Experts find flaws in hundreds of tests that check AI safety and effectiveness
Experts have found weaknesses, some serious, in hundreds of tests used to check the safety and effectiveness of new artificial intelligence models being released into the world,Computer scientists from the British government’s AI Security Institute, and experts at universities including Stanford, Berkeley and Oxford, examined more than 440 benchmarks that provide an important safety net,They found flaws that “undermine the validity of the resulting claims”, that “almost all … have weaknesses in at least one area”, and resulting scores might be “irrelevant or even misleading”,Many of the benchmarks are used to evaluate the latest AI models released by the big technology companies, said the study’s lead author, Andrew Bean, a researcher at the Oxford Internet Institute,In the absence of nationwide AI regulation in the UK and US, benchmarks are used to check if new AIs are safe, align to human interests and achieve their claimed capabilities in reasoning, maths and coding.
The investigation into the tests comes amid rising concern over the safety and effectiveness of AIs, which are being released at a high pace by competing technology companies,Some have recently been forced to withdraw or tighten restrictions on AIs after they contributed to harms ranging from character defamation to suicide,“Benchmarks underpin nearly all claims about advances in AI,” Bean said,“But without shared definitions and sound measurement, it becomes hard to know whether models are genuinely improving or just appearing to,”Google this weekend withdrew one of its latest AIs, Gemma, after it made up unfounded allegations about a US senator having a non-consensual sexual relationship with a state trooper including fake links to news stories.
“There has never been such an accusation, there is no such individual, and there are no such new stories,” Marsha Blackburn, a Republican senator from Tennessee, told Sundar Pichai, Google’s chief executive, in a letter,“This is not a harmless hallucination,It is an act of defamation produced and distributed by a Google-owned AI model,A publicly accessible tool that invents false criminal allegations about a sitting US senator represents a catastrophic failure of oversight and ethical responsibility,”Google said its Gemma models were built for AI developers and researchers, not for factual assistance or for consumers.
It withdrew them from its AI Studio platform after what it described as “reports of non-developers trying to use them”.“Hallucinations – where models simply make things up about all types of things – and sycophancy – where models tell users what they want to hear – are challenges across the AI industry, particularly smaller open models like Gemma,” it said.“We remain committed to minimising hallucinations and continually improving all our models.”Sign up to TechScapeA weekly dive in to how technology is shaping our livesafter newsletter promotionLast week, Character.ai, the popular chatbot startup, banned teenagers from engaging in open-ended conversations with its AI chatbots.
It followed a series of controversies, including a 14-year-old killing himself in Florida after becoming obsessed with an AI-powered chatbot that his mother claimed had manipulated him into taking his own life, and a US lawsuit from the family of a teenager who claimed a chatbot manipulated him to self-harm and encouraged him to murder his parents.The research examined widely available benchmarks but leading AI companies also have their own internal benchmarks that were not examined.It concluded there was a “pressing need for shared standards and best practices”.Bean said a “shocking” finding was that only a small minority (16%) of the benchmarks used uncertainty estimates or statistical tests to show how likely a benchmark was to be accurate.In other cases where benchmarks set out to evaluate an AI’s characteristics – for example its “harmlessness” – the definition of the concept being examined was contested or ill-defined, rendering the benchmark less useful.
The best public interest journalism relies on first-hand accounts from people in the know,If you have something to share on this subject, you can contact us confidentially using the following methods,Secure Messaging in the Guardian appThe Guardian app has a tool to send tips about stories,Messages are end to end encrypted and concealed within the routine activity that every Guardian mobile app performs,This prevents an observer from knowing that you are communicating with us at all, let alone what is being said.
If you don't already have the Guardian app, download it (iOS/Android) and go to the menu.Select ‘Secure Messaging’.SecureDrop, instant messengers, email, telephone and postIf you can safely use the Tor network without being observed or monitored, you can send messages and documents to the Guardian via our SecureDrop platform.Finally, our guide at theguardian.com/tips lists several ways to contact us securely, and discusses the pros and cons of each.
