobbyists argued that if Mr. Reyes were tapped, the company would flex its muscles in the Senate to block his confirmation.

“Google plays hardball beyond what most companies are willing to do,” said Jon Bruning, the former Nebraska attorney general who was part of the battle against Google. Google’s effort helped keep Mr. Reyes from being nominated, according to people familiar with the matter.


Normally I find these stories compelling, but the WSJ here has shifted into a “Google tries to get political influence, and that’s got to be a bad thing” mode. What’s noticeable about the whole story is that it feels as though it portrays Schmidt’s and Googlers’ support of the Democrats as bad, and even misguided (“Populist conservatives are particularly hostile to big tech, given its size and pervasive influence, as well as its support for immigration rights and other causes that clash with their economic nationalism” says an earlier passage).

I’ve included this because although it just shows politics working as politics does in the US, the WSJ’s tone demonstrates something else – an undercurrent of disdain, even hostility, for those who backed the loser.

That doesn’t seem good in a newspaper, or in a country. (Oh, and also: “internet ills such as online sex trafficking”? Is that different from real-world sex trafficking? See what I mean about the tone?)
link to this extract

Samsung unveils new management to quell leadership crisis • The New York Times

Raymond Zhong:


The troubles at Samsung’s management have not diminished its status as an innovative — and highly profitable — powerhouse in the technology world. That point was underscored on Tuesday, when the company reported a record-high profit for the second quarter in a row.

But the newly appointed executives, who have been drawn from Samsung’s deep ranks of professional managers, could help ensure the company can run while the fate of the powerful family that built the Samsung empire remains under a cloud.

The three executives appointed on Tuesday to run Samsung’s major business units are Kim Kinam, who will lead its lucrative components business; H.S. Kim, its new consumer-electronics chief; and D. J. Koh, who will lead the mobile-device division. The three men are expected to serve as co-chief executives once they are elevated to the company’s board, a decision that requires shareholder approval.

Kim Kinam succeeds Kwon Oh-hyun, who was widely credited with building up the components business — a major revenue driver for Samsung as other smartphone makers bought up its chips and displays.


Samsung just announced record profits, principally driven by semiconductor sales (up 51% year-on-year; memory was up 65%). Mobile was substantially up, but that’s comparing to the Note debacle. Mobile revenues and profits actually fell slightly sequentially.

So Apple, and all the companies reliant on Samsung for OLED and other displays and for memory, have contributed to this. Anyhow, if this is a company in crisis, imagine how it’s going to be when it’s focussed.
link to this extract

New research from Nvidia shows that the era of easily-faked, AI-generated photos is quickly emerging • Quartz

Dave Gershgorn:


Three years ago, after an argument at a bar with some fellow artificial intelligence researchers, Ph.D student Ian Goodfellow cobbled together a new way for AI to think about creating images. The idea was simple: one algorithm tries to generate a realistic image of an object or a scene, while another algorithm tries to decide whether that image is real or fake.

The two algorithms are adversaries—each trying to beat the other in the interest of creating the final best image—and this technique, now called “generative adversarial networks” (GANs) has quickly become a cornerstone of AI research. Goodfellow is now building a group at Google dedicated to studying their use, while Facebook, Adobe, and others are figuring out how to use the technique for themselves. Uses for data generated this way span from healthcare to fake news: machines could generate their own realistic training data so private patient records don’t need to be used, while photo-realistic video could be used to falsify a presidential address.

Until this month, it seemed that GAN-generated images that could fool a human viewer were years off. But last week research released by Nvidia, a manufacturer of graphics processing units that has cornered the market on deep learning hardware, shows that this method can now be used to generate high-resolution, believable images of celebrities, scenery, and objects. GAN-created images are also already being sold as replacements for fashion photographers—a startup called Mad Street Den told Quartz earlier this month it’s working with North American retailers to replace clothing images on websites with generated images.


link to this extract

Live updates: Facebook, Google and Twitter testified before Congress today • Recode

Kurt Wagner and Tony Romm did a liveblog, which is well worth a look:


4.40pm ET: Sen. Mazie Hirono just asked the million dollar question to Facebook’s Colin Stretch:

Hirono: “In an election where a total of about 115,000 votes would have changed the outcome, can you say that the false and misleading propaganda people saw on your Facebook didn’t have an impact on the election?”

Stretch: “Senator, we’re not well positioned to judge why any one person or an entire electorate voted as it did.”

Stretch added, again, that the content Facebook found was a “very small fraction”of everything available on Facebook and that it still shouldn’t have been there. But Hirono’s question is the big one: How much influence did this misinformation campaign actually have? It’s seems like it will be virtually impossible to quantify.

