Weather forecasting can be done more accurately and more quickly with using machine learning systems, Google DeepMind has shown. CC-licensed photo by Chic Bee 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 10 links for you. Bright prospects. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. Observations and links welcome.
Here’s how violent extremists are exploiting generative AI tools • WIRED
David Gilbert:
»
For years, Big Tech platforms have worked hard to create databases of known violent extremist content, known as hashing databases, which are shared across platforms to quickly and automatically remove such content from the internet. But according to Hadley, his colleagues are now picking up around 5,000 examples of AI-generated content each week. This includes images shared in recent weeks by groups linked to Hezbollah and Hamas that appear designed to influence the narrative around the Israel-Hamas war.
“Give it six months or so, the possibility that [they] are manipulating imagery to break hashing is really concerning,” Hadley says. “The tech sector has done so well to build automated technology, terrorists could well start using gen AI to evade what’s already been done.”
Other examples that researchers at Tech Against Terrorism have uncovered in recent months have included a neo-Nazi messaging channel sharing AI-generated imagery created using racist and antisemitic prompts pasted into an app available on the Google Play store; far-right figures producing a “guide to memetic warfare” advising others on how to use AI-generated image tools to create extremist memes; the Islamic State publishing a tech support guide on how to securely use generative AI tools; a pro-IS user of an archiving service claiming to have used an AI-based automatic speech recognition (ASR) system to transcribe Arabic language IS propaganda; and a pro-al-Qaeda outlet publishing several posters with images highly likely to have been created using a generative AI platform.
«
Nothing is bringing iMessage to its Android phone • The Verge
Wes Davis:
»
Nothing Phone 2 owners get blue bubbles now. The company shared it has added iMessage to its newest phone through a new “Nothing Chats” app powered by the messaging platform Sunbird. The feature will be available to users in North America, the EU, and other European countries starting this Friday, November 17th.
Nothing writes on its page that it’s doing this because “messaging services are dividing phone users,” and it wants “to break those barriers down.” But doing so here requires you to trust Sunbird. Nothing’s FAQ says Sunbird’s “architecture provides a system to deliver a message from one user to another without ever storing it at any point in its journey,” and that messages aren’t stored on its servers.
Marques Brownlee has also had a preview of Nothing Chats. He confirmed with Nothing that, similar to how other iMessage-to-Android bridge services have worked before, “…it’s literally signing in on some Mac Mini in a server farm somewhere, and that Mac Mini will then do all of the routing for you to make this happen.”
Nothing’s US head of PR, Jane Nho, told The Verge in an email that Sunbird stores user iCloud credentials as a token “in an encrypted database” and associated with one of its Mac Minis in the US or Europe, depending on the user’s location, that then act as a relay for iMessages sent via the app. She added that, after two weeks of inactivity, Sunbird deletes the account information.
But you’re still giving them access to your iCloud account to make this work, and as we’ve all learned over the years, companies don’t always do what they say they will. It’s worth reviewing Sunbird’s privacy policy and keeping a very skeptical mind about it.
«
Sunbird doesn’t explain how it does this. My understanding is that iMessages require key exchange, and that the private/public keypair is generated by the device itself. How does Sunbird generate an appropriate hardware key for the Android device? At Pocket Lint, Jason Cipriani wags a big finger and says no, don’t do this: it requires Sunbird signing into your iCloud account on a Mac it controls:
»
“You’re more or less giving Sunbird access to your entire Apple ID/iCloud account, and if you’re someone who uses Apple’s services, that’s a scary thought.”
«
Also, jeepers, people: just use WhatsApp, or Signal. Platform-specific messaging apps are so 2010s.
unique link to this extract
Robotic putting greens, mixed reality, loud spectators: this is golf?! • WIRED
Steven Levy:
»
Cameron Young slides a driver from his bag. He stares at a hole referred to as Texas Hill Country. It’s new to him—a par 4 with sand hazards and rough to avoid. The 26-year-old is in the top 20 in the Official World Golf Ranking, but he’s not sure how to proceed. He turns to his companion, former pro Roberto Castro. “What’s going on here?” Young asks.
