Want to book your flight on United Airlines using your zippy new agent? Sorry, you can’t – they’re blocked. CC-licensed photo by Margalit Francus 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. Come book with me. I’m @charlesarthur on Twitter. On Threads: charles_arthur. On Mastodon: https://newsie.social/@charlesarthur. On Bluesky: @charlesarthur.bsky.social. Observations and links welcome.
AI has made it ‘close to impossible’ to assess university coursework, regulator says • The Guardian
Richard Adams:
»
Universities should go back to traditional exams and in-person presentations to grade students after lecturers complained that AI has made it “close to impossible” to assess coursework, according to an investigation by England’s higher education regulator.
The Office for Students (OfS) said the credibility of degrees from England’s universities was under threat from the prolific use of AI in essays, dissertations and “open book” exams, with students reporting that their peers boasted of using AI undetected by staff.
The regulator warned universities and other higher education institutions that failing to design “valid and reliable measures of students’ knowledge and skills” to combat AI could breach their conditions of registration and lead to sanctions being imposed.
Josh Fleming, the director of strategy and delivery at the OfS, said AI use in higher education “presents a generational risk to the system as it stands, threatening to undermine students’ learning and institutions’ ability to assess them”.
Fleming added: “We need to see far greater efforts to tackle these risks, including far greater use of AI-free assessment.
“The scale and pace of both the challenge and opportunity presents a novel challenge for the sector and us as the regulator. We will continue to work with universities and colleges to help them reap the benefits of AI while avoiding the risks.”
The OfS said many institutions had already redesigned how they assessed students to reduce potential AI misuse, including greater use of in-person written exams or oral examinations known as vivas.
«
This is arguably the quickest and best response to the spread of AI compared to many other sectors where it’s had a deleterious effect. As long, that is, as universities implement this.
unique link to this extract
Jensen Huang said to have had relationship with Nvidia executive who reports to him • The New York Times
Ryan Mac, Tripp Mickle and Georgia Gee:
»
Jensen Huang, the leader of Nvidia who is under investigation for an affair with a woman who US intelligence determined is a Chinese spy, also had an extramarital relationship with an executive who reports to him, according to five people who worked at the company.
The relationship, which has not been previously reported, was with Kimberly Powell, Nvidia’s vice president of health care, said the people, who spoke under the condition of anonymity to describe sensitive information. Under Nvidia’s code of conduct, employees are told to “avoid situations where our actions create conflict” including “personal relationships,” and the company states that it “promptly investigates reports of misconduct.” Many corporations have policies against managers having relationships with subordinates.
Mr. Huang, 63, and Ms. Powell, 48, were captured on video being affectionate at a San Jose, Calif., hotel during a Nvidia conference in March, according to two of the people who worked at the company. The New York Times has not seen the video.
Both Mr. Huang and Ms. Powell are married to other people.
«
The NY Times is going to pick away at this story endlessly, because it knows there’s a goldmine. Meanwhile, Nvidia starts to look less like a well-run company and more like one guy’s personal fiefdom. (The WSJ suggests that the board will try to force him out.)
unique link to this extract
US Treasury Secretary Bessent confronted Commerce Secretary Lutnick over Nvidia customer Megaspeed, sources say • Axios
Maria Curi and Marc Caputo:
»
Treasury Secretary Scott Bessent confronted Commerce Secretary Howard Lutnick over a failure to place Nvidia customer Megaspeed on a list of companies that pose national security concerns, sources familiar told Axios.
The Commerce Department oversees the export of sensitive US technologies and Lutnick’s actions — or lack thereof — are increasingly under scrutiny following revelations that Nvidia CEO Jensen Huang had an affair with Megaspeed employee and suspected Chinese spy Alice Huang.
At a White House Situation Room meeting on Thursday after the midterm elections, Bessent pressed Lutnick on why Megaspeed had not been added to the Commerce Department’s entity list, to which Lutnick responded he would “look into it.”
Commerce Undersecretary Jeff Kessler also was repeatedly warned about Megaspeed in deputy meetings.
But Kessler deferred to Lutnick, saying “That’s a decision for the secretary.”Sources told Axios that Lutnick has spent the last year “undoing” efforts by other administration officials to hold Megaspeed — a company accused of illegally diverting US chips to China — accountable because of concerns it would escalate the trade war with China.
«
The rats continue fighting in the sack over the fallout from this. “I’ll look into it” is code for “I realise I should have done this but I don’t want to admit I was wrong”. The fun is all flowing uphill to gather at Lutnick’s feet.
unique link to this extract
Tennis • Six Finger Punch
This was the result when someone asked Claude to code a fun game for them, and to really go wild. The opening screen looks just like Pong, doesn’t it? Try playing it.
unique link to this extract
Google DeepMind just mathematically verified key parts of Signal • WIRED via Ars Technica
Lily Hay Newman:
»
Google DeepMind researchers used agentic AI to mathematically prove, or “formally verify,” a number of the Signal protocol’s key cryptographic guarantees—a landmark development for the widely used cryptography and AI itself.
Conducted as a sort of proof-of-concept project in just four weeks, the work is significant in two ways. It serves as further independent validation of the cryptography that protects massive quantities of the world’s communications on platforms like the Signal app, Google Messages, and WhatsApp. And it demonstrates Google DeepMind’s formal verification capabilities using a new AI framework called AlphaProof Code.
More broadly, AI-aided formal verification is having a moment, but it is unclear how practical it will be for widespread use and how effective it can be at securing large amounts of code. Proponents of formal verification have long referred to it as a sort of holy grail in software security, because it uses math to check and validate that software is functioning as intended—in other words, that it doesn’t have critical vulnerabilities.
