Now using them an average of 4.5 hours per day for “digital entertainment”. CC-licensed photo by Esther Vargas 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. Don’t mention the health! I’m @charlesarthur on Twitter. Observations and links welcome.
Did YouTube help China’s anti-Hong Kong propaganda go viral? • Quartz
A Chinese-made propaganda video about the Hong Kong protests went viral, apparently thanks to YouTube’s algorithm. The video—called “Who’s behind Hong Kong protests?”—argues that US agents are stirring protests in Hong Kong, and has more than half a million views on YouTube. It was created by China’s state broadcaster, China Global Television Network.
On August 24th, YouTube recommended it six times more than the average video when users searched for “Hong Kong protestors,” according to AlgoTransparency, making it one of the most-recommended videos on that subject. The most-viewed video on Hong Kong protests by the Wall Street Journal, BBC, and New York Times all had fewer views as of October 31st 2019.
The Chinese government has repeatedly claimed that the US is behind protestors, though it has failed to present any evidence. Attempts to present Hong Kong protests as an American uprising often involve outright falsehoods, such as Chinese media reports that a toy weapon was a US army grenade launcher.
AlgoTransparency, created by former YouTube engineer Guillaume Chaslot, analyzes videos recommended on thousands of channels daily. The data is based on blank profiles; as YouTube recommends videos based on users’ history, different individuals may have different recommendations. Overall, YouTube’s algorithm was created to optimize for watch time which, Chaslot has shown, often leads to YouTube recommending more extreme videos in a bid to capture attention. “You can go from more radical video to more radical video,” says Chaslot. “There’s a rabbit-hole effect.”
YouTube said it disagreed with Algotransparency’s methodology, data, and conclusions, and was unable to reproduce Algotransparency’s results.
How surprising, given that you can just look at the number of times the video has been watched, and note that it’s not available inside China.
unique link to this extract
‘Alarming’ loss of insects and spiders recorded • BBC News
Insects and spiders are declining in forests and grasslands across Germany, according to new research.
Scientists have described the findings as “alarming”, saying the losses are driven by intensive agriculture.
They are calling for a “paradigm shift” in land-use policy to preserve habitat for the likes of butterflies, bugs and flying insects.
Recent studies have reported widespread declines in insect populations around the world.
The latest analysis, published in the journal, Nature, confirms that some insect species are being pushed down the path to extinction. It is becoming clearer and clearer that the drivers of insect decline are related to farming practices, said Dr Sebastian Seibold of the Technical University of Munich in Freising, Germany.
“Our study confirms that insect decline is real – it might be even more widespread then previously thought considering, for example, that also forests are experiencing declines in insect populations,” he told BBC News.
Consumers spend 4.5 hours per day on digital entertainment • Midia Research
The average consumer now spends an estimated 30.73 hours per week on digital entertainment. This works out to approximately 4.5 hours per day.
Allocating eight hours of work and seven hours of sleep, consumers then have about 4.5 hours per day left to do things like eat, commute, socialise, run errands, and conduct general life admin.
Watching TV/ streaming is the most prevalent activity among consumers, followed by ‘doing nothing’ and then ‘listening to music’.
With every minute of spare attention competing brands and interests are fighting for, consumers with the available spend are choosing to use it to enable their own choice and control over when and what they see, do, and listen to.
There is nowhere else to grow, engagement-wise, other than fighting for digital attention share amongst existing propositions.
A data point; they’re holding a conference to discuss how best to do this on November 20 in London.
unique link to this extract
Apple Watch forced Fitbit to sell itself • Above Avalon
Apple didn’t just steal customers away from Fitbit. In such a scenario, Fitbit may actually have had a chance to survive as the company could have had a means to respond competitively. Apple ended up doing something that ultimately proved far worse for Fitbit. The Apple Watch altered the fundamentals underpinning the wrist wearables industry. This left Fitbit unable to remain relevant in a rapidly-changing marketplace.
Apple placed a bet that wrist real estate was being undervalued. The Swiss had dropped the ball and were primarily selling the wrist as a place for intangibles with high-end mechanical watches. Instead of following Fitbit and selling a $99 dedicated fitness tracker, Apple looked at the wrist as being a great place for additional utility beyond just telling time or tracking one’s fitness and health. Apple turned health and fitness tracking from a business into a feature. The Apple Watch redefined utility on the wrist.
This change led to consumers wanting more from wrist wearables. Apple Watch established a stronghold at the premium end of the market. Taking a page from its product strategy playbook, Apple then methodically began to lower entry-level Apple Watch pricing, which had the impact of removing oxygen from increasingly lower price segments. Fitbit was squeezed as the company had no viable way to compete directly with Apple Watch.
