The arrival of AI has meant that job applications are being overwhelmed – so recruiters are having a rethink about how to screen applicants. CC-licensed photo by Innov8social 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 9 links for you. Help unwanted. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
Tech tries a new message: you’d love data centres if China wasn’t making you hate data centres • Business Insider
Peter Kafka:
»
Everyone* hates data centres now.
How did that happen?
There are lots of theories: People are worried about the environment. Or electric bills. Or their jobs. Or imperious Big Tech billionaires ordering them around. All of the above.
Also: It’s China’s fault.
That last one doesn’t come up in polling. But it’s quite popular on certain corners of the pro-AI internet.
The theory spelled out: China wants Americans to turn against AI data centres, so that America will fall behind China in the AI race. So it is using the internet to brainwash Americans to hate data centres. A psyop, per prominent VC Garry Tan.
Is that possible? Maaaaaaaybe: In June, OpenAI said it had evidence some China-linked accounts were trying to use ChatGPT to generate anti-AI content they could distribute on social media. But that same report said it didn’t have any evidence that those efforts had shifted public opinion.
And after celebrity investor Kevin O’Leary accused opponents of his Utah data centre of working on behalf of China, he later acknowledged that he had “no evidence” that was true, and walked the claim back (so did Fox News).
It would not be surprising if it eventually turns out that there was a coordinated social media campaign to try to influence how Americans feel about AI and data centres. All kinds of people use social media to try to influence sentiment. And after the 2016 election, social media platforms started paying extra attention to documenting covert efforts that tried to do that.
It’s also possible that it’s easier for bad actors to act badly on social media right now. In the years after Elon Musk bought Twitter, social media platforms have gutted the trust and safety teams that used to monitor those platforms. Outside watchdogs, like the Stanford Internet Observatory, have been shuttered or neutered.
Then again: everyone who owns a big social media platform in the US is also pouring many tens of billions into AI, much of which is going to US data centre projects.
«
Tech bros get terribly upset when their pet projects aren’t greeted adoringly by the wider populace (see also: crypto) and China is an obvious rival here. Suspicion is far from proof, though.
unique link to this extract
Yahoo, the internet’s ‘OG’, wants to win over Gen Z • Financial Times
Daniel Thomas:
»
Yahoo wants to convince a new generation of users that one of the internet’s oldest services can be one of the new AI winners, with the Apollo-backed company seeking to develop its brand alongside the launch of an in-house AI tool and consumer products.
Chief executive Jim Lanzone told the FT that the company’s heritage gave it credibility with younger users not alive when it launched in 1994 in a “vintage way”. He described Yahoo as the internet’s “OG” — a slang term meaning “original gangster”.
He argued that familiarity has become an advantage as trust in some newer technology groups has eroded. “There’s something inherently trustworthy about [being] the ‘OG’ of the space,” he added. “We are the original guide to the internet . . . We want to get the brand back out there again.”
Lanzone, who joined Yahoo from Tinder in 2021 shortly after Apollo acquired the company from Verizon, has sought to recast the internet pioneer as a leaner consumer tech group focused on its biggest brands.
The push comes as Yahoo seeks to prove it can carve out a role in the AI era despite having spent years overshadowed by larger rivals.
While the company still reaches hundreds of millions of users through products including Finance, Sports, News and email services, it remains a challenger in search and other core internet services dominated by Google and more AI-native competitors.
Yahoo plans to fully roll out an AI-powered “answer engine” called Scout later this year alongside several editorial and consumer products, part of a broader effort to return the business to growth.
«
New CEO? ✔️
“Build on our brand”? ✔️
“Still got millions of users”? ✔️
Just about to roll out a new product using AI to generate news/answers/whatever? ✔️
The Yahoo story just goes around and around. Every few years it changes hands and someone new comes aboard and says the same things. Then it continues to lose money, there’s a restructuring and after a bit it changes hands and…
unique link to this extract
It should be harder to apply for a job. No, really • WIRED
Kate Taylor:
»
“A year ago, every recruiter would tell me: We want to make it as easy as possible to apply for jobs,” offering candidates a “seamless experience,” Ophir Samson, the head of voice AI for Greenhouse, tells WIRED. Samson joined the recruiting platform earlier this year, when it acquired his startup, which used AI to conduct job interviews. “What they got was 2,000 applicants in 24 hours for a job. That is a shitty experience for everyone.”