4:51 pm ET: Sen. Al Franken is now asking Facebook’s Stretch about the company’s embarrassing ad targeting feature, which allowed marketers to target ads for groups of users who identified as “Jew haters” and other inappropriate labels. Facebook claims it didn’t know those targeting categories existed, claiming they were created by a software algorithm. Franken was in disbelief over the fact that Facebook was unaware that those options were available to advertisers.


link to this extract – Deep learning models with demos


Deep learning models with demos
Sortable and searchable compilation of pre-trained deep learning models. With demos and code.

Pre-trained models are deep learning model weights that you can download and use without training. Note that computation is not done in the browser


This stuff is all going to be available anywhere, everywhere, all the time.
link to this extract

Super Mario Run’s 200 million downloads didn’t result in ‘acceptable profit’ for Nintendo • The Verge

Andrew Webster:


Nintendo’s first mobile game, Super Mario Run, was enormously popular — but that doesn’t mean it was a success for the company. During its most recent earnings report, Nintendo revealed that Mario Run has been downloaded 200 million times, 90% of which came from outside of Japan. However, Nintendo says that despite these big numbers, the game has “not yet reached an acceptable profit point.” While Nintendo didn’t reveal any specifics with regards to conversion rates, a big sticking point for many with Super Mario Run was its comparatively large price point; it’s free to download, but requires a one-time fee of $9.99 to unlock the whole game.

In contrast, Fire Emblem Heroes — which utilizes a more typical free-to-play structure, with plentiful microtransactions — has been a much more lucrative title for Nintendo. The company didn’t release specific numbers for the game, but says that Heroes’ success has largely been due to its continual updates since the game debuted in February. “For this title, we listened to the voices of our consumers and provided continual updates,” Nintendo says. “As a result, we are on track to meet our overall business objectives, including our profit objectives.”


Apparently with 7m Switches sold, 2m bought Super Mario Odyssey in its first three days. That’s a hell of a conversion rate. I bet that’s reached an acceptable profit point.

Nintendo remains a puzzle: feeling its way in mobile but wedded to its console roots.
link to this extract

iPhone X review: how we tested (and tricked) FaceID • WSJ

Joanna Stern:


A few other things I found during my testing:

• Hats, scarves, earrings, glasses and other everyday accessories are all good. Putting these on didn’t slow down the login process, although my glasses seemed to trip it up a few times, possibly when I held the phone too close to my face. Apple says you should hold the phone 10 to 20 inches out.
• Two different pairs of sunglasses I tried didn’t get in the way of things, either. Although direct sunlight did—I had to adjust my positioning slightly to unlock the phone.
• It hasn’t failed me in the dark. In the back of a dark car at night, and in a pitch-black bedroom, it unlocked.
• Apple says FaceID learns and grows with you, which is why smacking on a handlebar mustache or eyebrows that resemble a raccoon freaks it out. Grow a killer ‘stache over time and it should be fine, though a Groucho ‘stache may never work, as the system needs to see eyes, nose and, yes, a mouth.
• The login process didn’t slow down when registering faces of various skin colors and tones, as other more-primitive facial-recognition systems have been known to do.
• Using Apple Pay is the place I missed TouchID most. While you just have to glance at the phone to authenticate, it’s still something else to think about at the cash register. It may never feel as natural as using a finger.

I tried hard to get a photo to unlock the X: I taped a cutout photo of my face to my face, I pasted it on a big Popsicle stick, I even tried holding a photo on my iPhone 7 up to the new iPhone X. While the Galaxy Note 8 was fooled more than once with the paper cutout, none of this made the X flinch. If the X recognizes a face but it isn’t yours, the lock icon will jiggle.

It makes sense that a photo didn’t work. Apple’s sophisticated system uses two cameras and projected infrared dots to measure the depth of your face. You know what has depth? A mask, a real theatrical mask.


No, that didn’t work either. Notice the fact that it discriminates better than the Samsung. And how they beat it? Needed 8-year-old triplet twins. (Apple says that under 12, FaceID isn’t as reliable.)
link to this extract

PR • OneThirtySeven

Matt Alexander, on the annoyance expressed by a number of more “traditional” tech outlets (where “traditional” means “up to 10 years old”) at a number of YouTubers having got a hands-on use of the iPhone X which they published ahead of said outlets:


it’s worth bearing in mind that Apple’s goal isn’t for you to produce a multi-thousand word treatise about the Face ID mechanism for your audience of people who are statistically most likely to have already pre-ordered the product.

