You can now get a chatbot that will generate novel proteins in response to a text prompt. It’s programming for amino acids. CC-licensed photo by Simon Cockell 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. Unfolded. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
OpenAI unites with Jony Ive in $6.5bn deal to create AI devices • The New York Times
Mike Isaac and Cade Metz:
»
On Wednesday, Sam Altman, OpenAI’s chief executive, said the company was paying $6.5bn to buy IO, a one-year-old start-up created by Jony Ive, a former top Apple executive who designed the iPhone. The all-stock deal, which effectively unites Silicon Valley royalty, is intended to usher in what the two men call “a new family of products” for the age of artificial general intelligence, or A.G.I., which is shorthand for a future technology that achieves human-level intelligence.
The deal, which is OpenAI’s biggest acquisition, will bring in Mr. Ive and his team of roughly 55 engineers and researchers. LoveFrom will assume creative and design responsibilities across OpenAI and build hardware that helps people better interact with the technology.
In a joint interview, Mr. Ive and Mr. Altman declined to say what such devices could look like and how they might work, but they said they hoped to share details next year. Mr. Ive, 58, framed the ambitions as galactic, with the aim of creating “amazing products that elevate humanity.”
“We’ve been waiting for the next big thing for 20 years,” Mr. Altman, 40, added. “We want to bring people something beyond the legacy products we’ve been using for so long.”
Mr. Altman and Mr. Ive are effectively looking beyond an era of smartphones, which have been people’s signature personal device since the iPhone debuted in 2007. If the two men succeed — and it is a very big if — they could spur what is known as “ambient computing.” Rather than typing and taking photographs on smartphones, future devices like pendants or glasses that use A.I. could process the world in real time, fielding questions and analyzing images and sounds in seamless ways.
Mr. Altman had invested in Humane, a company that pursued this kind of vision with the creation of an A.I. pin. But the start-up folded not long after its product flopped.
«
You can see that Humane’s product is a sort of platonic ideal for Ive. (By the way, NYT, do you not give people honorifics? He’s Sir Jony Ive.) But the next best would be smart glasses that would let you walk and talk to your device while absorbing the world that you see and acting on it. Things just got interesting.
unique link to this extract
Fortnite is finally back on US iPhones • The Verge
Jay Peters:
»
Fortnite is once again available on the iOS App Store in the US, according to Epic Games. Epic says it has returned to the Epic Games Store and AltStore as well.
Apple kicked Fortnite off the App Store nearly five years ago after Epic Games added its own in-app payment system to the game, which violated Apple’s rules. But after a major court ruling in Epic Games v. Apple that forced Apple to not take fees from purchases made outside of apps, the game is available to play on US iPhones once again.
[After Apple was evidently delaying its release following resubmission earlier this month] Epic asked the judge in the Epic v. Apple case to order Apple to review its Fortnite submission on May 16th. On Monday, the judge said in a filing that Apple is “fully capable of resolving this issue without further briefing or a hearing,” and that if a resolution wasn’t reached, the Apple official who “is personally responsible for ensuring compliance” would have to appear at a hearing next Tuesday.
However, shortly after Fortnite returned to the App Store on Tuesday, Epic and Apple filed a joint notice saying that they have “resolved all issues” from Epic’s May 16th filing. Apple didn’t immediately reply to a request for comment.
Epic also recently rolled out a new promotion to encourage players to use its payment systems: if you use Epic’s system in Fortnite, Rocket League, or Fall Guys on PC, iOS, Android, and the web, the company will give you 20% back in Epic Rewards that can be used for other purchases in its games or on the Epic Games Store.
«
Strange how the threat of having to turn up in court got Apple to hit the “OK” button.
unique link to this extract
‘Fortnite’ players are already making AI Darth Vader swear • WIRED
Megan Farokhmanesh:
»
On Friday, Epic Games announced Darth Vader would be returning to Fortnite as an in-game boss—but this time, players would be able to chat with him through conversational AI. “Ask him all your pressing questions about the Force, the Galactic Empire … or you know, a good strat for the last Storm circle,” Epic said in its announcement.
Unfortunately, players had other plans. Mere hours after Vader appeared in Fortnite, gamers began posting clips of AI Vader going rogue.
“What freaking fucking food is that Darth Vader? Tell me,” says streamer Loserfruit in one clip posted to X. “Freaking? Fucking? Such vulgarity does not become you,” Vader replies. (A spokesperson for Epic Games, Cat McCormack, told WIRED that it pushed a hotfix “within 30 minutes of this happening in-game, so this shouldn’t happen again.”)
Later, in a conversation about possible romantic partners, Loserfruit prompts Vader into replying “You speak of breasts, Loserfruit? I trust you are referring to the armored chestplates.”
