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 11 links for you. Just enough. I’m @charlesarthur on Twitter. Observations and links welcome.
Never in the last 70 years has a major advanced economy left a free-trade area. Brexit is providing the first real-world evidence of the costs that come from undoing the intricate bonds of globalization.
It is of course an extreme case of deglobalization: The European Union’s single market for goods, services, capital and labor is much more integrated than other free trade zones. Yet many of the barriers that are bound to rise between Britain and its partners, such as on regulations, trade penalties and immigration, are similar to those cropping up in the wider world, such as between the US and its partners.
Measuring the effect of Brexit is complicated by the fact it hasn’t happened yet. British and European leaders met Wednesday in an effort to bridge differences on a post-Brexit deal. Without a deal, Britain could see tariff and nontariff barriers snap back to the maximum the World Trade Organization permits.
Yet without a single tariff going up, Brexit has clearly extracted a price. This can be seen by comparing Britain to a basket of peer economies whose performance closely tracked Britain’s until it voted to leave the EU in June 2016. Pierre Lafourcade, Arend Kapteyn and John Wraith of UBS construct such a synthetic Britain from a blend of other members of the Organization for Economic Cooperation and Development.
Actual and synthetic Britain track each other closely from 1995 to mid-2016, then diverge: Actual British output is now 2.1% below this counterfactual. UBS attributes this divergence primarily to household consumption, which is now 1.7% below its counterfactual, and investment, which is 4% lower.
Completely as economic theory would predict – comparative advantage and so on.
link to this extract
What would the world look like without Facebook? Chartbeat had a glimpse into that on Aug. 3, 2018, when Facebook went down for 45 minutes and traffic patterns across the web changed in an instant. What did people do? According to our data, they went directly to publishers’ mobile apps and sites (as well as to search engines) to get their information fix. This window into consumer behavior reflects broader changes we see taking hold this year around content discovery, particularly on mobile. This is good news for publishers.
Despite volatility driven by algorithm shifts and intense news cycles, user demand for content (represented by traffic across the web) is quite stable. But the sources of that traffic are anything but static. In fact, we’ve seen a major reversal in the specific sources driving traffic to publisher sites in the last year.
• Mobile traffic has seen double-digit growth and surpassed desktop, which saw double-digit declines.
• On mobile, Facebook is down nearly 40% since January 2017, while Google Search has seen a 2x growth in that same time period. That means increases in Google Search referral traffic have more than offset any declines in Facebook referral traffic.
• Additionally — and of significant importance — mobile direct traffic to publishers is now greater than traffic sent by Facebook to publishers’ sites. This means consumers are now more likely to get their news by typing in a publisher URL or opening an app than by being referred through Facebook.
Intel has struggled mightily the past few months, but it may be able to retrieve some lost love by showing strong data-centre growth and progress on rolling chips using a long-overdue manufacturing process.
Intel is scheduled to report quarterly earnings after the close of markets on Thursday. On Monday, rumors circulated that Intel was killing off its 10nm manufacturing process following a string of delays on the year, but Intel was quick to deny the reports.
“10nm,” where “nm” means nanometers, refers to how small a chip maker can make the transistors that go on a computer chip, with the general rule being that smaller transistors are faster and more efficient in using power. Advanced Micro Devices Inc. AMD, has been chipping away at Intel’s dominance as its 7nm chip manufacturing process has been hailed as equal or even superior to Intel’s.
That is just the latest problem for Intel, which has had a trying year. The chip maker was hit late in 2017 by news of twin vulnerabilities baked into its chips, then dumped its chief executive — who has not been replaced on a permanent basis — while dealing with a shortage of chips thought to stem from manufacturing-process issues.
The reports of killing off the 10nm came from SemiAccurate, which called it “struggling”. Intel’s denial feels like one of those “in good time we’ll agree” denials.
If correct, then that really is the end of Moore’s Law for Intel.
link to this extract
Gmail creator and YC partner Paul Buchheit on joining Google, how to become a great engineer and happiness • Triplebyte Blog
Q: If you’re thinking about joining a startup, how do you tell if the founders are like Larry and Sergey or if they’re an Elizabeth Holmes?
PB: Right, that’s the worst combination: smart and full of s**t. I think you have to interview them a little bit. Ask hard questions and see if they give direct, insightful answers, or if they’re evasive and dismissive. It also helps if there is a product you can try. I would avoid startups that have a ton of hype and no product.
Q: Generally, when you are interviewing with a startup, how should you decide if it is the right company for you?
PB: Looking back, one of the things that really impressed me about Google, which is probably good advice for anyone choosing a startup to work at, was that the interviewers all asked really smart questions. They asked things that only people who really knew their stuff could have answered well. Urs asked me, “Let’s say you have a server, and it’s running really slowly for some reason, how do you diagnose the cause?” To answer that question, you actually have to understand systems really well.
Their questions required being able to think at all these different levels: “Is there something going on in the kernel? Do you understand that hard drives are not these magical things which spit out information? Do you know why random access takes time?”
