Start Up No.1936: Apple’s headset details (maybe) leak, AI plagiarism-bot?, the trouble with insulating homes, and more

If you were looking for something you can put on hot chips andon hot chips, we have good news. CC-licensed photo by jeffreyw 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. Yes, you put it on chips. I’m @charlesarthur on Twitter. Observations and links welcome.

How Apple’s upcoming mixed-reality headset will work • Bloomberg via Yahoo

Mark Gurman:


Apple’s long-anticipated mixed-reality headset is an ambitious attempt to create a 3D version of the iPhone’s operating system, with eye- and hand-tracking systems that could set the technology apart from rival products.

The roughly $3,000 device, due later this year under the likely name of Reality Pro, will take a novel approach to virtual meetings and immersive video, aiming to shake up a VR industry currently dominated by Meta Platforms Inc. It’s a high-stakes gambit for Apple, which is expanding into its first major new product category since releasing a smartwatch in 2015, and the company needs to wow consumers.

Apple is pushing into an uncertain market with a premium-priced product. The company’s 1,000-person-plus Technology Development Group has spent more than seven years on the project, and Apple is counting on it to become a new revenue source — especially with sales growth poised to stall this year.

But virtual reality has proven a challenge for the biggest titans of technology. Though some projections have the industry topping $100 billion by the decade’s end, headsets are still seen as niche items — and Meta has lost billions on its efforts.

Apple’s goal is to bring something new to the table. The eye- and hand-tracking capabilities will be a major selling point for the device, according to people familiar with the product, which is expected to cost roughly twice the price of rival devices. Its core features will include advanced FaceTime-based videoconferencing and meeting rooms.

The headset also will be able to show immersive video content, serve as an external display for a connected Mac, and replicate many functions of iPhones and iPads.


Lots more detail about how it will work (external cameras, hand sensors). Still find it hard to believe the price tag, and that Apple really thinks there’s enough demand to make this even halfway viable.
unique link to this extract

CNET’s AI journalist appears to have committed extensive plagiarism • Futurism

Jon Christian:


a new Futurism investigation found extensive evidence that the CNET AI’s work has demonstrated deep structural and phrasing similarities to articles previously published elsewhere, without giving credit. In other words, it looks like the bot directly plagiarized the work of Red Ventures competitors, as well as human writers at Bankrate and even CNET itself.

Jeff Schatten, a professor at Washington and Lee University who has been examining the rise of AI-enabled misconduct, reviewed numerous examples of the bot’s apparent cribbing that we provided. He found that they “clearly” rose to the level of plagiarism.

We asked Schatten what would happen if a student turned in an essay with a comparable number of similarities to existing documents with no attribution.

“They would be sent to the student-run ethics council and given the repeated nature of the behavior would almost certainly be expelled from the university,” he replied.

The bot’s misbehavior ranges from verbatim copying to moderate edits to significant rephrasings, all without properly crediting the original. In at least some of its articles, it appears that virtually every sentence maps directly onto something previously published elsewhere.


To be honest, though it would be nice if it were blatant, this is pretty thin gruel. The topics and sentences that they pick are the sort where you would be able to feel your soul draining away out of your ears. “How to sign up for low-balance alerts.” “How to avoid overdraft and NSF fees.” I say it’s good to make computers write this sort of junk to save humans the psychic pain.

But I understand the motivation to stomp on CNet’s private equity owners by any means available.
unique link to this extract

M2 Max: under the hood with the latest Apple Silicon • Creative Strategies

Ben Bajarin:


Throughout these past five days, and using a tool to monitor M2 Max’s wattage use and frequency of the (efficiency) e-cores and (performance) p-cores, I believe Apple made two subtle changes. First, they made high-performance cores more efficient. When designing chips to be more efficient, and thus draw less power, one goal is to optimize to perform their workload and then get to a zero state as quickly as possible. While monitoring my Mac with M2 Max and my old Mac with M1 Max, I noticed the high-performance cores reach a zero state, and interestingly the M1 Max cores do not.

Perhaps more to the point, the same observation can be made while M2 Max handles a more robust workload. You can see the high-performance cores during the performance benchmark both reach a higher frequency of ~3.5 GHz and drop to a zero state whereas the M1 Max performance cores do not throttle as high or as low. This appears to be the first time Apple M-series silicon was designed to have the high-performance cores reach a zero state.

Apple didn’t stop at just making M2 Pro/Max high-performance cores more performant AND more efficient. They also made the efficiency cores more performant! I didn’t notice it until I looked at the above image more closely, and looked at the log files of core frequency more closely over the course of my working day.  The efficiency cores on M2 Max can operate at a higher frequency than M1 Max.  In the above image, during a GeekBench CPU benchmark, the efficiency cores throttled to ~2.5 GHz on M2 Max. While M1 Max, during the same workload, the efficiency cores throttled to ~2.0 GHz.

These subtle generational improvements in the M2 CPU design are the reasons I think Apple is able to tout around an hour of battery life increase over previous M1 Apple Silicon.