In a clip from a different streamer, Vader can be heard talking about carcinogens before saying a slur typically used against queer men that can also be slang for cigarettes. The streamer can be heard screaming “HE SAID IT! HE SAID IT!” before running away in glee.
«
On the one hand, childish, but on the other, shows the impossibility of keeping these things inside what you think are its guardrails. Chatbots are like modern genies, and people really work at getting an extra spell out of them.
unique link to this extract
From chalkboards to chatbots: evaluating the impact of generative AI on learning outcomes in Nigeria • World Bank
Martín De Simone et al:
»
This paper examines whether generative artificial intelligence, specifically large language models (LLMs), can help solve that problem. We evaluate a six-week after-school tutoring program in Nigeria that used a publicly available LLM (ChatGPT-4) to support students in learning English. First-year secondary students from nine public schools in Benin City were invited to participate; from this pool, 52% of eligible students expressed interest, and participants were randomly selected from among them. Those assigned to the intervention attended twelve 90-minute sessions in computer labs, engaging in curriculum-aligned activities guided by teachers. We use a randomized controlled trial (RCT) design to estimate the causal impact of the program on learning outcomes.
We present three main sets of results. First, we show that students selected to participate in the program score 0.31 standard deviation higher in the final assessment that was delivered at the end of the intervention. We find strong statistically-significant intent-to-treat (ITT) effects on all sections of that assessment: English skills (which included the majority of questions, 0.24 σ), digital skills (0.14 σ), AI skills (0.31 σ) and an Item Response Theory (IRT) composite score of each student’s exam (0.26 σ). We also show that the intervention yielded strong positive results on the regular English curricular exam of the third term.
This result is important because the content evaluated in that exam was broader than the one covered during the six weeks of the intervention and included the content of the entire year.
«
This seems like a very significant finding: it’s relatively cheap and it works. In its way it reminds me of the famous paper which found that Tanzanian fishermen who had mobile phones could significant improve their takings because they knew which port would have the largest demand for their catch.
unique link to this extract
I told AI to make me a protein. Here’s what it came up with • Nature
Ewen Callaway:
»
Last month, a team led by Fajie Yuan, a machine-learning scientist at Westlake University in Hangzhou, China, showed that a text-to-protein model his team developed can design functional proteins, including lab-tested enzymes and fluorescent proteins, that are original in their designs and not similar to existing molecules. “We are the first to design a functional enzyme using only text,” Yuan says. “It’s just like science fiction.”
The model, called Pinal, is one of several protein-design AIs that can be directed with ordinary language — as opposed to a protein sequence or the structure-guided specifications typical of most such AIs.
But it’s early days for these bio-AI models, says Anthony Gitter, a computational biologist at the University of Wisconsin–Madison. “I see it as a high-risk, high-reward area,” he says.
Teaching biological AI models to communicate in English (or any language) typically involves exposing them to text descriptions of biological data. Yuan’s team trained Pinal using 1.7 billion descriptions of the structures, functions and other characteristics of different proteins. After some extra training, the model could take a prompt and churn out hundreds of sequence designs. The model has a web interface, and its code and parameters needed to run the model are freely available.
One prompt that the researchers used was ‘Please design a protein that is an alcohol dehydrogenase’, referring to an alcohol-metabolizing enzyme. Yuan and his colleagues then used other computational tools to identify the most promising designs and, working with a biologist collaborator, tested their enzymatic activity.
Two of the eight alcohol dehydrogenase designs successfully catalysed the breakdown of alcohol, albeit much less efficiently than natural enzymes.
«
Like programming, proteins are a bounded problem space, so they’re a good topic for LLMs (or maybe LPMs – large protein models).
unique link to this extract
South Korea’s robot chefs worry human workers – and disappoint customers • Rest of World
Michelle Kim:
»
On sweltering summer days, chef Park Jeong-eun would cook makguksu, an earthy Korean dish made with buckwheat noodles steeped in a tangy, ice-cold broth, topped with spicy gochujang paste. Truck drivers would come from faraway places to the Munmak rest stop, on the highway in the mountainous Gangwon-do province in South Korea, to eat her food.
That was until February 2024, when three robot chefs took over the kitchen at Munmak. The restaurant’s menu has since changed, away from local delicacies like makguksu and slow-cooked beef stews to easily automatable dishes such as ramen, udon, and varieties of Korean stews. The robots speed through 150 meals every hour, nearly double what Park can make by hand.
When longtime patrons learn their beloved menu items are no more, they gasp and walk out the door, she recalled to Rest of World.