Castro consults with their caddie and reports, “It’s 312 to that bunker there.”
Young makes clean contact. The ball lofts skyward.
But there’s no sky above him. On this steamy day in late October, Young is in an air-conditioned soundstage on the back lot of Universal Studios in Orlando, Florida. The building once hosted Nickelodeon TV shows. The “caddie” Castro consulted is virtual—it lives on a 15-inch tablet. The tee is on a patch of natural grass the width of a large mattress. It sits atop wooden pallets on a concrete floor.
Young’s golf ball hits a billboard-sized screen 35 yards away. The dimpled sphere falls meekly to the ground, while up on the giant display its virtual successor continues its flight. A phalanx of supersensitive radar trackers and hi-res cameras sends data to a bank of computer servers that calculate velocity and spin to show how the ball will bounce and where it will ultimately settle on the vista of the screen.
Young’s ball lands in the digital rough. He walks over to a tray of two-inch-high Bermuda grass mixed with rye. The screen now shows him closer to his goal, an 8-iron away. He swings, the ball thuds against the display again, and seconds later his virtual ball lands just outside the green.
…Many pro golfers practice using room-sized simulators in their personal gym, and weekend warriors commonly visit golf centers with plenty of tech. That’s not what Young is up to. He’s testing a system for real competition that will be aired on prime time, with $20m of prize money at stake. He’s one of 24 pros, including golf legends Tiger Woods and Rory McIlroy, who are involved in the most ambitious effort yet to merge e-gaming and actual pro sports. It’s called TGL, allegedly not an acronym for The Golf League, but three TV-friendly letters that don’t mean anything.
TGL’s first event will take place on January 9 inside a $50m–plus, custom-built arena with an inflatable dome in Palm Beach Gardens, Florida. A 200,000-pound [91 tonne] turntable will support an 800,000-pound [363 tonne] green that will shape-shift to give each hole its character. A 4K screen will rival the goliath displays of Taylor Swift concerts. The stands will accommodate around 1,600 live spectators, who are encouraged to boisterously violate golf’s finicky silence rule. Players themselves will be mic’d up, in hopes that their trash talk might go viral online.
«
Strange things that the combination of money, TV and empty airtime will make people try.
unique link to this extract
Andreessen Horowitz invests in Civitai, which profits from nonconsensual AI porn • 404 Media
Emanuel Maiberg:
»
Andreessen Horowitz, also known as a16z, the influential Silicon Valley venture capital firm that was an early investor in Facebook, Lyft, and other tech giants, has invested in Civitai, a giant platform for sharing AI models that enables and profits from the creation of AI generated nonconsensual sexual images of real people. That includes launching a feature where people can list “bounties” for others to create AI models of specific targets.
Civitai said that it raised $5.1m in a seed funding round led by a16z.
A16z’s official website, which includes a jobs board with open positions at companies in its portfolio, currently lists five jobs at Civitai. According to a16z’s site, these jobs were posted more than 30 days ago.
A16z regularly announces investments the company is making on its site, but has not publicly announced its investment in Civitai yet.
A16z did not respond to a request for comment. When asked about a16z’s investment in Civitai over Discord, a community engagement manager at Civitai told 404 Media that “There will be a press release/announcement shortly.” Civitai then published a press release confirming the investment minutes after 404 Media reached out for comment.
In August, 404 Media published an investigation into Civitai, which explained how the platform works, and enables the creation of AI-generated nonconsensual sexual images, and profits from it. Civitai allows users to share modified models of the open source text-to-image AI tool Stable Diffusion. These modified models are often trained on images of celebrities, influencers, YouTubers, and athletes, almost exclusively women, to recreate their likeness. Those models can then be combined with AI models that are trained on porn in order to instantly generate nonconsensual sexual images.
