Wind farms: now DeepMind’s AI is forecasting their output. CC-licensed photo by reynermedia on Flickr.
A selection of 9 links for you. Tested on humans for irritancy. I’m @charlesarthur on Twitter. Observations and links welcome.
US Cyber Command operation disrupted internet access of Russian troll factory on day of 2018 midterms • The Washington Post
The US military blocked Internet access to an infamous Russian entity seeking to sow discord among Americans during the 2018 midterms, several US officials said, a warning that the Kremlin’s operations against the United States are not cost-free.
The strike on the Internet Research Agency in St. Petersburg, a company underwritten by an oligarch close to President Vladi mir Putin, was part of the first offensive cyber campaign against Russia designed to thwart attempts to interfere with a US election, the officials said.
“They basically took the IRA offline,” according to one individual familiar with the matter who, like others, spoke on the condition of anonymity to discuss classified information. “They shut them down.”
The operation marked the first muscle-flexing by US Cyber Command, with intelligence from the National Security Agency, under new authorities it was granted by President Trump and Congress last year to bolster offensive capabilities.
Whether the impact of the St. Petersburg action will be long-lasting remains to be seen. Russia’s tactics are evolving, and some analysts were skeptical the strike would deter the Russian troll factory or Putin, who, according to US intelligence officials, ordered an “influence” campaign in 2016 to undermine faith in US democracy. US officials have also assessed that the Internet Research Agency works on behalf of the Kremlin.
Could they block the Trump family’s Twitter and Instagram access next?
link to this extract
over the past decade, wind farms have become an important source of carbon-free electricity as the cost of turbines has plummeted and adoption has surged. However, the variable nature of wind itself makes it an unpredictable energy source—less useful than one that can reliably deliver power at a set time.
In search of a solution to this problem, last year DeepMind and Google started applying machine learning algorithms to 700 megawatts of wind power capacity in the central United States. These wind farms—part of Google’s global fleet of renewable energy projects—collectively generate as much electricity as is needed by a medium-sized city.
Using a neural network trained on widely available weather forecasts and historical turbine data, we configured the DeepMind system to predict wind power output 36 hours ahead of actual generation. Based on these predictions, our model recommends how to make optimal hourly delivery commitments to the power grid a full day in advance. This is important, because energy sources that can be scheduled (i.e. can deliver a set amount of electricity at a set time) are often more valuable to the grid.
Although we continue to refine our algorithm, our use of machine learning across our wind farms has produced positive results. To date, machine learning has boosted the value of our wind energy by roughly 20%, compared to the baseline scenario of no time-based commitments to the grid.
[Imagine that] Amazon sends a delivery van to my home filled with hard drives containing all its sales and user browsing data for the past year. What do I do with it?
Keep in mind, this trove is worth billions. Accounting rules don’t call (yet) for tech companies to specify their data as a separate asset on the balance sheet, but by any reasonable valuation, Amazon’s purchase data is worth an immense fortune … to Amazon.
That’s because Amazon has built an expansive ecommerce presence, a ruthlessly efficient recommendation and advertising engine, and a mind-bogglingly complex warehouse and fulfillment operation around the data on those hard drives. Ditto Google, Uber, Airbnb, and every other company you’d identify as an “oil field” in this tired metaphor.
Sure, you could maybe sell some of that data—there are companies that would love to know Amazon’s sales data or Google’s search queries or Uber’s routing and pricing history. But here’s the key thing: Those interested outside parties are competitors, and the owners of the data would never in a million years sell it. [Unlike oil, sold between multiple sellers and buyers,] Uber isn’t selling data to Lyft, Amazon isn’t selling data to Walmart, and Airbnb sure isn’t selling user lists to Hotels.com…
…The annual revenue per user for Facebook globally is about $25. In the US and Canada, it’s about $130. Don’t spend it all in one place.
That’s even assuming it’s owed to you by Facebook. Many of the high-value ad placements, such as that creepy ad for the product you browsed but didn’t buy on the web somewhere, are driven by data that Facebook doesn’t own. That outside party, be it Zappos or Walgreens, engages in some data-joining acrobatics to tell Facebook whom to show ads to, but the data itself isn’t shared with the social network; advertisers don’t trust Facebook either.