The point about the Swiss essentially selling intangibles – “you’ll be able to pass it to your kids!” – is a good one. And turning someone’s USP into just another feature is, yes, a great way to kill them. (Dropbox has so far managed to avoid that, but it’s close.)
unique link to this extract
Apple said to partner with Valve to make AR headset • Apple Insider
Citing supply chain reports, Digitimes claims that Apple has partnered with the game developer, Valve, in order to produce an Augmented Reality headset. As well as developing the headset with Steam, Apple is said to be using familiar suppliers Pegatron and Quanta Computer to assemble the device. It’s not expected to be released before the second half of 2020 at the earliest.
Valve wouldn’t be providing the VR experiences itself, but could use the Steam distribution system to distribute them. Back in 2017, Valve leveraged Apple’s macOS High Sierra’s eGPU and virtual reality support to bring a beta version of its SteamVR to the Mac.
Then for macOS Mojave, Apple worked with both Valve and HTC to support their HTC Vive Pro headset for VR.
This timescale reported by Digitimes is later than previously predicted by analyst Ming-Chi Kuo.
Or as 9to5 Mac calls it, “the ever-questionable Digitimes”. This sounds like supply chain noise, but – to mix metaphors – probably not smoke without some fire in there. The timing sounds about right too. The big question remains whether there’s anything compelling there. I’m not sure games quite cut it.
unique link to this extract
Samsung is shutting down its custom CPU division • Android Authority
Samsung filed a Worker Adjustment and Retraining (WARN) letter in Texas, according to The Statesman, notifying the state that 290 employees will be laid off as part of its CPU unit being shut down. The layoffs reportedly go into effect from December 31.
The Korean manufacturer confirmed the news to Android Authority, while also explaining the reasoning behind the decision.
“Based upon a thorough assessment of our System LSI [large scale integration – ed] business and the need to stay competitive in the global market, Samsung has decided to transition part of our US-based R&D teams in Austin and San Jose,” the company told us in a statement, adding that it remained committed to its US workforce.
It’s unclear what exactly this means for Samsung’s custom CPU plans for 2020 and beyond. Samsung’s Mongoose CPU cores were mostly used in its flagship Exynos processors, starting with 2016’s Exynos 8890 in the Galaxy S7. But our own testing with the Galaxy S10 series revealed that while the Exynos chipset offered better single core performance than the Snapdragon variant, the Snapdragon version beat it in most other key areas.
If Samsung is indeed abandoning its custom CPU cores for flagship phones, then it’s likely that the firm will adopt Arm CPUs or semi-custom versions of these CPUs for future devices. Huawei currently uses Arm CPUs in its flagships, while Qualcomm uses tweaked versions of these cores in its Snapdragon 800-series of high-end processors. Qualcomm in particular previously used fully custom CPU designs for several years before transitioning to a semi-custom model instead.
Surprising: this means that Apple and Huawei would be the only companies designing their own CPU cores rather than using the ARM basics. Samsung is doing a fair bit of retrenching lately, though. Perhaps it decided it could never get ahead of Qualcomm.
Releasing Spleeter: Deezer Research source separation engine • Deezer
While not a broadly known topic, the problem of source separation has interested a large community of music signal researchers for a couple of decades now. It starts from a simple observation: music recordings are usually a mix of several individual instrument tracks (lead vocal, drums, bass, piano etc..). The task of music source separation is: given a mix can we recover these separate tracks (sometimes called stems)? This has many potential applications: think remixes, upmixing, active listening, educational purposes, but also pre-processing for other tasks such as transcription.
From a Mix of many instruments, a source separation engine like Spleeter outputs a set of individual tracks or stems.
Interestingly, our brain is very good at isolating instruments. Just focus on one of the instrument of this track [the Rolling Stones’s ‘Angie’] (say the lead vocal for instance) and you will be able to hear it quite distinctively from the others. Yet that’s not really separation, you still hear all the other parts. In many cases, it may not be possible to exactly recover the individual tracks that have been mixed together. The challenge is thus to approximate them the best we can, that is to say as close as possible to the originals without creating too much distortions.
Open source and fast. Look forward to more remixes!
unique link to this extract
Undercover reporter reveals life in a Polish troll farm • The Guardian
It is as common an occurrence on Polish Twitter as you are likely to get: a pair of conservative activists pouring scorn on the country’s divided liberal opposition.
“I burst out laughing!” writes Girl from Żoliborz, a self-described “traditionalist” commenting on a newspaper story about a former campaign adviser to Barack Obama and Emmanuel Macron coming to Warsaw to address a group of liberal activists.
“The opposition has nothing to offer. That’s why they use nonsense to pull the wool over people’s eyes,” replies Magda Rostocka, whose profile tells her almost 4,400 followers she is “left-handed with her heart on the right”.
In reality, neither woman existed. Both accounts were run by the paid employees of a small marketing company based in the city of Wrocław in southwest Poland.