I only interviewed at one other company and they asked stupid questions like, “name the seven layers of the OSI Networking Stack,” or something that you’d pull out of a textbook, not things that were actually interesting.
Also, when I first went to work at Google, I had the opposite feeling I described having at Intel. I was excited. I woke up in the morning and was excited to go to work. There was this buzz of productivity in the office all the time. I think that’s one way to know if a startup is doing well: When you go into their office, you can just tell. Are people busy working, or are they sitting around on Twitter wasting time? Are people showing up because they have to, or are they eagerly working because they’re excited? Google was a really energizing place to be back then.
And plenty more. It’s fascinating.
link to this extract
Gerrit de Vynck:
Uninstall tracking exploits a core element of Apple Inc.’s and Google’s mobile operating systems: push notifications. Developers have always been able to use so-called silent push notifications to ping installed apps at regular intervals without alerting the user—to refresh an inbox or social media feed while the app is running in the background, for example. But if the app doesn’t ping the developer back, the app is logged as uninstalled, and the uninstall tracking tools add those changes to the file associated with the given mobile device’s unique advertising ID, details that make it easy to identify just who’s holding the phone and advertise the app to them wherever they go.
The tools violate Apple and Google policies against using silent push notifications to build advertising audiences, says Alex Austin, CEO of Branch Metrics Inc., which makes software for developers but chose not to create an uninstall tracker. “It’s just generally sketchy to track people around the internet after they’ve opted out of using your product,” he says, adding that he expects Apple and Google to crack down on the practice soon. Apple and Google didn’t respond to requests for comment.
At its best, uninstall tracking can be used to fix bugs or otherwise refine apps without having to bother users with surveys or more intrusive tools. But the ability to abuse the system beyond its original intent exemplifies the bind that accompanies the modern internet, says [EFF tech policy director Jeremy] Gillula.
How likely that Apple or Google tries to find some way to block this? Apple more likely than Google.
link to this extract
Undercover researchers for Which? set up dedicated Amazon and Facebook accounts and requested to join several of the “rewards for reviews” groups.
“They were instructed to order a specified item through Amazon, write a review and share a link to the review once it was published. Following the successful publication of the review, a refund for the cost of the item would then be paid via PayPal,” said Which?
But the Which? investigators turned the tables on the fake review factories by posting their honest opinion on the product.
In one example, the investigator gave the product – a smartwatch – a two-star review. “They were told by the seller to rewrite it because the product was free, so it “is the default to give five-star evaluation”, said Which?
In another, the investigator was told that a “refund will be done after a good five-star review with some photo” after receiving a pair of wireless headphones. But after posting a three-star review with photos they were told they would not be refunded unless they wrote a five-star review. The investigator refused, so did not get refunded for the purchase.
When the Guardian searched the Amazon UK Reviewers Facebook group – which has more than 25,000 members – it found postings appearing almost every couple of minutes from companies around the world offering to pay for positive reviews. For example, on Friday, one company was seeking “UK reviewers only” for a “4k action camera waits for review Refund via Paypal just send me your amazon profile”.
Mark J. Girouard, an employment attorney at Nilan Johnson Lewis, says one of his clients was vetting a company selling a resume screening tool, but didn’t want to make the decision until they knew what the algorithm was prioritizing in a person’s CV.
After an audit of the algorithm, the resume screening company found that the algorithm found two factors to be most indicative of job performance: their name was Jared, and whether they played high school lacrosse. Girouard’s client did not use the tool.
“It’s a really great representation of part of the problem with these systems, that your results are only as good as your training data,” Girouard said. “There was probably a hugely statistically significant correlation between those two data points and performance, but you’d be hard-pressed to argue that those were actually important to performance.”
The community of researchers and technologists studying artificial intelligence have warned that this could be possible in any similar AI algorithm that learns about people using historical data.
In 2016, Pinboard creator Maciej Cegłowski called machine learning “money laundering for bias.”
“It’s a clean, mathematical apparatus that gives the status quo the aura of logical inevitability. The numbers don’t lie,” Cegłowski said.
Misyrlena Egkolfopoulou and Claire Boston:
The $2bn bond offering, which will be issued in dollars and euros, comes just a week after the company reported a bigger jump in subscribers than Wall Street analysts expected. The bonds would push the cash-burning company’s debt load above $10bn for the first time. Netflix’s market value has soared almost 70% this year to about $140bn.
The US portion of the 10.5-year bond may yield around 6.375%, while the euro notes could pay 4.625%, according to people with knowledge of the matter. Netflix paid less than 6% when it last tapped the market in April, in part because underlying Treasury yields were lower.
“To me it feels a bit like a win-win situation,” said John McClain, a high-yield money manager at Diamond Hill Capital, which oversees $22.6bn including Netflix debt. “You’re buying the highest-quality, high-yield business at yields that are fairly close to the overall market. It’s low-cost funding for them, especially relative to the cost of issuing new equity.”