Good applicants can’t stand out, while bad candidates slip through the cracks. Recruiters, Samson says, face the tortuous task of judging thousands of candidates, every single day. Instead of taking the time to search LinkedIn for dream employees, they spend hours in applicant tracking systems, tweaking filters to try to determine whose applications are worth a 30-second skim.
Now, recruiters are telling Samson, “actually, we kind of want friction,” he says. “The friction is good. We want to make it harder.”
The desire to remove pain points in the application process makes sense: If the highest number of qualified candidates apply for a role, it stands to reason that a company can maximize the chances of finding the best possible fit.
…Many recruiters are turning to AI tools themselves, hoping to wrangle the sea of applications into something more manageable. Companies have used automated applicant tracking systems to screen résumés for at least a decade; now, many are using agents to conduct first-round interviews with candidates.
“We frequently hear from recruiters: They have 1,000 applications, but they know that only 30 of them are serious,” Greenhouse’s Samson says. An AI interview can sort the unserious from the serious, he says, offering recruiters context beyond a résumé.
While Samson says that these AI interviews are not friction for friction’s sake, they have the added bonus of weeding out bots and halfhearted candidates. Curran says she expects some companies to turn to knock-out questions that cut down the application pool via strict criteria. Others might add skills testing earlier in the process.
«
Seriously, the whole thing sounds broken, and needs radical rethinking about how to find humans to do jobs. If they still want that.
unique link to this extract
The biggest carbon polluters are to blame for half of US western water lost to climate change, research says • Inside Climate News
Jake Bolster:
»
Some of the largest fossil fuel companies on Earth are disproportionately responsible for the water crisis in the American West, according to a first-of-its-kind study published Tuesday.
From 2014 to 2024, climate change sapped over a third of the region’s typical annual snowpack and decreased streamflows by about 13 percent, the authors found in the paper, which appeared in Nature’s Communications Earth and Environment journal. About half of those losses are attributable to emissions from “carbon majors,” 122 fossil fuel companies and cement producers, the study concluded.
That carbon-major pollution has drained the West of trillions of gallons of water, about the full capacity of Lake Mead, the largest reservoir in the U.S., the study found.
“There are millions of people that rely on these water resources,” said Carly Phillips, a senior research scientist at the Union of Concerned Scientists and one of the study’s co-authors. “The consequences of these changes are really evident for people on the ground.”
…By subtracting the carbon majors’ emissions from the total emissions between 2014 to 2024, and modeling each scenario against a hypothetical Earth without climate change, Phillips and her co-authors determined the role of carbon majors’ emissions in regional snowpack and riverflow declines.
This practice, known as attribution science, represents a growing body of work that links extreme weather events and rapid environmental changes to the greenhouse gas effect, about three-quarters of which comes from burning fossil fuels. In the West, as temperatures rise, more precipitation may fall as rain instead of snow, which can melt earlier, scrambling water forecasts.
«
Two planned datacentres will have higher UK carbon emissions than ExxonMobil, analysis finds • The Guardian
Pippa Neill:
»
Analysis by the non-profit group Foxglove reveals that when fully operational, the Wapseys Wood datacentre in Buckinghamshire and the Quest Park datacentre in Bedfordshire will use 1.3GW of power and will produce more than 4.5m tonnes of carbon emissions.
ExxonMobil emitted 3.9m tonnes in 2023, according to the global energy thinktank Ember.
AI datacentres use a massive amount of electricity to power servers and cooling equipment, and to run facility systems. There are currently 315 datacentres in the queue to connect to the UK’s electricity grid, representing 73GW of demand. This is almost double the peak winter energy demand for the entire UK.
Because of the long wait to connect to the grid, the Wapseys Wood and Quest Park datacentres are seeking to build their own on-site gas-fired power stations.
Based on government figures on the carbon intensity of electricity generation, Foxglove has calculated these projects will produce 4.577m tonnes of carbon emissions per year.
A government spokesperson said these figures were “misleading” because of the assumption that both projects will run at 100% utilisation from day one, which they said was “not accurate”.
