Start Up No.1951: Musk goes mad on the algorithm, TikTok aims at Meta headsets, the lossy compression discussion, and more


In the US, two storefronts are hit by cars.. every day. And the vast majority aren’t ram raids. Yet there’s a simple fix. CC-licensed photo by Scott Hughes 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.


There’s another post due this Friday at the Social Warming Substack at about 0845 UK time.


A selection of 9 links for you. What if you had someone who wanted to be the main chracter, though? I’m @charlesarthur on Twitter. Observations and links welcome.


Twitter is just showing everyone all of Elon’s tweets now • The Verge

Emma Roth:

»

For many of us, Twitter’s “For You” is full of tweets and replies to tweets from Elon Musk. Not everyone is getting the Elon-first feed, but on Monday afternoon, more than a few people noticed something was different.

Several of us here at The Verge are seeing more Musk replies than usual, and I personally counted five at the very top of my feed, with many more sprinkled in between tweets from other users. The same is true for some accounts that don’t even follow Elon Musk.

This comes just days after Musk complained that his tweets weren’t getting enough views — and even fired an engineer over it.

As reported by Platformer’s Zoë Schiffer and Casey Newton, internal Twitter data indicates that while Musk’s account rose to peak popularity in search rankings in April 2022, engagement has since dropped significantly, and engineers found no issue with Twitter’s algorithm.

Over the weekend, Musk said Twitter rolled out some sort of change to fix this “visibility” issue, with the billionaire CEO stating that 95% of his tweets weren’t “getting delivered.” I’m not sure if this is at all related to this Elon-filled feed, but I’m hoping Twitter fixes this issue soon — unless the new mandate is to get the boss more views by any means necessary.

«

Tom Warren had a tweet which showed this perfectly.

Meanwhile, the abrupt changes to the API (with a limited free tier and more pricey paid tier/s) has been delayed again because everyone showed so much “enthusiasm” for it. That’s not how I usually understand enthusiasm to work. This discussion on Hacker News of what people think of the upcoming API changes is quite educational: companies probably aren’t going to wear it.

Pulling the company out of this self-imposed dive at the ground is going to be a hell of an achievement – if that’s what happens.
unique link to this extract


Benefits of improved brakes on oil trains miscalculated by US • Claims Journal

Matthew Brown:

»

President Donald Trump’s administration miscalculated the potential benefits of putting better brakes on trains that haul explosive fuels when it scrapped an Obama-era rule over cost concerns, The Associated Press has found.

A government analysis used by the administration to justify the cancellation omitted up to $117m in estimated future damages from train derailments that could be avoided by using electronic brakes. Revelation of the error stoked renewed criticism Thursday from the rule’s supporters who called the analysis biased.

Department of Transportation officials acknowledged the mistake after it was discovered by the AP during a review of federal documents but said it does not change their decision not to install the brakes.

Safety advocates, transportation union leaders and Democratic lawmakers oppose the administration’s decision to kill the brake rule, which was included in a package of rail safety measures enacted in 2015 under President Barack Obama following dozens of accidents by trains hauling oil and ethanol in the US and Canada.

…The deadliest happened in Canada in 2013, when an unattended train carrying crude oil rolled down an incline, came off the tracks in the town of Lac-Megantic and exploded into a massive ball of fire, killing 47 people and obliterating much of the Quebec community’s downtown.

There have been other fiery crashes and fuel spills in Alabama, Oregon, Montana, Virginia, West Virginia, North Dakota and Illinois.

«

Of course when the rule was suggested, lobbyists for the rail and oil industries said it would be too expensive yet ineffective. Seems to me that not setting off massive balls of fire is desirable. (Another example of regulation that you need.) (Thanks drew for the link.)
unique link to this extract


TikTok’s parent takes on Meta in battle for virtual-reality market • WSJ

Meghan Bobrowsky and Stu Woo:

»

Two years ago, ByteDance bought Pico, a Chinese startup that makes VR headsets. That launched a new front in the Chinese company’s competition with Meta, whose Instagram and Facebook services have been battling for users and advertising dollars against TikTok as the short-video app soared in popularity.

Pico’s headset shipments have since jumped, turning it into a small but fast-rising No.2 to Meta in the global market, according to industry data, even though Pico doesn’t sell its consumer headsets in the US. 

Mark Zuckerberg in 2021 renamed Facebook to Meta in part to reflect his bet on the metaverse, a more immersive version of the internet to be experienced largely through virtual-reality headsets. The company has been spending heavily on that concept. In its latest quarterly results, Meta said there were more than 200 apps on its VR devices that have generated over $1m each in sales, although total revenue in Meta’s Reality Labs segment was down 17% in the quarter due to lower Quest 2 headset sales.