Rather, looking at the past year or so — and particularly the past month and a half — their goal has been to accomplish the following:

• Create a multitude of reasons for consumers, of all types, to justify spending $1,000 on a phone. Although it’s a subsidized cost, many consumers have not considered the true cost of their iPhone ever. Now, with many press outlets leading with the $1,000 total cost angle, Apple needs to combat that perception.

• Assure users that Face ID is better and more secure than Touch ID.

• Emphasize that the “notch” is not the dealbreaker that the tech press (and its audience), primarily, overhyped around the iPhone X announcement.

• Combat the theory that the phone is going to be next-to-impossible to purchase. Combining the elements above, they need to show that it’s going to be available to normal people.

• Finally, show some of the fun of the device, rather than the technical prowess and industrial design.

How would Apple go about accomplishing these goals?

Simply put, they’d create a crashing wave, of sorts, of press around the product, which would enable them to control and manipulate consumer perception of the news, regardless of how more technical reviewers may feel.


Exactly right. We’re well past the point where a smartphone review needs to be some spec-obsessed dive into precisely how many millimetres have been shaved off a bezel. (I felt we reached that point a couple of years ago.) What Apple wants people to get is: how does it feel to own this phone? That so many early reviewers’ response has been to generate an animoji shows that that’s the right choice.

People who think they know PR and marketing consistently underestimate Apple’s grasp of it. Having watched it operate for over 20 years, I think I can spot when it has made a mistake. This wasn’t a mistake.
link to this extract

Building the best possible driver inside Waymo’s castle • TechCrunch

Darrell Etherington:


Waymo classifies anything from Levels 1 through 3 as technically “driver assist” features, according to Krafcik, and this is an “important divide” which Waymo has observed first hand, concluding early on that it’s not an area they’re interested in pursuing.

Krafcik revealed that one of the first products Waymo considered bringing to market back in 2012 and 2013 was a highway driving assist feature, which would handle everything. between onramp and exit, but that also required drivers to be fully attentive to the road and their surroundings while it was in operation.

The results, per Krafcik, were downright frightening: footage taken from the vehicles of Google employees testing the highway assist features, which the company showed us during the briefing, including people texting, doing makeup, fumbling around their seat for charge cables and even, in one particularly grievous instance, sleeping while driving 55 MPH on a freeway.

“We shut down this aspect of the project a couple of days after seeing that,” Krafcik said. “The better you make the driver assist technologies… the more likely the human behind the wheel is to fall asleep. And then when the vehicle says hey I need you to take over, they lack contextual awareness.”

This is why Waymo has been very vocal in the past and today about focusing on Level 4 (full autonomy within specific ‘domains’ or geographies and conditions) and Level 5 (full, unqualified autonomy).


“Lacks contextual awareness” is a nice way to say “won’t know what the hell is going on”. Reminds of the old joke – “I want to die peacefully in my sleep, like my father, not screaming in terror like his passengers.”
link to this extract

An iPhone lover’s review of the Google Pixel 2 • BirchTree

Matt Birchler:


The metal back is coated in a plastic material that is somewhat grippy, but also lets the phone slide into my pocket easily. It feels comfortable without feeling cheap. I do find it odd that they would make a metal back to the phone and then cover it with plastic though. If they had just used plastic for the entire thing they could have added wireless charging, something I very much miss from this phone. I had just gotten used to it with the iPhone 8 and had converted most of my charing spots to wireless. Being forced to use a wire for all charging needs feels like stepping backwards.

The back of the phone also has a fingerprint reader. This is far from the first phone to do it, but it’s the first phone I’ve owned with a rear-mounted fingerprint reader. I haven’t been using it for too long, but I don’t love it personally. People say this location is great because it’s were your index finger naturally is when you’re holding the phone, but my index finger simply does not rest there when I’m using the phone. I can put my finger there easily enough when I pic up the phone to unlock it, but my hand shimmies down the phone to actually use it. I’m about an inch below it and need to stretch to reach it, which is not comfortable. This is most noticeable when trying to authenticate 1Password to fill a form or to make a payment on the Google Play Store.

I also have an issue with the back mounted reader when I’m at my desk at work or driving in the car where my phone is on a stand. The back-mounted fingerprint reader is not accessible in either of these common orientations, so I see this screen a lot:

I’ve entered my PIN more in the past week on this phone than I have in the past year on the iPhone because the reader simply isn’t in a place I can always reach.


He has plenty to say about the processing that generates the portrait effect in the camera too. TL;DR: he likes the phone a lot. But it falls a little short of the iPhone 8 Plus here and there.
link to this extract

Errata, corrigenda and ai no corrida: none notified

You can now sign up to receive each day’s Start Up post by email. You’ll need to click a confirmation link, so no spam.

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.