Netflix said in a statement that it will use proceeds from the offering to continue to acquire and fund new content. The company said last week that it expects to burn about $3bn in cash this year as it continues to prioritize original series and movies.
That’s not even close to serious gearing. Netflix is going to get miles in front of everyone with this. And those are pretty attractive yields; I bet it will have no trouble at all selling it. Hardly “junk”.
link to this extract
You can make a realpolitik case for continuing to engage with Saudi Arabia. Just like my coffee companion [a paid lobbyist for Russian interests] five years ago did for continuing to engage with Russia. See how well that turned out, how since then Russia has become so much more enlightened, so progressive, such a glorious contributor to the commonwealth of nations? …Oh. Saudi Arabia is different, yes, but in a worse way; it’s so sensitive to criticism, overreacts so wildly and violently, because it is fundamentally a fragile state. Nassim Taleb, who predicted the collapse of Syria and its civil war before it happened, has predicted a similar fate for Saudi Arabia.
I don’t think the Trump administration is going to continue its support for Saudi Arabia’s new and erratic leadership for fear of the human or economic consequences if they do otherwise. “Trump’s Razor:” the stupidest reason is most likely to be correct. Here, that means the administration doesn’t want to walk back their Saudi support because they think that will make them look weak. Similarly, who are we kidding, VCs who take money from Saudi LPs aren’t doing so in order to help prop up the Pax Americana; it’s purely because they want the money, and nobody else is prepared to throw around $45bn in cash.
Right now, though, and for the foreseeable future, sovereign Saudi money is tainted, poisoned, blood money.
“Trump’s Razor”. Nice. (Concept originated back in July 2016, by Josh Marshall, about Trump wanting to reverse his decision to have Mike Pence as his vice-presidential candidate; named by John Scalzi.)
link to this extract
Brian Hayes, author of the book Infrastructure, on the explosions that blew up mains gas-connected buildings in Massachusetts in September, which was caused by a feedback loop that wasn’t actually a loop – so it pushed up pressure because its readings said the pressure was too low, measured in the wrong pipes:
when you open the valve to increase the inflow of gas, you expect the pressure to increase. (Or, in some circumstances, to decrease more slowly. In any event, the sign of the second derivative should be positive.) If that doesn’t happen, the control law would call for making an even stronger correction, opening the valve further and forcing still more gas into the pipeline. But you, in your wisdom, might pause to consider the possible causes of this anomaly. Perhaps pressure is falling because a backhoe just ruptured a gas main. Or, as in Lawrence last month, maybe the pressure isn’t actually falling at all; you’re looking at sensors plugged into the wrong pipes. Opening the valve further could make matters worse.
Could we build an automatic control system with this kind of situational awareness? Control theory offers many options beyond the simple feedback loop. We might add a supervisory loop that essentially controls the controller and sets the set point. And there is an extensive literature on predictive control, where the controller has a built-in mathematical model of the plant, and uses it to find the best trajectory from the current state to the desired state. But neither of these techniques is commonly used for the kind of last-ditch safety measures that might have saved those homes in the Merrimack Valley. More often, when events get too weird, the controller is designed to give up, bail out, and leave it to the humans. That’s what happened in Lawrence.
This is a fascinating little discussion (with a couple of other accidents, including the notorious Air France 447 from Rio de Janeiro) which leaves much to think about. It also reminded me of control theory, which I haven’t had to think of in decades. (Via Ben Thompson.)
link to this extract
Jeff Atwood, co-founder of Stack Overflow (used by gazillions of flummoxed coders, including me):
I am honored and humbled by the public utility that Stack Overflow has unlocked for a whole generation of programmers. But I didn’t do that.
• You did, when you contributed a well researched question to Stack Overflow.
• You did, when you contributed a succinct and clear answer to Stack Overflow.
• You did, when you edited a question or answer on Stack Overflow to make it better.
All those “fun size” units of Q&A collectively contributed by working programmers from all around the world ended up building a Creative Commons resource that truly rivals Wikipedia within our field. That’s … incredible, actually.
But success stories are boring. The world is filled with people that basically got lucky, and subsequently can’t stop telling people how it was all of their hard work and moxie that made it happen. I find failure much more instructive, and when building a business and planning for the future, I take on the role of Abyss Domain Expert™ and begin a staring contest [quoting Nietzsche: “if you gaze long into an abyss, the abyss gazes also into you”). It’s just a little something I like to do, you know … for me.
Thus, what I’d like to do right now is peer into that glorious abyss for a bit and introspect about the challenges I see facing Stack Overflow for the next 10 years.
The fact that SO (as it gets called all over the place) is principally and was always intended to be a curated wiki and that it is so enormously useful, just like Wikipedia (even if one dislikes the sausage-making process in the latter), seems to me to indicate something important about curated wikis v pretty much every other form of unmediated content collection system.
Tightly curating knowledge is obviously a more bounded problem than lightly curating opinion (as in social media). But why does the latter break down so easily into abuse? Because of the light curating, or the nature of the content?
link to this extract
Errata, corrigenda and ai no corrida: none notified