Meta held 90% of the market share about a year ago, according to research firm International Data Corp. By the third quarter of 2022—the latest period for which data is available—its market share had dropped to about 75%. Market share for Pico more than tripled over the same period to about 15%. No other VR headset maker held more than 3% of the market. 

Meta’s headset shipments in the third quarter declined 48% from a year earlier, IDC’s data shows. ByteDance’s Pico was the only headset maker to increase shipments, in a market that was estimated to be worth $4bn as of 2022.

«

Feels like it’s going to be another of those inside China/outside China tech markets, like smartphones and search engines.
unique link to this extract


How Spotify’s podcast bet went wrong • Semafor

Max Tani:

»

By 2021, Spotify had paid to sign some of the biggest names in podcasting, and it was ready to start squeezing its competitors.

The Swedish audio streaming giant had spent hundreds of millions of dollars to purchase podcast production companies and big name creators in the hopes of luring new subscribers to the platform. Now, Spotify chief content officer Dawn Ostroff — a TV veteran most famous for bringing Gossip Girl to the CW — was ready to stop many of these creators and companies from sharing podcasts on Apple and Amazon, and keep the content exclusively on Spotify.

Then Bill Simmons sent an email to her boss.

Simmons had sold the sports and pop culture audio empire The Ringer to Spotify a year earlier for $200m. Now he wrote Spotify CEO Daniel Ek to argue for keeping the Ringer’s mass audience on Apple and its advertising revenue, driven by the explosion of sports betting.

Simmons won the argument. But that 2021 dispute exposed deep questions about the strategy behind Spotify’s billion-dollar bet on podcasting. In January, Spotify pushed out Ostroff and canceled nearly a dozen shows at its highest-profile podcast investment, the studio Gimlet. Podcasting was a “big drag on our business in 2022,” the company’s chief revenue officer said earlier this month.

“In hindsight, I probably got a little carried away and overinvested relative to the uncertainty we saw shaping up in the market,” Ek said on an earnings call in January. “So we are shifting to focus on tightening our spend and becoming more efficient.”

«

Lots of noise but.. doesn’t seem to add up to much. The biggest claim, that Joe Rogan’s podcast contract ends this year, was flatly contradicted by Spotify after the article published.
unique link to this extract


Adversary drones are spying on the US and the Pentagon acts like they’re UFOs • The Drive

Tyler Rogoway:

»

We may not know the identities of all the mysterious craft that American military personnel and others have been seeing in the skies as of late, but I have seen more than enough to tell you that it is clear that a very terrestrial adversary is toying with us in our own backyard using relatively simple technologies—drones and balloons—and making off with what could be the biggest intelligence haul of a generation. While that may disappoint some who hope the origins of all these events are far more exotic in nature, the strategic implications of these bold operations, which have been happening for years, undeterred, are absolutely massive.

Our team here at The War Zone has spent the last two years indirectly laying out a case for the hypothesis that many of the events involving supposed UFOs, or unidentified aerial phenomena (UAP), as they are now often called, over the last decade are actually the manifestation of foreign adversaries harnessing advances in lower-end unmanned aerial vehicle technology, and even simpler platforms, to gather intelligence of extreme fidelity on some of America’s most sensitive warfighting capabilities. Now, considering all the news on this topic in recent weeks, including our own major story on a series of bizarre incidents involving US Navy destroyers and ‘UAP’ off the Southern California coast in 2019, it’s time to not only sum up our case, but to discuss the broader implications of these revelations, what needs to be done about them, and the Pentagon’s fledgling ‘UAP Task Force’ as a whole.

«

Makes far better sense than absolutely any other hypothesis out there.
unique link to this extract


New iMac not expected to launch until late 2023 at earliest • MacRumors

Joe Rossignol:

»

Apple has no plans to launch a new 24-inch iMac until late 2023 at the earliest, according to Bloomberg’s Mark Gurman. In his newsletter today, he reiterated his expectation that Apple will skip updating the iMac with the M2 chip and instead wait to release a model with the M3 chip, which has yet to be announced.

“I haven’t seen anything to indicate there will be a new iMac until the M3 chip generation, which won’t arrive until the tail end of this year at the earliest or next year,” wrote Gurman. “So if you want to stick with the iMac, you’ll just have to sit tight.”

Apple’s M3 chip is expected to be manufactured based on TSMC’s latest 3nm process, providing additional performance and power efficiency improvements. The M3 chip is also expected to be used in a new MacBook Air rumored to launch by the second half of 2023, and potentially in future versions of the 13-inch MacBook Pro and Mac mini. By comparison, the M2 chip is built on TSMC’s second-generation 5nm process.

Apple last updated the iMac in April 2021 with the M1 chip and a new ultra-thin design available in seven colors, including green, yellow, orange, pink, purple, blue, and silver.