Tom Hegarty, head of communications at Foxglove, said assuming a 100% utilisation rate followed standard industry practice.
Kierra Box, a trade and environmental regulation campaigner at Friends of the Earth, said the aim was for datacentres to increase load and run at full capacity, and so it was “valid to look at those top level predictions”.
«
Those datacentres had better be running 24/7 on finding ways to remove CO2 from the atmosphere cheaply and quickly, because otherwise they’re just making the whole problem worse.
unique link to this extract
Logging in was never supposed to be this complicated • The Atlantic
Will Oremus:
»
Lately, the process of accessing apps and websites has grown dizzying. “I’m in password hell,” a colleague confided to me recently. “Every login attempt is playing the lottery.” The problem is not just the sheer number of digital accounts, all of which require their own password, ideally unique and unguessable (which usually also means un-memorizable). It’s the pileup of solutions to that problem: all of the different layers of software offering to remember your passwords, the six-digit codes sent to your phone, the authenticator apps, the passkeys.
Browsers and operating systems compete with third-party password managers to generate, store, and autofill your credentials. Most of my logins live in 1Password, but others are stored in Apple’s Passwords app (usually by accident), and still others were created via “single sign-on” features such as those offered by Google and Facebook. I vaguely recall certain apps—Facebook is one—prompting me many years ago to copy down a set of “login recovery codes” that would become extremely important should I ever get locked out of my account. Whatever device I stored them on was probably wiped and sold long ago.
…Cybersecurity experts will tell you that passkeys—which typically entail logging in to apps or websites with a face or fingerprint scan instead of a password—are the future. When they work properly, they’re both painless and secure. “It may get to a point where we’re all using passkeys and we don’t even know we’re using passkeys,” Lorrie Cranor, a computer-science professor at Carnegie Mellon University who researches privacy and cybersecurity, told me. “I just say ‘Log in’ and I smile at it, and it gets my passkey, and I’m logged in.”
But that future isn’t here yet, and in the present, passkeys have added a new layer of frustration. Part of the problem is that virtually no one seems to know what they are. “The fact that even I have trouble explaining them should be a clue,” Cranor said. Her past research on the rollout of two-factor authentication suggests that most people tend to resist new security features for as long as they can. Companies can nudge them, but for websites and apps trying to attract and retain users, pushing people to adopt an unfamiliar authentication technology risks scaring them away.
«
Apple cuts jobs in Siri, Vision Pro immersive video, and gaming teams • Bloomberg via Business Standard
Mark Gurman:
»
Apple Inc. is cutting jobs across teams responsible for the Siri digital assistant and the Vision Pro headset, part of an effort to focus on new devices and artificial intelligence.
Apple is also laying off employees in its Intelligent Systems Experience, a team within its software engineering organisation responsible for some of the AI features on its devices, according to people with knowledge of the matter.
As part of the changes, the company is largely shutting down a Vision Pro team focused on gaming for the headset and reducing the size of a unit responsible for producing the device’s immersive video content, said the people, who asked not to be identified because the personnel moves haven’t been announced.
Altogether, the cuts affect more than 200 people, with about 100 positions eliminated from the Vision Pro organization and another 100 from the Siri and software teams.…The Vision Pro changes were described to employees as an effort to realign priorities based on how customers are using the device and to better direct resources toward other future products.
Immersive video has been a key selling point for the Vision Pro since its launch. But producing each 3D video requires multiple crews of Apple employees and contractors, and an episode in an immersive video series can cost several million dollars to film. Those expenses have strained the business given the limited number of active Vision Pro users.
The Vision Pro was seen as a groundbreaking product when it launched in February 2024, but the cumbersome nature of the headset and its now-$3,699 price have made it a hard sell.
Apple isn’t abandoning immersive video entirely. The company plans to produce a smaller number of videos in-house while encouraging third-party companies to develop content for the device. Although Apple has maintained a team focused on gaming features for the Vision Pro, the headset has gained limited traction among gamers.
«
It’s still astonishing how Apple hasn’t figured out the cost-effective way to make immersive video.
unique link to this extract
Opinion: Twitter broke the Left. X will break the Right • The New York Times
Ezra Klein:
»
In his 1922 book “Public Opinion,” Walter Lippmann argued that the central problem of democracy is that the image of the world inside our heads differs endlessly and unpredictably from the world as it actually exists.