Doesn’t this mean that eventually the performance and efficiency cores will meet in the middle, at the nirvana of performant efficiency and efficient performance?
unique link to this extract

Insulation only provides short-term reduction in household gas consumption • University of Cambridge


Insulating the lofts and cavity walls of existing UK housing stock only reduces gas consumption for the first year or two, with all energy savings vanishing by the fourth year after a retrofit, according to research from policy experts at the University of Cambridge.

The latest study is the first to track in detail household gas use across England and Wales for at least five years both before and after insulation installation.

Researchers analysed gas consumption patterns of more than 55,000 dwellings over twelve years (2005-2017), and found that cavity wall insulation led to an average 7% drop in gas during the first year. This shrank to 2.7% in the second, and by the fourth year, any energy savings were negligible.   

Loft insulation was half as effective as cavity wall, with an initial fall in gas consumption of around 4% on average, dropping to 1.8% after one year and becoming insignificant by the second year. For households with conservatories*, any gains in energy efficiency disappeared after the first year.  

The findings suggests that when it comes to home insulation there may be a significant ‘rebound effect’: any savings through energy efficiency get cancelled out by a steady increase in energy use.**

The UK Treasury recently announced some £6 billion in funding to reduce the energy consumption of buildings and industry by 15% over the next eight years, with a major focus on insulation retrofits across the residential sector. 

Researchers behind the study, published in the journal Energy Economics, say it is extremely difficult to identify specific causes of the ‘rebound effect’ they found, but behaviours such as turning up the heating, opening windows in stuffy rooms or building extensions may all contribute.


We like to be comfortable, and get used to it! Though this is the most perverse, disheartening news of the day.
unique link to this extract

Twitter drops from “tiny” to “tinier” as a referral source for news publishers • Nieman Journalism Lab

Laura Hazard Owen:


“The big, scary, existential question is, will social media continue to be a traffic source for a news organization? Or will it become just a storytelling platform or just a marketing platform?”

That’s the question that one publisher asked in a Digiday piece about how Twitter is declining as a traffic source for publishers.

The question was recently echoed — and answered — by Semafor editor-in-chief Ben Smith. “When you get beyond the drama of Twitter and the flickers of life on your Facebook feed, what we’re seeing is the end of the whole social media age in news,” he wrote on Christmas Eve 2022.

Twitter never drove much traffic to news publishers. Back in 2016, social analytics firm (which was acquired by Automattic in 2021) found that “Twitter generates 1.5% of traffic for typical news organizations.”


And the drop has been pretty dramatic, mostly from Twitter getting rid of the (human-curated) Moments feature. This from the Digiday piece:


Twitter referral traffic to a dozen major publishers’ websites declined, on average, by 12% in December 2022 compared to November 2022, according to an analysis by Similarweb, a data analytics company that monitors web traffic. Some publishers — such as The Washington Post, The Wall Street Journal, CNN, The New York Times, USA Today, the BBC and Yahoo — each saw referral traffic from Twitter fall between 10% and 18% month over month.


All this can be true, and yet Twitter’s relevance isn’t about the traffic it drives, but the discussion it ferments. Nadhim Zahawi’s peculiar tax returns gained notoriety on Twitter in a matter of a few days, as did Boris Johnson’s arrangements for a loan with a chum who then got a plum job at the BBC. Twitter’s value, such as it is, is in being the market square where the news gets discussed, rather than made.
unique link to this extract

I’m a noted music critic. can AI do my job? • TIDAL Magazine

Simon Reynolds:


I made the instructions [to ChatGPT] as simple and straightforward as possible:

Write an essay in the style of music critic Simon Reynolds that expresses skeptical views about A.I. taking over the role of the music critic.

Within seconds, the program served up 200 words of disconcertingly clear and well-organized argument. The effect was at once mind-blowing and underwhelming. Although millions of my own sentences can be found on the internet, the program proved unable to duplicate any stylistic mannerisms. The argument itself struck me as averagely intelligent, making entry-level points about how A.I.-generated prose is necessarily deficient in empathy and nuance, and how it would lack the unique and personal perspective of a human critic. (I had to give the chatbot points for self-awareness, at least.) Similar formulations on the same topic, substituting the names of music journalists with highly individual prose voices, produced equally neutral and characterless results.

The program is addictive, with tremendous scope for time-suck. You keep thinking that if you bang away at it, trying out ever more outlandish approaches or finely tuned commands, it’ll suddenly and dramatically improve. But I found that the “voice” remained consistent: earnest, plodding, attuned to bland generalities rather than arresting specifics, and irritatingly fair-minded. Not promising attributes for a critic! 

The even-handedness is probably the most significant defect, when it comes to the prospect of a chatbot usurping people like me. Asked to compose a fierce critique of a particular record, it produced the prim rejoinder “I am not programmed to write negative reviews …  my primary goal is to be helpful and supportive.” This is not the primary goal or proper role of a critic: brutal honesty, to the point of unkindness, is closer. But the truth is that ChatGPT is just as incapable of writing a rave review. Instead, it surveys the range of existing viewpoints and gathers a balanced array of pros and cons. It sits on the fence.