«
You know, I’m beginning to think that a16z is a bit skeevy.
unique link to this extract
Google DeepMind’s weather AI can forecast extreme weather faster and more accurately • MIT Technology Review
Melissa Heikkilä:
»
In research published in Science on Tuesday, Google DeepMind’s model, GraphCast, was able to predict weather conditions up to 10 days in advance, more accurately and much faster than the current gold standard. GraphCast outperformed the model from the European Centre for Medium-Range Weather Forecasts (ECMWF) in more than 90% of over 1,300 test areas. And on predictions for Earth’s troposphere—the lowest part of the atmosphere, where most weather happens—GraphCast outperformed the ECMWF’s model on more than 99% of weather variables, such as rain and air temperature
Crucially, GraphCast can also offer meteorologists accurate warnings, much earlier than standard models, of conditions such as extreme temperatures and the paths of cyclones. In September, GraphCast accurately predicted that Hurricane Lee would make landfall in Nova Scotia nine days in advance, says Rémi Lam, a staff research scientist at Google DeepMind. Traditional weather forecasting models pinpointed the hurricane to Nova Scotia only six days in advance.
“Weather prediction is one of the most challenging problems that humanity has been working on for a long, long time. And if you look at what has happened in the last few years with climate change, this is an incredibly important problem,” says Pushmeet Kohli, the vice president of research at Google DeepMind.
Traditionally, meteorologists use massive computer simulations to make weather predictions. They are very energy intensive and time consuming to run, because the simulations take into account many physics-based equations and different weather variables such as temperature, precipitation, pressure, wind, humidity, and cloudiness, one by one.
GraphCast uses machine learning to do these calculations in under a minute. Instead of using the physics-based equations, it bases its predictions on four decades of historical weather data. GraphCast uses graph neural networks, which map Earth’s surface into more than a million grid points. At each grid point, the model predicts the temperature, wind speed and direction, and mean sea-level pressure, as well as other conditions like humidity. The neural network is then able to find patterns and draw conclusions about what will happen next for each of these data points.
«
Surprise! Historical data is a good predictor of future weather patterns. Unfortunately “It still lags behind conventional weather forecasting models in some areas, such as precipitation, Dueben says”. In other words, it’s not going to take over rain prediction – the thing we really want – just yet.
unique link to this extract
Chatbots may ‘hallucinate’ more often than many realise • The New York Times
Cade Metz:
»
a new start-up called Vectara, founded by former Google employees, is trying to figure out how often chatbots veer from the truth. The company’s research estimates that even in situations designed to prevent it from happening, chatbots invent information at least 3% of the time — and as high as 27%.
Experts call this chatbot behavior “hallucination.” It may not be a problem for people tinkering with chatbots on their personal computers, but it is a serious issue for anyone using this technology with court documents, medical information or sensitive business data.
Because these chatbots can respond to almost any request in an unlimited number of ways, there is no way of definitively determining how often they hallucinate. “You would have to look at all of the world’s information,” said Simon Hughes, the Vectara researcher who led the project.
Dr. Hughes and his team asked these systems to perform a single, straightforward task that is readily verified: Summarize news articles. Even then, the chatbots persistently invented information.
“We gave the system 10 to 20 facts and asked for a summary of those facts,” said Amr Awadallah, the chief executive of Vectara and a former Google executive. “That the system can still introduce errors is a fundamental problem.”
The researchers argue that when these chatbots perform other tasks — beyond mere summarization — hallucination rates may be higher.
Their research also showed that hallucination rates vary widely among the leading AI companies. OpenAI’s technologies had the lowest rate, around 3%. Systems from Meta, which owns Facebook and Instagram, hovered around 5%. The Claude 2 system offered by Anthropic, an OpenAI rival also based in San Francisco, topped 8%. A Google system, Palm chat, had the highest rate at 27%.