We track every topic across the internet to identify growing trends. Select a category below to explore our example data.
Your options are companies, products, industries and styles. There’s a free monthly newsletter with “two exponentially growing trends”, a $29/mo newsletter with 7 exponentially growing trends plus 3 “hyper growth” trends, or a $479/mo newsletter with 10 exponential and 5 hyper growth trends.
I think monetised newsletters is a growth trend.
link to this extract
• Oct. 2 2018: Jamal Khashoggi is last seen entering the Saudi consulate in Istanbul. Soon, The Washington Post raises the alarm and demands answers from the Saudi government.
• Oct. 7: Turkey asserts that Khashoggi has been killed inside the Saudi consulate. The Washington Post escalates its relentless coverage, demanding justice for Jamal.
• Oct. 15: A hashtag linking Jeff Bezos to The Washington Post’s reporting appears on Saudi social media: “Boycott Amazon.” The first tweet with the hashtag proclaims: “To the people of the Great Kingdom of Saudi Arabia: The leftist Jeff Bezos is the owner of The Washington Post, the newspaper of evil and betrayal … we have to defend our country and boycott Amazon.”
• Nov. 2: The Washington Post publishes an op-ed by Turkish president Recep Tayyip Erdogan, whom the Saudi government accuses of “politicizing” Khashoggi’s murder.
• Nov. 3: A spike in anti-Bezos tweets begins in Saudi social media showing clear signs of government manipulation. In addition to “Boycott Amazon,” two additional hashtags are launched, one calling for the boycott of Souq.com, an Amazon-owned Middle Eastern e-commerce company. The initial tweet reads: “We as Saudis will never accept to be attacked by The Washington Post in the morning, only to buy products from Amazon and Souq.com by night! Strange that all three companies are owned by the same Jew [sic] who attacks us by day, and sells us products by night!”
• Nov. 4: Several news outlets notice the calls to boycott Amazon on Saudi social media–and at least one points to clear signs of manipulation…
…• Jan. 7, 2019: AMI’s Dylan Howard finally sends a message advising Bezos that they have the story and will publish within three days
• Jan. 9: AMI announces that they have “successfully completed the refinancing of all outstanding debt.” Although the company was emphatic that no foreign investors were directly involved, AMI refused to name their white knights. They have not ruled out that the Saudis may have invested indirectly, as you’ll see below.
• Jan. 10: The National Enquirer publishes a special-edition mid-week issue, an unprecedented 12-page story about Bezos’ affair. Although Bezos is known to most Americans as the world’s richest man and the founder of Amazon, on the magazine cover, beneath the photo of Bezos are the words, “The Owner of The Washington Post”—the same way the Saudi Twitter campaigns had been describing him for months.
When my wife finally hits ‘send’, the text gets sent to the SMS Service Center (SMSC), which then routes the message to me. I’m upstairs and my phone is on, so I receive the text in a handful of seconds, but what if my phone were off? Surely my phone can’t accept a message when it’s not receiving any power, so the SMSC has to do something with the text.
If the SMSC can’t find my phone, my wife’s message will just bounce around in its system until the moment my phone reconnects, at which point it sends the text out immediately. I like to think of the SMSC continuously checking every online phone to see if its mine like a puppy waiting for its human by the door: is that smell my human? No. Is that smell my human? No. Is this smell my human? YESYESJUMPNOW.
The validity period (VP) bytes tell the carrier how long the puppy will wait before it gets bored and finds a new home. It’s either a timestamp or a duration, and it basically says “if you don’t see the recipient phone pop online in the next however-many days, just don’t bother sending it.” The default validity period for a text is 10,080 minutes, which means if it takes me more than seven days to turn my phone back on, I’ll never receive her text.
Something about lost phones and lost texts seems apt here. (Thanks, John Naughton.)
link to this extract
The company has been using the sleep-tracking feature for several months with testers at secret sites around its Cupertino, California, headquarters, according to people familiar with the work. If the functionality is successful in the testing stages, the company plans to add it to the Apple Watch by 2020, according to one of the people. The company has released new versions of the Apple Watch each fall since 2016.