“Our customers say the dishes we used to cook tasted much better than what the robots serve now,” Park, 58, said. “Even though the robots have lightened my workload, I’ve lost my sense of pride in our food.”
Park now finds refuge scrubbing dishes in the back of the kitchen, away from the counter, where customers barrage her with harsh complaints about the food. Sometimes, they return their ramen bowls untouched in protest, she said.
«
But! South Korea needs its robots because the population is ageing: those over 60 (as Park will soon be) make up a quarter of the workforce. Maybe get used to the new taste of the ramen?
unique link to this extract
The spy factory: how Russia used Brazil to create deep-cover spies • The New York Times
Michael Schwirtz and Jane Bradley:
»
For years, a New York Times investigation found, Russia used Brazil as a launchpad for its most elite intelligence officers, known as illegals. In an audacious and far-reaching operation, the spies shed their Russian pasts. They started businesses, made friends and had love affairs — events that, over many years, became the building blocks of entirely new identities.
Major Russian spy operations have been uncovered in the past, including in the United States in 2010. This was different. The goal was not to spy on Brazil, but to become Brazilian. Once cloaked in credible back stories, they would set off for the United States, Europe or the Middle East and begin working in earnest.
The Russians essentially turned Brazil into an assembly line for deep-cover operatives like Mr. Shmyrev.
One started a jewelry business. Another was a blond, blue-eyed model. A third was admitted into an American university. There was a Brazilian researcher who landed work in Norway, and a married couple who eventually went to Portugal.
Then it all came crashing down.
For the past three years, Brazilian counterintelligence agents have quietly and methodically hunted these spies. Through painstaking police work, these agents discovered a pattern that allowed them to identify the spies, one by one.
Agents have uncovered at least nine Russian officers operating under Brazilian cover identities, according to documents and interviews. Six have never been publicly identified until now. The investigation has already spanned at least eight countries, officials said, with intelligence coming from the United States, Israel, the Netherlands, Uruguay and other Western security services.«
Easy to forget that all this stuff goes on, and on, under the surface.
unique link to this extract
Judge slams lawyers for ‘bogus AI-generated research’ • The Verge
Emma Roth:
»
A California judge slammed a pair of law firms for the undisclosed use of AI after he received a supplemental brief with “numerous false, inaccurate, and misleading legal citations and quotations.” In a ruling submitted last week, Judge Michael Wilner imposed $31,000 in sanctions against the law firms involved, saying “no reasonably competent attorney should out-source research and writing” to AI, as pointed out by law professors Eric Goldman and Blake Reid on Bluesky.
“I read their brief, was persuaded (or at least intrigued) by the authorities that they cited, and looked up the decisions to learn more about them – only to find that they didn’t exist,” Judge Wilner writes. “That’s scary. It almost led to the scarier outcome (from my perspective) of including those bogus materials in a judicial order.”
As noted in the filing, a plaintiff’s legal representative for a civil lawsuit against State Farm used AI to generate an outline for a supplemental brief. However, this outline contained “bogus AI-generated research” when it was sent to a separate law firm, K&L Gates, which added the information to a brief. “No attorney or staff member at either firm apparently cite-checked or otherwise reviewed that research before filing the brief,” Judge Wilner writes.
When Judge Wilner reviewed the brief, he found that “at least two of the authorities cited do not exist at all.”
«
As I’ve said previously, one has to think that this might have slipped past other judges if the two sides aren’t careful enough.
unique link to this extract
From birth to gene-edited in six months: custom CRISPR therapy breaks speed limits • Ars Technica
Beth Mole:
»
researchers in Philadelphia appear to have successfully treated a six-month-old baby boy, called KJ, with a personalized CRISPR gene-editing therapy. The treatment corrects an ultra-rare mutation in KJ that breaks a liver enzyme. That enzyme is required to convert ammonia, a byproduct of metabolism, to urea, a waste product released in urine. Without treatment, ammonia would build up to dangerous levels in KJ—and he would have a 50% chance of dying in infancy.
While the gene-editing treatment isn’t a complete cure, and long-term success is still uncertain, KJ’s condition has improved and stabilized. And the treatment’s positive results appear to be a first for personalizing gene editing.
Now, who doesn’t love a good story about a seemingly miraculous medical treatment saving a cute, chubby-cheeked baby? But, this story delivers more than an adorable bundle of joy; the big triumph is the striking timeline of the treatment’s development—and the fact that it provides a template for how to treat other babies with ultra-rare mutations.
«
The mutation was recognised within days (on its own, a remarkable bit of medicine) but the process of taking cells, finding the targets in the infant’s genome, proving it, getting approval, doing animal tests, all happened at breakneck speed – and, apparently, successfully.
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 