But what the employee pretending to be Magda Rostocka did not know is that the colleague pretending to be Girl from Żoliborz was an undercover reporter who had infiltrated the company, giving rare insight into the means by which fake social media accounts are being used by private firms to influence unsuspecting voters and consumers…
…The accounts produced both leftwing and rightwing content, attracting attention, credibility and support from other social media users, who could then be rallied in support of the company’s clients.
“The aim is to build credibility with people from both sides of the political divide. Once you have won someone’s trust by reflecting their own views back at them, you are in a position to influence them,” said Wojciech Cieśla, who oversaw the investigation in collaboration with Investigate Europe, a consortium of European investigative reporters.
“Reading these communications, you can see how the leftwing and rightwing accounts would receive their daily instructions, how they would be marshalled and directed like two flanks of the same army on a battlefield.”
Which sort of explains why Twitter doesn’t need to turn down political ads: they’re already there, earning it money from fake accounts run by real people which attract other ad money. But here’s a neat twist to this story: “A majority of Cat@Net’s employees are understood to be disabled, allowing the company to derive substantial public subsidies from Poland’s National Disabled Rehabilitation Fund.”
unique link to this extract
Did Bitcoin price manipulation fuel 2017 surge? Study says yes • Bloomberg
Matthew Leising and Matt Robinson:
A Texas academic created a stir last year by alleging that Bitcoin’s astronomical surge in 2017 was probably triggered by manipulation. He’s now doubling down with a striking new claim: a single market whale was likely behind the misconduct, seemingly with the power to move prices at will.
One entity on the cryptocurrency exchange Bitfinex appears capable of sending the price of Bitcoin higher when it falls below certain thresholds, according to University of Texas Professor John Griffin and Ohio State University’s Amin Shams. Griffin and Shams, who have updated a paper they first published in 2018, say the transactions rely on Tether, a widely used digital token that is meant to hold its value at $1.
“Our results suggest instead of thousands of investors moving the price of Bitcoin, it’s just one large one,” Griffin said in an interview. “Years from now, people will be surprised to learn investors handed over billions to people they didn’t know and who faced little oversight.”
Tether rejected the claims, with General Counsel Stuart Hoegner arguing in a statement that the paper is “foundationally flawed” because it is based on an insufficient data set. The research was probably published to back a “parasitic lawsuit,” the general counsel added…
…The authors examined Tether and Bitcoin transactions from March 1, 2017 to March 31, 2018, concluding that Bitcoin purchases on Bitfinex increased whenever Bitcoin’s value fell by certain increments. Griffin and Shams didn’t name the entity on Bitfinex that they think was responsible. They shared their updated research with Bloomberg News.
“This pattern is only present in periods following printing of Tether, driven by a single large account holder, and not observed by other exchanges,” they wrote in their new peer-reviewed paper, set to be published in a forthcoming Journal of Finance. “Simulations show that these patterns are highly unlikely to be due to chance. This one large player or entity either exhibited clairvoyant market timing or exerted an extremely large price impact on Bitcoin that is not observed in aggregate flows from other smaller traders.”
In other words: lots of people got really, really suckered. And those who cling to it are still being suckered due to the lack of transparency in those exchanges.
unique link to this extract
The trillion dollar lawsuit, part one • (Two Truths and a Take) Substack
In terms of monetary evaporation, the accusations here are ten times the size of WeWork, and at least ten times as interesting.
In my opinion, this is the most interesting story in tech this year and nothing else comes close.
In this and next week’s newsletter, I’m going to walk you through the accusations being made in this complaint, along with context from others in the community who’ve been paying attention, so you can understand how truly shocking this all is. It’s a complex story, so I’ve broken down the allegations into three parts. The first two we’ll talk about today, and the third next week:
Allegation #1: The 2017 Bitcoin Bubble was market manipulation, and Tether was how they did it
Allegation #2: Tether became a systemically important, money laundering conduit for the crypto ecosystem
Allegation #3: They might’ve gotten away with it, too, if they hadn’t gotten robbed while busy scamming
To be clear: everything I’m about to talk about in the next two weeks are allegations.
The mechanism he explains for how it works is simple, and persuasive.
unique link to this extract
Errata, corrigenda and ai no corrida: none notified
re. CPU cores:
1- I don’t think Huawei is doing its own. They’ve got custom stuff elsewhere on the chip (AI, 4G/wifi, not sure about imaging), but AFAIK the cores themselves are vanilla ARM; Anandtech uses them as benchmarks of a pure implementation of ARM’s designs.
2- The other likely candidate for custom cores would be Qualcomm, and I’m not sure about them. I’m sure they significantly scaled back their core customizations (their chips used to be completely unrelated to ARMs sample designs), I’m not sure they stopped them entirely. Again, there’s custom stuff elsewhere on the SoC, but specifically in the cores, at best little, at worst none (I think they still do something w/ the I/O and RAM)
It’s a credit to ARM that OEMs are not finding obvious optimizations; plus the pace has picked up between process shrinks, and other parts of the SoC are more in need of attention (AI, data, imaging in and out, drivers).