Before the AI era, it was incredibly time-consuming for people to describe the characteristics of software they wanted to prove, turn these summaries into formulas, and then develop the proof. Agentic AI can significantly cut down the time of the latter project, enabling analyses like DeepMind’s Signal evaluation. But even this analysis was limited in scope—it did not, for example, formally verify the entire Signal mobile app—and DeepMind researchers emphasize that their findings are just one milestone in a larger process.
“We are very excited by this result, but there is a long distance to go before we are able to scale this to all the software of the world, and we want to do that,” Swarat Chaudhuri, a DeepMind research scientist, tells WIRED.
«
And you know Google really does mean all the software in the world.
unique link to this extract
Anthropic targets valuation of up to $1.4tn in stock market flotation • The Guardian
Dan Milmo:
»
Anthropic is seeking a valuation of up to $1.4tn (£1.1tn) in its imminent stock market flotation, short of the $2tn that had been expected before investors started to question the durability of the AI boom.
The developer of the Claude chatbot was widely reported to be targeting a $2tn initial public offering on Wall Street but its sights have been set lower, according to a US regulatory filing on Tuesday.
Anthropic is aiming to raise up to $64.4bn by selling 825m shares in the IPO priced up to $78 a share. The California-based company will also raise a further $26bn via a separate private placement deal with cornerstone investors including the Google parent, Alphabet, and the chipmaker Nvidia.
An extra $9.7bn could be raised from a so-called greenshoe offering, which kicks in if investor demand is high and would take the total amount raised by Anthropic to approximately $100bn if the shares sold at the top end of the price range.
At the upper bracket of the $70-$78 per share price range, Anthropic would be worth $1.4tn on a fully diluted basis – which means taking into account all the common shares in the company in existence, including employee stock options.
Trading in Anthropic shares is expected to begin next week.
«
OpenAI nowhere in sight.
unique link to this extract
United won’t let your AI agent book your flight • The Verge
Lauren Feiner:
»
United says it won’t let personal AI assistants book tickets on the airline, citing both privacy and regulatory concerns.
The company says it hasn’t yet seen many people try to use AI agents to book their flights, The Wall Street Journal reports, in line with observations from other businesses like Doordash. But United is concerned that agents could expose customers’ information or make purchases they didn’t intend, and potentially skirt federal rules about the kind of information airlines need to share with customers — like having them attest that their bags don’t have explosives.
While agents can still show users United flights, they won’t be able to actually complete the transaction on their behalf. The decision, and others like it, could limit the potential usefulness of newer personal AI assistants like Meta’s Muse and OpenAI’s Dots. “To be honest, I don’t want to create another middleman,” United CEO Scott Kirby said at an event on Tuesday, according to the Journal.
Across sectors, businesses are grappling with whether to let AI agents transact on their sites, and United is not alone in blocking access for unauthorized agents. eBay appeared to block agents from browsing the site, and several major banks have also placed restrictions on them. United suggests it could reconsider its decision based on future advancements.
«
Second-class citizens. Well, not even citizens. How soon before there’s a court case in the US arguing that agents should be given “human” status?
unique link to this extract
For Silicon Valley’s leaders, getting too close to Trump was a mistake • The Atlantic
Nick Clegg:
»
First, tech leaders said that uniting the power of Washington and Silicon Valley would help America in its fight with China for AI supremacy. Yet over the past 18 months, China has been closing the gap with the United States by most objective measures, including model performance and chip development. In some important respects, China has outpaced America altogether, deploying open-weight models faster, developing a more sustainable energy strategy, and training world-class researchers and engineers at a larger scale.
Second, Silicon Valley leaders said that partnering with the government would help assuage public concerns about AI and overcome the “woke” opposition to Big Tech. Yet societal attitudes toward AI in the US are much more negative now than they ever were under President Biden, and worse than almost anywhere else on the planet. Fostering the very backlash they claimed they were trying to confront is a peculiar triumph of hubris and incompetence.
Third, tech executives hoped that earning Trump’s good graces would allow them to avoid unwelcome regulatory incursions outside the US, especially in Europe. The Trump administration has undoubtedly helped with this in some ways. For example, Commerce Secretary Howard Lutnick has become an important ally for tech leaders trying to fight a number of European Union enforcement decisions. But by any other meaningful measure the approach has been a disaster: Anti–Big Tech, and (alas) anti-American, sentiment has never been higher in large parts of the world, and most major economies—including Canada’s, India’s, Germany’s, and Brazil’s—are now scrambling to have less, not more, of what Silicon Valley produces.
Since Silicon Valley pivoted to Trump, then, its scorecard has looked something like this: chief rival (China) strengthened; domestic support weakened; global markets infuriated.
«
Clegg, who was of course head of global affairs at Meta until shortly before Trump’s second inauguration, offers a number of reasons why. “Misunderstanding politics” is a particularly brutal one. Isn’t that what they paid him for?
unique link to this extract
Amazon saves millions by replacing Sonnet 4.6 with Strands Decider • Strands Agents
Carr Onstott and Mehul Jaiswal:
»
For a while, Amazon and AWS teams relied on Claude Sonnet 4.6 to label tickets in an internal system without breaking the bank.
Sonnet 4.6 fit the goldilocks choice of balancing intelligence with cost. Considering it dealt with one of the world’s largest internal ticketing systems, the nine month old model did well. However, two Amazon engineers recently swapped out Sonnet 4.6 for Strands Decider, a decision model we launched a month ago, that performed just as well at 98% less cost. This change will save millions of dollars of token costs in a year.
Carr and Mehul, who work on the internal ticketing system, walk through how this happened below.
«
This is the sort of thing that 20 years ago would have been about a switch from MySQL to some giant unstructured database where it worked even better. Now we’re on to agents doing things and saving money, but the explanation is just as complex if you’re not deep into it.
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