It also makes stuff up, which isn’t really what you want. When it starts getting all its facts right, watch out. Until then, it’s going to make things much worse.
unique link to this extract

GPU cooler tested with ketchup, potatoes, and cheese as thermal paste • Tom’s Hardware

Aaron Klotz:


The test system used a Radeon R7 240 with a 30W TDP, with temperature readings from a five-minute run of Furmark. As such, these tests aren’t a great indicator of the long-term feasibility of using a potato to cool your chip, so here’s a statement of the obvious: Don’t try this at home.

The user shared a spreadsheet showing the findings, including 22 different tested thermal “paste” materials. The list includes several standard thermal pads of different sizes, including Arctic TP2 0.5mm, 1mm, 1.5mm, Arctic TP3 1mm, 1.5mm, EC360 Blue 0.5mm, EC360 Gold 1mm, 0.5mm EKWB, and Thermal Grizzly Minus 8 thermal pads.

With those relatively safe choices out of the way, next up is the unusual substances not designed for thermal conductivity in a GPU application, including double-sided aluminum copper tape, cheese slices, potato slices, ketchup, copper paste, and Penaten Creme for diaper rashes. The enthusiast also used a broad range of toothpastes, including a few brands you might not recognize, like Amasan T12, Silber Wl.paste, Kupferpaste, and a no-named toothpaste with no branding. 


The ketchup did really quite well: as well as some of the commercial products.
unique link to this extract

The chess world’s new villain: a cat named Mittens • WSJ

Andrew Beaton and Joshua Robinson:


Mittens—or technically the chess bot known as Mittens—might look cute. Her listed chess rating of a single point seems innocuous. But her play over the past few weeks, which has bedeviled regular pawn-pushers, grandmasters, and champions who could play for the world title, is downright terrifying. And as it turns out, people are gluttons for punishment.

Since introduced this bot with the avatar of a cuddly, big-eyed kitten on Jan. 1, the obsession with playing her has been astonishing. Mittens has crashed the website through its sheer popularity and helped drive more people to play chess than even “The Queen’s Gambit.” has averaged 27.5 million games played per day in January and is on track for more than 850 million games this month—40% more than any month in the company’s history. A video that American grandmaster Hikaru Nakamura posted to YouTube titled “Mittens The Chess Bot Will Make You Quit Chess” has already racked up more than three million views. 

“This bot is a psycho,” the streamer and International Master Levy Rozman tweeted after a vicious checkmate this month. A day later, he added, “The chess world has to unite against Mittens.” He was joking, mostly. 

Mittens is a meme, a piece of artificial intelligence and a super grandmaster who also happens to reflect the broader evolution in modern chess. The game is no longer old, stuffy and dominated by theoretical conversations about different lines of a d5 opening. It’s young, buzzy and proof that cats still rule the internet.

…Mittens is designed to be skillful enough to beat the best chess players on the planet but uses particularly grueling tactics. Becker thought it would be “way more demoralizing and funny” if, instead of simply smashing opponents, Mittens grinded down opponents through painstaking positional battles, similar to the tactics Russian grandmaster Anatoly Karpov used to become world champion.

It hasn’t been difficult for Becker to see the reactions to his masterpiece. Nakamura, who could manage only a draw against Mittens, bluntly said in a video, “This cat is extremely patient, which is kind of annoying. I’m not going to lie.”


Masochism and high-level chess go together well, it seems.
unique link to this extract

Is life in the UK really as bad as the numbers suggest? Yes, it is • Financial Times

Tim Harford:


if the forecasts are bad, it is the scene in the rear-view mirror that is truly horrifying. The British economy is in a generation-long slough of despond, a slow-burning economic catastrophe. Real household disposable income per capita has barely increased for 15 years.

This is not normal. Since 1948, this measure of spending power reliably increased in the UK, doubling every 30 years. It was about twice as high in 1978 as in 1948 and was in touching distance of doubling again by 2008, before the financial crisis intervened. Today, it’s back at those pre-crisis levels.

It’s worth lingering on this point because it is so extraordinary. Had the pre-crisis trend continued, the typical Brit would by now be 40% richer. Instead, no progress has been made at all. No wonder the Institute for Fiscal Studies is now talking of a second lost decade.

Go back and look for historical precedents for this, and you will not find much. In the National Institute Economic Review, economic historians Nick Crafts and Terence Mills examined the growth in labour productivity over the very long run. (This is defined as the total output of the UK economy divided by the total number of hours worked; labour productivity is closely connected to material standards of living.) They do find worse runs of performance — 1760 to 1800 was not much fun — but none within living memory. Nowhere in 260 years of data do they find a sharper shortfall from the previous trend. The past 15 years have been a disappointment on a scale that previous generations of British economists could hardly have imagined.


What Harford doesn’t explain is why things are so bad. Brexit is a significant part of it, and the conditions that led people to vote that way are another. The absurd imbalance towards the financial services industry has to be another.
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 Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

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