«
The problem isn’t so much the fact of hallucination – we often like it when humans make stuff up, a phenomenon we call “stories” – but that we can’t predict or necessarily spot where it’s happening.
unique link to this extract
Samsung unveils ChatGPT alternative Samsung Gauss that can generate text, code and images • TechCrunch
Kate Park:
»
Just a few days after OpenAI’s developer event, Samsung unveiled its own generative AI model, Samsung Gauss, at the Samsung AI Forum 2023.
Samsung Gauss, developed by the tech giant’s research unit Samsung Research, consists of three tools: Samsung Gauss Language, Samsung Gauss Code and Samsung Gauss Image.
Samsung Gauss Language is a large language model that can understand human language and answer questions like ChatGPT. It can be used to increase productivity in several ways. For instance, it can help you write and edit emails, summarize documents and translate languages. Samsung plans to incorporate the large language model into its devices like phones, laptops and tablets to make the company’s smart devices a bit smarter. When asked if it supports both English and Korean as interaction languages, a spokesperson of Samsung declined to comment on it.
Samsung Gauss Code, which works with its code assistant called code.i, focuses more specifically on development code. The idea is that Samsung Gauss Code could help developers write code quickly. Samsung said the AI model for code will support “code description and test case generation through an interactive interface.”
As for Samsung Gauss Image, as the name suggests, it will be an image generation and editing feature. For instance, it could be used to convert a low-resolution image into a high-resolution one.
«
AI envy is absolutely a thing now, and given that Samsung is the company most given to technology envy, this was inevitable, as is its gradual sunsetting and/or supplanting by users over the next few years in favour of something from Google.
unique link to this extract
August 2021: A $1.5m ‘women-led’ NFT project was actually run by dudes • Inverse
Chris Stokel-Walker, in August 20212:
»
The NFT market, like many tech-centric areas, has traditionally been dominated by men, and a women-led project was a much-welcomed change.
The three women behind it — Cindy and Andrea, the U.S.-based marketer and developer, respectively, and Kelda, the Norwegian artist and “ideologist” — were supposedly striking a note for female empowerment. Their head-and-shoulders illustrations of slender women in different guises — which users could pay Ethereum to mint and own — even merited a passing mention in The New Yorker.
But the story behind the project was a lie. The three women purportedly running Fame Lady Squad weren’t women at all. They were Russian men, according to research by NFT enthusiast and fellow Russian Fedor Linnik. And they are allegedly behind other NFT collectible series that claim to be one thing, but are in actuality something else entirely.
The story began with whispered rumors last month, and came to a conclusion, of sorts, this week. After an uprising within the community of investors that bought into the project to the tune of nearly $1.5 million, the Russian men behind Fame Lady Squad have ceded control of the project to actual women, including a self-employed realtor in Canada, Ashley Smith.
…“These guys are just cynically exploiting the Western, left-liberal agenda of protecting female rights and stuff like that,” says Linnik. He points to the fact that at least two of the original team alleged to be behind Fame Lady Squad have previously lived or studied in Canada as an indication that the decision to misrepresent their gender when launching the project was a cynical one. “I believe these guys understand Western society pretty well, and that’s why they can manipulate us easily.”
On Monday, Linnik posted a Twitter thread laying out what he knew. The men who had pretended to be women moved quickly to try and limit the reputational damage. On Tuesday, in a lengthy Twitter thread of their own, the originators apologized for misleading the world. “But it doesn’t mean it’s a scam or a fraud,” they wrote.
«
Nooooo, not a scam at all. NFT stats says average price was $123.20. “Current floor price $0.05”. A couple of years old, but Stokel-Walker pointed to it in the Guardian Technology mailout on Tuesday, and I couldn’t resist.
unique link to this extract
China’s spending on green energy is causing a global glut • WSJ
Sha Hua and Phred Dvorak:
»
China’s newest solar-energy manufacturers include a dairy farmer and a toy maker.