Reeves refuses to say she will stick to manifesto pledge on tax rises and insists she must face world ‘as it is’ – UK politics live
Reeves is now taking questions.Beth Rigby from Sky News goes first.Q: Will you stick to your manifesto promise not to raise the taxes that working people pay? And, if you won’t, doesn’t that make a mockery of the trust people put in you at the election?Reeves replied:I will set out the individual policies of the budget until the 26th of November. That’s not what today is about. Today is about setting the context up for that budget
Divine dining: Australian church restaurants claim their own devout followings
At these places of worship, secular and churchgoing diners place their orders for coffee, curry puffs and za’atar pastries, served with kindnessGet our weekend culture and lifestyle emailOn Sunday mornings, thousands stream through Our Lady of Lebanon Co-Cathedral, a Lebanese Maronite Catholic church in Sydney’s western suburbs. In between back-to-back mass services, worshippers rush to its onsite cafe, Five Loaves.“Sunday is our busiest day,” says Yasmin Salim, who has fronted the counter for eight years. Lines are long and diners’ appetites are large: a single customer might ask for 10 pizzas and 10 pastries flavoured with za’atar, the Middle Eastern herb mix. “It’s like at Maccas, everyone wants their french fries,” says Salim
BP signals more cost cuts on way after fall in profits
BP has said it will ramp up efforts to hive off parts of the business, as the energy company reported a drop in profits in its latest quarter.The company reported an underlying profit of $2.2bn (£1.7bn) in the three months ended in September. It marked a slowdown against its previous quarter, when it made a profit of $2
Elon Musk’s $1tn Tesla pay deal to be rejected by huge Norway wealth fund
Norway’s sovereign wealth fund has said it will vote against a $1tn (£765bn) pay package for the Tesla chief executive, Elon Musk.The fund, which is the biggest national wealth fund in the world, said that while it appreciated the “the significant value created under Mr Musk’s visionary role” it would vote against his performance award.“We are concerned about the total size of the award, dilution and lack of mitigation of key person risk – consistent with our views on executive compensation,” it said. “We will continue to seek constructive dialogue with Tesla on this and other topics.”The warning from Norges Bank, which is the seventh biggest single shareholder in Tesla with a stake worth $17bn, comes two days before the carmaker hosts its annual shareholder meeting
French taxi driver cleared of stealing from David Lammy after fare dispute
A French taxi driver accused of stealing money and luggage from David Lammy has been acquitted due to lack of evidence, a prosecutor said.Nassim Mimun, 40, drove the deputy prime minister and his wife, Nicola Green, more than 600km (370 miles) from Forli, near Bologna in northern Italy, to the ski resort of Flaine in the French Alps on 11 April.But at the end of the journey the “tone escalated” over the cost of the fare, the Bonneville prosecutor Boris Duffau said in May.The driver, from the south-eastern city of Avignon, then left with his passengers’ bags in the boot of his car. “He dropped them off the next day at a municipal police station” but that was considered theft due to the length of time he had them in his possession, Duffau said
Josh O’Connor: the shape-shifting star who became cinema’s most wanted
He came to prominence with his portrayal of Prince Charles in The Crown, and now it seems that Josh O’Connor might be primed for his own coronation.The British actor is in three major films between now and January – better known to film-lovers as awards season.He leads Kelly Reichardt’s art heist drama Mastermind, which opened in UK cinemas last week; stars opposite Paul Mescal in the period romance drama The History of Sound; and takes the central role in Wake Up Dead Man, Rian Johnson’s third instalment in the Knives Out mystery franchise.There’s also the persistent industry chatter that he’s among those being considered for the next James Bond. “This Is The Autumn Of Josh O’Connor,” declared Vogue recently, while GQ wondered, “How Josh O’Connor Became the Thinking Man’s Leading Man”
We need clarity on big pharma’s tax breaks | Letters
UK factories return to growth after JLR restarts operations; US manufacturing exports hit by tariffs – as it happened
Australia’s richest person Gina Rinehart spotted at Donald Trump’s Halloween party
Victims robbed of £4bn in ‘insulting’ car loan redress scheme, say claims firms
Delivery firm DPD accused of ‘revenge’ sacking drivers who criticised pay cuts
Kimberly-Clark to buy Tylenol maker Kenvue in landmark $40bn merger