«

Why the rush? Nobody’s doing mission-critical video processing on an iMac. The M2 Mac mini plus a screen of your choice will do the job. The iMac is positioned as a low-end consumer product.
unique link to this extract


Journalism is lossy compression. Isn’t everything? • Whither news?

Jeff Jarvis:

»

To predict the next, best word in a sequence is a different task from finding the correct answer to a math problem or verifying a factual assertion or searching for the best match to a query. This is not to say that these functions cannot be added onto large-language models as rhetorical machines. As Google and Microsoft are about to learn, these functions damned well better be bolted together before LLMs are unleashed on the world with the promise of accuracy.

When media report on these new technologies they too often ignore underlying lessons about what they say about us. They too often set high expectations — ChatGPT can replace search! — and then delight in shooting down those expectations — ChatGPT made mistakes!

[Ted] Chiang [in his New Yorker article, linked yesterday] wishes ChatGPT to search and calculate and compose and when it is not good at those tasks, he all but dismisses the utility of LLMs. As a writer, he just might be engaging in wishful thinking.

…We are early our progression of learning what we can do with new technologies such as large-language models. It may be too early to use them in certain circumstances (e.g., search) but it is also too early to dismiss them.

«

Certainly he’s right that journalism is quick to praise things and quick to shoot them down, but to complain about Chiang’s “lossy compression” analogy because we don’t document absolutely everything about every event (whether in libraries or news journalism) seems weak. And it wasn’t journalists who pushed Microsoft to include ChatGPT in Bing. (Thanks wendyg, again, for the link.)
unique link to this extract


7-Eleven to pay $91M to Bensenville man who lost both legs in storefront crash • NBC Chicago

»

A 57-year-old suburban man who became a double amputee after a car pinned his legs against the front of a Bensenville 7-Eleven will receive a $91m payout from the convenience store chain.

The 2017 crash was one of the thousands of similar incidents identified in discovery for the case, collisions that frequently resulted in crippling injuries, said James Power, one of the attorneys representing the plaintiff, who wished to be identified as “Carl” to avoid drawing attention to his windfall.

The settlement was approved by a Cook County judge on Monday, the day the case had been set for a jury trial, and is the largest pre-trial settlement in a personal injury case in state history, Power said. Joseph Power Jr., Larry Rogers Jr. and Louis Berns also represented the plaintiff.

“(Carl) was in shock. Just silent amazement,” Power said, noting that his client had been hospitalized for a month after the crash and now walks using prosthetic legs. “He has been through a lot of pain.”

The case was the first in which attorneys had access to some 15 years of reports from 7-Eleven, which identified some 6,253 storefront crashes at 7-Eleven stores across the country, Power said. Data from a previous lawsuit against the company identified another 1,525 crashes between 1991 and 1996.

The crashes could have been prevented if 7-Elevens had installed bollards — thick posts anchored in the ground — between storefronts and parking spaces, Power said.

…“We have evidence 7-Eleven had been getting sued for these kinds of incidents going back to 1990,” Power said, noting the total number of storefront crashes identified in the case indicated that, on average, a car crashed into a 7-Eleven store about once a day.

«

An amazing instance either of corporate indifference or corporate lack of communication. It probably would have cost a lot less than $91m to fit the bollards. And the stats suggest on average there are *two* storefront crashes across the US per day. (Thanks Patrick for the link.)
unique link to this extract


Please don’t trust Twitter with your credit card info • Garbage Day

Ryan Broderick:

»

Imperfect social media tools gave anyone the ability to publish their thoughts and communicate. And the men who got rich from building those tools suddenly realized they didn’t want to have to read what the rest of us had to say (about them). And they have spent the years since the start of the pandemic losing billions of dollars building embarrassing failed attempts at retrofitting the internet into a country club.

They create invite-only “social audio” apps, they try and convince us to buy cryptocurrency or expensive VR helmets or change the algorithms that power our apps to only prioritize their own content. They bend themselves into pretzels because they can’t seem to grasp that they aren’t cool or popular and they won’t ever be, no matter how much money they make.

Even in the techno-feudalist future they all salivate over, as long as the serfs can make content, which is, ironically enough, the only way these guys make their money, we will still be able to post about how much they suck.

«

Ryan constantly has the viewpoint that gets underneath the flotsam on the internet (even though often he’s writing about the flotsam – such as in this case, about Elon Musk’s idiotic sort-of plans about ignoring blocklists not created by Twitter Blue users.)
unique link to this extract


• Why do social networks drive us a little mad?
• Why does angry content seem to dominate what we see?
• How much of a role do algorithms play in affecting what we see and do online?
• What can we do about it?
• Did Facebook have any inkling of what was coming in Myanmar in 2016?

Read Social Warming, my latest book, and find answers – and more.


Errata, corrigenda and ai no corrida: none notified

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.