The new entrants are examples of a green energy spending binge in China that is fueling the country’s rapid build-out of renewable energy while also creating a glut of solar components that is rippling through the industry and stymying attempts to build such manufacturing elsewhere, particularly in Europe.
Since the start of the year, prices for Chinese polysilicon, the building block of solar panels, are down 50% and panels down 40%, according to data tracker OPIS, which is owned by Dow Jones.
Inside China, some companies fear a green bubble is about to pop. China’s state-guided economy spent nearly $80bn on clean-energy manufacturing last year, around 90% of all such investment worldwide, BloombergNEF estimates. The country’s annual spending on green energy overall has increased by more than $180bn a year since 2019, the International Energy Agency says.
The rush of funding has attracted an unusual array of companies to the bustling business. Last summer, Chinese dairy giant Royal Group unveiled plans for three new projects. There was a farm with 10,000 milk cows, a dairy processing plant and a $1.5bn factory to make solar cells and panels.
“The solar industry is improving over the long term, and the market potential is huge,” Royal Group wrote in a document outlining the project last year. More recently, Royal Group said it wants to create synergies between its core agricultural business and photovoltaics, “and promote solar technology to empower dairy owners to reduce costs and increase efficiency,” the company said in a response to The Wall Street Journal.
The milk manufacturer wasn’t alone in jumping on China’s solar bandwagon in the past two years. Other newbies include a jewelry chain, a producer of pollution-control equipment and a pharmaceutical company.
The newcomers are helping an ambitious wind and solar push in China—this year alone the country is set to install roughly as much solar as the U.S. has in total, Rystad Energy estimates.
«
One feels a disapproving tone in this story: how dare China fund a product that’s in huge demand (and, it’s careful to point out, which the US invented in the 1950s) and make its price crater so more people can benefit from it? Why can’t dairy farmers and toymakers just stick to their knitting?
unique link to this extract
The Humane AI Pin is a bizarre cross between Google Glass and a pager • Ars Technica
Ron Amadeo:
»
Not since Magic Leap has a “next-generation” hardware company been so hyped while showing so little. Everyone in the tech world has been freaking out about this new pocket protector thing that wants to “replace your smartphone.” It’s called the “Humane AI Pin.” As far as we can tell, it’s a $700 screenless voice assistant box and, like all smartphone-ish devices released in the last 10 years, it has some AI in it. It’s as if Google Glass had a baby with a pager from the 1990s.
«
Amadeo writes absolutely brutal reviews of hardware. I heartily approve. Savour this one particularly. (He’s also scathingly sceptical about the Nothing/iMessage/Sunbird promise.)
At the end, he asks:
»
Why wasn’t this just a smartwatch? Some of the OpenAI-powered responses are pretty neat, but there’s no reason not to have that just show up on a screen or be read aloud by a smartwatch.
«
That really is the question. This thing more and more looks like an adjunct to a smartphone, not a replacement. And an excellent question in the comments about this suit-lapel-worn (ladies, how do you feel about that?) device:
»
There are so many places to start, but I think my very very favorite is: does Silicon Valley know that, come November, when we go outdoors most of us are wearing coats?
«
| • 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 
The lesson I draw from that _Wired_ article is: AI art is real! People are using it in their work (of political propaganda), it’s not just a fancy demo. This is like living through the era when motion pictures were invented (think “Triumph Of The Will”). I’m sure there were articles back then roughly saying: “Oh my god, can you imagine this – that dictator *MADE A MOVIE* showing him as heroic, and it was all *disinformation*. That’s soooo scary, be very afraid, what new mechanistic plague have those amoral eggheads wrought upon us? Tyrannical governments are going be using this newfangled “cinema” stuff to show deep-fabrications of events which never happened, woe is us …”
I’m by no means a fan of Andreessen. However, with him now apparently getting the same sort of Great Satan media hate-on as Elon Musk, combined with the Biden Executive Order on AI, I can’t help but think that under his wretched plutocrat ideology, he does have a point.