In an interview, Apple’s Tim Cook says that Stevie Wonder was given a demonstration of the Vision Pro. Thus raising a whole new set of questions. CC-licensed photo by Jon Lebkowsky 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 coming this week at the Social Warming Substack on Friday at 0845 UK time. Free signup.

---

A selection of 9 links for you. Unsigned, unsealed, undelivered. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.

---

Russian crypto criminals helped UK drug gangs launder lockdown cash • BBC News

Dominic Casciani:

»

A multi-billion-dollar money laundering operation that formed when UK gangs were struggling to offload cash during lockdown has been uncovered by the National Crime Agency.

Discovering the Russian-speaking network embedded in the UK’s street drugs market is the biggest success against money laundering in a decade, say investigators.

The global operation, based in Moscow, has been taking dirty money from crime gangs for a fee, and allowing them to exchange it for untraceable cryptocurrency, protecting drugs profits from detection. The network has also been used by the Russian state to fund espionage.

The network stretches across 30 countries, and 84 people have so far been arrested, including 71 in the UK, the NCA and its partners told reporters at a briefing earlier this week.

UK Security Minister Dan Jarvis said the operation “exposed Russian kleptocrats, drug gangs and cyber criminals – all of whom relied on the flow of dirty money”.

On Wednesday, the United States Treasury sanctioned the key figures at the top of the network.

Ekaterina Zhdanova, the head of a Moscow-based cryptocurrency network called Smart, has been identified as being at the heart of the operation. She was previously sanctioned by US authorities in November 2023 for allegedly moving money for Russian elites.

«

Cryptocurrency, you say? How surprising.
unique link to this extract

---

Australia struggling with oversupply of solar power • ABC News

Daniel Mercer:

»

Amid the growing warmth and increasingly volatile weather of an approaching summer, Australia passed a remarkable milestone this week.

The number of homes and businesses with a solar installation clicked past 4 million — barely 20 years since there was practically none anywhere in the country.

It is a love affair that shows few signs of stopping.

And it’s a technology that is having ever greater effects, not just on the bills of its household users but on the very energy system itself.

At no time of the year is that effect more obvious than spring, when solar output soars as the days grow longer and sunnier but demand remains subdued as mild temperatures mean people leave their air conditioners switched off.

Such has been the extraordinary production of solar in Australia this spring, the entire state of South Australia has — at various times — met all of its electricity needs from the technology. What South Australia could not use itself, it exported to other states.

And everywhere, it seems, demand for power from the grid — that is, demand for power not being met by rooftop solar — has fallen to record lows.

But all of this solar is prompting some hard questions, and gnashing of teeth, for one, simple reason — there is, at times, too much solar power in Australia’s electricity systems to handle.

«

0 gigawatts in 2010, past 22GW this year. As problems go, having too much energy is one of the nice ones.
unique link to this extract

---

Tim Cook wants Apple to literally save your life • WIRED

Steven Levy:

»

SL: Will you open up Apple apps like Mail and Messages to other companies to use in their AI systems? How are you thinking about privacy there?

TC: We’ll always consider the privacy implications. We don’t accept that there’s a trade-off between great privacy and great intelligence. Much of Apple Intelligence runs on the device, but for some users we need more powerful models. So we crafted private cloud compute that essentially has the same privacy and security as your device does. We just kept plugging at it until we came up with the right idea.

…SL: When you’re thinking about things late at night, don’t you sometimes ask what it would mean if computers had superhuman intelligence?

TC: Oh, of course. Not just for Apple, but for the world. There’s so much extraordinary benefit for humanity. Are there some things you have to have guardrails on? Of course. We’re very deeply considerate about things that we do and don’t do. I hope that others are as well. AGI itself is a ways away, at a minimum. We’ll sort out along the way what the guardrails need to be in such an environment.

…SL: I heard that Stevie Wonder had a demo of the Vision Pro and loved it. How did that work?

TC: He’s a friend of Apple and it’s great to get feedback from Stevie. And of course his artistry is just unparalleled. One of the common threads running through Apple over time is that we don’t bolt on accessibility at the end of the design process. It’s embedded. So getting his feedback was key.