Several second-tier OEMs had announced custom SoCs (based on vanilla cores) but gave up, e.g Xiaomi; Samsung only ever sold Exynos to 1 OEM for a minor model (something from BBK IIRC) but that may be due to deals with Qualcomm, not Exynos’ intrinsic qualities.
Makes life easier for hackers too, not having 2 entirely different S and Note models to bootload, root, custom-distro…
PS: the bit about Samsung’s cores beating Qualcomm in single core is interesting. Apple has been making hay by heavily pushing that specific metric when other stuff is probably more important (multi-core, radio, RAM, storage, imaging, AI…).
There’s a bit of a tunnel effect going on. Android OEMs would love to one-up Apple, but don’t have the PR power to re-orient the discourse to those more relevant metrics, so Samsung kinda tried to fight on Apple’s single-core turf. It didn’t pan out because a) Apple is way ahead and b) it doesn’t matter that much in practice, users even testers where noticing the multicore & battery life compromises more than the single-core gains.
Re. Fitbit: I’m hoping this is mostly a case of using pocket money to preempt the competition snatching up a vaguely-maybe someday semi-valuable asset.
Google’s insistence on going full-OEM and mainstream puzzles me:
1- Android’s ecosystem has glaring issues, none of them attributable to OEMs shyness w/ innovation, premium and inversely value… Android does not need more OEMs
2- Google’s forte is subsystems, not whole products. Let them work on apps, dev chains, cameras…
3- Android needs Long-Term Support (HW and SW) on the phone side, and a non-health Smartwatch for the 80+% of people who either don’t exercise or don’t track it but “social” like crazy. Let Google do that…
Xiaomi just released Another phone that looks nice.
The branding is messy. It’s a Mi not Redmi when it really is a direct upgrade to Redmi’s wildly successful Note: no moving parts, IR blaster, FM radio, probably SD slot (unconfirmed yet, but at 40€ for an extra 128GB of internal flash and 2GB RAM, no that important; Apple charges 3x that and has the gall to offer a 64GB base model). Has with AMOLED, better specs, much better camera than the Redmi Note.
Called the CC9 Pro in China, might be Mi 9 Lite Pro in Europe since the CC9 is the Mi 9 Lite. Xiaomi copying Apple again, this time w/ ridiculous names ;-p
I’m curious how good the main camera will be. It has a ridiculous pixel count, but also a very large sensor – which is way more important. Also, 3 extra cameras (depth, wide angle, 5x telephoto). Apparently there’s OIS and laser autofocus on all, which is excellent and has been sorely missing on Xiaomi stuff since 2yrs ago when they started going for pixel count instead of quality.
At 350€ for the base model, +40 for more RAM and storage, +50 for even better camera, this might become my recommendation for amateur photographers. Few of those, but they’ve been forced to buy premium up to now.
Oh, the Premium Edition already garnered DxO’s top honors ex-aequo with a Huwaei: https://www.dxomark.com/category/mobile-reviews/
Curious how he standard edition would fare…
Aaaaand apparently released tomorrow as the Mi Note 10 in Europe. Impressive how the China to EU delay is going from months to zero. And as long as you’re playing alt-naming bingo, no point in trying to be logical I guess ^^
OK, the pics are as awesome as expected, even in low light.
That phone is interesting if it’s a calculated move and not some accident:
1- it’s a top notch camera phone at $350, which should steal some steam from the premium segment for which cameras are the major selling point
2- it shows some serious segmentation, “CC” stands for “Creative and Cool”; clearly the rest of the specs apart for the camera are midrange. It takes balls to do an half-flagship like that (CPU performance is half a flagship’s, screen is 2.5K not 4K, no stereo..). It’s actually kind of weird Qualcomm has a SoC that good at pictures that isn’t wildly premium.
3- it points at the challenge for OEMs maybe not being grabbing share from each other, but convincing their own low-end users to move upmarket. If you take the Redmi Note as a baseline (which you should), one issue is why not a basic Redmi (much lower but still good-enough performance, slightly worse camera), and the other issue is what would you get for paying more. The basic CC9 brings creature comforts (AMOLED, slightly better everything); this CC9 Pro also, and adds that incredible camera, the Note Pro better gaming. All 3 are feasible upsells to the huge 2yo-Redmi-Note population.
4- We’ll see what Xiaomi comes out with next for gamers and premium all-rounder. They apparently have cancelled/postponed several devices, I’d guess because of their wholesale switch to AMOLED and better cams. Right now their lineup and branding is a mess (6 versions of the Mi 9, all good value but with no defining feature set: some have SD and jack some not, some have a periscope cam…)