“What each man does is based not on direct and certain knowledge, but on pictures made by himself or given to him,” Lippmann wrote. “If his atlas tells him that the world is flat, he will not sail near what he believes to be the edge of our planet for fear of falling off. If his maps include a fountain of eternal youth, a Ponce de Leon will go in quest of it. If someone digs up yellow dirt that looks like gold, he will for a time act exactly as if he had found gold. The way in which the world is imagined determines at any particular moment what men will do.”
There is no solution to this problem. The blazing otherness of reality will always defy the simplifications we impose upon it. But the representations we hold in our heads can be better or worse. The Mercator projection that recognizes a round earth is superior to a map that imagines oceans ending in an edge. The Mercator has its problems — Greenland appears as large as Africa, when it is 1/14th the size — but it is an improvement on its predecessors.
My contention about Twitter then and X or BlueSky or Truth Social now is that they shape the pictures in our heads for the worse. The common concern here is disinformation, the lies that spread like wildfire across these platforms. And that is a real problem. But I agree with Henry Farrell, a political scientist at Johns Hopkins, when he wrote, “The more important change is to our beliefs about what other people think, which we perpetually update based on social observation.”
What we consider reasonable to say or even think is built on what those around us say and think. We are all creatures of convention. These platforms surround us with unreasonable and antisocial voices, which are almost always more interesting than the reasonable and mannered ones. That can lead in useful directions, as every society preserves some injustice through convention and courtesy. But it can also lead in bizarre and destructive directions, as attention begins revolving around people ill suited to leadership and political elites lose sight of how off-putting their ideas have become.
…The problem the right faces in 2026 and will face in 2028 is closer to the problem the left faced after 2020: Through Musk, it dominates X, and so it has fully exposed itself to the reality-warping, prudence-dissolving effects of the X hothouse. In dismantling Twitter’s guardrails against outright racism and antisemitism, Musk has filled the right with Groypers and white supremacists who will be a constant problem in Republican politics for years. Democrats paid a price for forming their politics on Twitter; people on the right will pay a price for forming theirs on X.
«
Robot’s viral “face-covering” sprint at 2nd World Humanoid Robot Games is “spontaneous”, not designed by developers • Global Times
»
A humanoid robot that grabbed a 400-meter title at the 2nd World Humanoid Robot Games (WHRG) in Beijing while appearing to “cover its face with both hands” developed the unusual posture by itself during sprinting, its developers said, after clips of the run drew wide attention online.
Tien Kung Omni, built by the Beijing-based X-Humanoid, won the 400-meter run small-group final on Sunday in 45.66 seconds. Rather than swinging its arms forward and back as a human sprinter does, the robot ran with its center of gravity pitched forward and its arms held high near its face, swinging them widely as its hips rotated from side to side — a gait viewers likened to “a shy person hiding behind their hands.”
The posture was not what engineers had in mind. “We initially designed Tien Kung Omni’s gait around a human-style arm swing,” Han Gang, a motion control algorithm expert at the company, told the Global Times. In a simulated learning environment, the team gave the robot continuous rewards, and through repeated iteration on the data it gradually settled into a posture that suited it better. “It feels that this is more comfortable,” Han said.
The reason is a physical constraint, Han explained. Swinging the arms rapidly places a heavy load on the shoulder joints, which can cause them to heat up and drive up energy consumption — leaving the robot unable to sustain the full 400m. Running with its hands near its face allows the robot to complete the distance quickly while keeping its legs powerful and its arms under low load.
Different distances pose different problems, Han added. In endurance events such as the humanoid half-marathon, engineers have to pay closer attention to heat dissipation in the hip joints after long periods of running. Over shorter distances such as the 100m and 400m, hip overload is much less of a concern.
«
We forget that the way we run is inherently limited by all sorts of things about our bodies that we’re completely unaware of because we accept them as “how it is”. But we learn to run, and everyone runs slightly differently, and there are better and worse ways for humans to run.
So it makes sense to let the machines figure it out for themselves too.
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
The Overspill: when there's more that I want to say 