«

It’s a typically anodyne interview. But I struggle with how Stevie Wonder would benefit from the Vision Pro. In 1999 he considered surgery to restore some simulacrum of sight, but didn’t go ahead with it.
unique link to this extract

---

How close is AI to human-level intelligence? • Nature

Anil Ananthaswamy:

»

will LLMs ever deliver AGI? One point in their favour is that the underlying transformer architecture can process and find statistical patterns in other types of information in addition to text, such as images and audio, provided that there is a way to appropriately tokenize those data. Andrew Wilson, who studies machine learning at New York University in New York City, and his colleagues showed that this might be because the different types of data all share a feature: such data sets have low ‘Kolmogorov complexity’, defined as the length of the shortest computer program that’s required to create them.

The researchers also showed that transformers are well-suited to learning about patterns in data with low Kolmogorov complexity and that this suitability grows with the size of the model. Transformers have the capacity to model a wide swathe of possibilities, increasing the chance that the training algorithm will discover an appropriate solution to a problem, and this ‘expressivity’ increases with size. These are, says Wilson, “some of the ingredients that we really need for universal learning”. Although Wilson thinks AGI is currently out of reach, he says that LLMs and other AI systems that use the transformer architecture have some of the key properties of AGI-like behaviour.

Yet there are also signs that transformer-based LLMs have limits. For a start, the data used to train the models are running out. Researchers at Epoch AI, an institute in San Francisco that studies trends in AI, estimate4 that the existing stock of publicly available textual data used for training might run out somewhere between 2026 and 2032. There are also signs that the gains being made by LLMs as they get bigger are not as great as they once were, although it’s not clear if this is related to there being less novelty in the data because so many have now been used, or something else. The latter would bode badly for LLMs.

Raia Hadsell, vice-president of research at Google DeepMind in London, raises another problem. The powerful transformer-based LLMs are trained to predict the next token, but this singular focus, she argues, is too limited to deliver AGI. Building models that instead generate solutions all at once or in large chunks could bring us closer to AGI, she says.

«

unique link to this extract

---

GenCast predicts weather and the risks of extreme conditions with state-of-the-art accuracy • Google DeepMind

Ilan Price and Matthew Wilson:

»

Because a perfect weather forecast is not possible, scientists and weather agencies use probabilistic ensemble forecasts, where the model predicts a range of likely weather scenarios. Such ensemble forecasts are more useful than relying on a single forecast, as they provide decision makers with a fuller picture of possible weather conditions in the coming days and weeks and how likely each scenario is.

Today, in a paper published in Nature, we present GenCast, our new high resolution (0.25°) AI ensemble model. GenCast provides better forecasts of both day-to-day weather and extreme events than the top operational system, the European Centre for Medium-Range Weather Forecasts’ (ECMWF) ENS, up to 15 days in advance. We’ll be releasing our model’s code, weights, and forecasts, to support the wider weather forecasting community.

«

AI being useful! Overdue!
unique link to this extract

---

What happened to Intel? • The Verge

Sean Hollister:

»

In a way, it all comes down to reversing one mistake, the ultimate bad bet — the one where Intel funded the technology that its competitors used to leap ahead.

Over a decade ago [2012, in fact – Overspill Ed], Intel spent billions investing in Dutch multinational ASML, which is today the most important company in chips. It’s the only firm in the world that manufactures machines capable of pulverizing a ball of tin, using high-power lasers, such that it emits an extremely tight wavelength of ultraviolet light to efficiently carve circuits into silicon wafers, a process known as EUV.

Intel initially believed in the tech, even carving out a $4.1bn stake in the company, then decided not to order the pricey machines. But Taiwan’s TSMC did — and went on to become the undisputed leader in silicon manufacturing, producing an estimated 90-plus% of the world’s “leading-edge logic chips.” Samsung ordered machines, too.

Gelsinger was not shy about calling Intel’s choice “a fundamental mistake” in our 2022 interview. “We were betting against it. How stupid could we be?”

So Gelsinger decided to embrace EUV, while simultaneously giving its technology departments a blank check to leapfrog TSMC. “I said, ‘You have an unlimited budget, and you are going to deliver five nodes in four years. We are going to get back to unquestioned process leadership.’”

«

Five nodes in four years is, translated, five generations of chip in four years. There was no way in the world that Intel was going to achieve that, unless three of them were already made.
unique link to this extract

---

Why did Intel fire CEO Pat Gelsinger? • SemiAccurate

Charlie Demerjian:

»

This difference lead to the company looking pretty bad for the pure finance folk but the technical observers knew better. Things were moving in the right direction more often than not and the painful financial news was simply Pat riding out the sh*tstorm that that was handed to him. If you followed the process side, you undoubtedly know about the 10nm debacle, but did you know that 14nm and even 22nm had many of the same issues? They were hidden but SemiAccurate documented them over the years.