Sleep tracking on the Apple Watch would reduce a competitive advantage that longtime fitness-wearable developer Fitbit has had on the market. Besides Fitbit, Withings – formally known as Nokia Health – also makes sleep-tracking gadgets.
Apple didn’t respond to requests for comment on the company’s plans.
A new Apple Watch wouldn’t be the iPhone maker’s first foray into sleep-tracking hardware. In May 2017, Apple acquired Finnish startup Beddit, which makes a sleep-tracking sensor strip. Apple sells the product on its website under the Beddit brand and launched an updated version at the end of last year.
Apple sells Beddit stuff? They kept that quiet.
link to this extract
I ran a somewhat popular indie site [MetaFilter] for 15 years, the last half or so with ample moderation. But to put the scale of the work in perspective, we were dealing with 10-15 thousand active people daily posting about 3,000 things. Slightly big numbers but still small enough you can wrap your head around them. Mostly day to day we broke up bickering matches between two grad students on the site. And even that was still a drag and after many years doing it I had to hang it up to take a break from the day to day stress.
People often say to me that Twitter or Facebook should be more like MetaFilter, but there’s no way the numbers work out. We had six people combing through hundreds of reported postings each day. On a scale many orders of magnitude larger, you can’t employ enough moderators to make sure everything gets a check. You can work off just reported stuff and that cuts down your workload, but it’s still a deluge when you’re talking about millions of things per day. How many moderators could even work at Google? Ten thousand? A hundred thousand? A million?
YouTube itself presents a special problem with no easy solution. Every minute of every day, hundreds of hours of video are uploaded to the service. That’s physically impossible for humans to watch it even if you had thousands of content mods working for YT full time around the world.
So everyone says “I guess AI will solve it” but then you have all of AI’s problems on top of it. Baby videos get flagged as porn because there’s too much skin tone filling the screen. Subtle forms of abuse aren’t picked up because the patterns don’t exist yet in the AI and every day is a cat-and-mouse game to stay head of AI. AI is prone to the same biases in the creators and will have negative effects down the line.
I don’t know how to counteract the effects of moderation, or how to mitigate the toll it takes on people.
We live in a world where we crave easy answers, because the questions are easy.
link to this extract
Stating that a pesticide is “the last thing you want to think about” as you raise a glass, PIRG wanted to highlight what it sees as a potential danger.
“No matter the efforts of brewers and vintners, we found that it is incredibly difficult to avoid the troubling reality that consumers will likely drink [the weedkiller] glyphosate at every happy hour and backyard barbecue around the country,” said US PIRG Education Fund’s Kara Cook-Schultz, who authored the study.
The 2018 Sutter Home Merlot was the wine with the highest concentration of glyphosate at 51.4 parts per billion, or ppb, while in the beer category, it was Tsingtao from Hong Kong with 49.7 ppb. The American beer with the largest trace was Coors Light with 31.1. ppb.
Organic adult beverages were also implicated in the U.S. PIRG research. For example, A 2016 Inkarri Malbec had 5.3 ppb and a 2017 Samuel Smith Organic Lager, 5.7 ppb.
William Reeves, a toxicologist for Bayer, which now owns Monsanto, accused the group of publicizing misleading information about pesticide residues in food.
“Assuming the greatest value reported, 51.4 ppb, is correct, a 125-pound adult would have to consume 308 gallons of wine per day, every day for life to reach the U.S. Environmental Protection Agency’s glyphosate exposure limit for humans,” he said “To put 308 gallons into context, that would be more than a bottle of wine every minute, for life, without sleeping.”
link to this extract
Errata, corrigenda and ai no corrida: many thanks to the multiple people who explained that U.S. is point-abbreviated while EC or EU is not because when telegrams and headlines used to be in ALL CAPS the story FOREIGN THREAT TO BOMB US would be ambiguous but FOREIGN THREAT TO BOMB U.S. isn’t.
However, we’re not doing stuff in all caps any more. Not telegrams, not headlines.
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.