Why was this mess allowed to not only fester but continue and grow? Because the internal incentive structure was so broken that it encouraged employees to lie for profit. Worse yet lies went unpunished. SemiAccurate has many emails, texts, and had conversations about meetings where this happened. An example would be when a design team asked the process side if node XYZ would be ready at time ABC with specs of DEF. Process would say yes it would, no question.

The first problem was that they knew it would not be ready on that date, not meet the intended specs, and usually wouldn’t be close. Design teams knew the other side was lying but what could they do? A few years later the process was indeed late, occasionally partially working, and met every letter of the law that governed bonus structures. Designers would then force a few working devices out to an OEM so that their bonuses, paid if device X shipped in quarter Y, did ship then. Sure yields were financially untenable but their new BMW had heated seats.

This isn’t to put the blame solely on the process side of the company, everyone lied. One great example was when Tim Cook met with Intel folk over their cellular modems. He directly asked someone I won’t name, “Will it be ready in time?”. The Intel exec said, “Yes”. He was lying, everyone on the Intel side knew he was lying but didn’t contradict the boss. From what we understand, Tim Cook also knew he was lying, and we know several Apple personnel in the room definitely knew it was well past a fib. If you have read this far, you understand how that program, and later the entire Apple/Intel relationship ended. It was for cause.

«

In my experience SemiAccurate is an accurate enough name for the content. Some parts of this may be true. The problem is figuring out what.
unique link to this extract

---

EU set to crack down on Asian online retailers Temu and Shein • Financial Times

Paola Tamma and Andy Bounds:

»

The EU is preparing a crackdown on the growing flood of packages from Asian online retailers such as Temu and Shein, following a big increase in ecommerce that largely evades EU custom checks.

Measures under consideration include a new tax on ecommerce platforms’ revenue and an administrative handling fee per item that would make most shipments less competitive, according to five people familiar with the discussions.

European trade commissioner Maroš Šefčovič has said about 4bn lower-value parcels will be flown to the EU this year, almost triple the number in 2022. The sheer volume and the fact that they are under the €150 threshold for custom duties means most are not checked, driving a rise in imports of dangerous goods such as toxic toys.

While the EU executive is targeting the business model of popular online platforms such as China’s Temu and Shein, which was founded in China but is now based in Singapore, no decisions had been taken and any action was complicated by international law, the people said. 

EU officials are worried about the undercutting of European competitors that face higher production costs to adhere to EU standards and the negative impact of cheap imports on high street retailers.

The bloc’s safety authorities have detected a growing number of dangerous and counterfeit goods, many of which are dispatched direct to consumers.

«

Does this count as a trade war/ Perhaps a quiet one?
unique link to this extract

---

April 2024: Ransomware attack has cost UnitedHealth $872m; total expected to surpass $1bn • The Record

Jonathan Greig, in April 2024:

»

The ransomware attack on a company owned by healthcare giant UnitedHealth Group (UHG) has so far caused $872m in losses, according to the corporation’s latest earnings report. 

UnitedHealth owns Change Healthcare, a key cog in the US healthcare industry that was crippled by a ransomware attack in February. Change Healthcare and UHG subsidiary Optum took hundreds of systems offline as a result of the incident and faced criticism from the White House and Congress over its handling of the ransomware attack.

On an earnings call, president and chief financial officer John Rex said the company earned $7.8bn in the first quarter but suffered $872m in “unfavourable cyberattack effects.”

“Of the $870 million, about $595 million were direct costs due to the clearinghouse platform restoration and other response efforts, including medical expenses directly relating to the temporary suspension of some care management activities. For the full year, we estimate these direct costs at $1 billion to $1.15 billion,” Rex said. 

“The other components affecting our results relates to the disruption of ongoing Change Healthcare business. This is driven by the loss of revenues associated with the affected services, all while incurring the support and costs to keep these capabilities fully ready to return to service.”

Depending on the timing of service restoration and a return of previous transaction volumes, the company estimates another $350m to $450m in losses for the rest of the year, Rex added. 

«

Why a story from April? Because on Wednesday night UnitedHealth’s CEO Brian Thompson was shot dead in a targeted hit by a gunman on a New York street. The motive is unknown, but this ransomware attack offers a big possibility. The NYPD doesn’t think the killer was a professional: his first shot hit Thompson in the calf, in front of a camera and a bystander (though he wore a mask).
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