Hello ODI Supporter,
A researcher who’s worked at both OpenAI and Anthropic, Jacob Coxon, quit this week with a viral resignation post that racked up 70 million views, accusing both companies of “gambling with our lives”. The post got Anthropic’s own alignment lead, Evan Hubinger, to publicly agree and put his own estimate at “more than 10%” odds of AI killing everyone within a decade, while admitting that Anthropic doesn’t, as yet, have a plan to solve alignment for anything that powerful. One of the ‘godfathers of AI’, Geoffrey Hinton, more or less agreed on BBC Newsnight, calling Hubinger’s 10% guess “not an unreasonable estimate”, although he also said that “nobody really knows how to estimate it”. Coxon ended up rowing back a bit on CNN. But still, the comments landed days after OpenAI’s chief scientist warned that “no one is prepared for the consequences” of where this is heading, and Paul Christiano, a former alignment lead at OpenAI, said the company “isn’t on track to reduce risk to an acceptable level” as he joined the non-profit board of the firm, which is quite the entrance. Lawmakers on both sides of the Atlantic piled in. Ted Cruz trotted out Elon Musk’s “10-20%” estimate on daytime TV, Andy Burnham told parliament that AI “poses risks to our national security”, and even Sheryl Crow warned that AI has the capacity to “eliminate us in order to continue”. And she should know about existential threats, have you heard her cover of Sweet Child O’ Mine?! But not everybody is convinced. Sandra Wachter from the Oxford Internet Institute called the whole extinction framing thing “a big distraction from real issues”, commentator and academic Gary Marcus pinned the kerfuffle on “a superabundance of hype” from OpenAI itself, and, as Dame Wendy Hall put it, the whole thing could just be “PR and marketing” from two companies hurtling towards IPOs. All of which makes for interesting timing, given that on the very same day Geoffrey Hinton was warning the UK parliament that superintelligence “could even lead to human extinction”, Matt Clifford, architect of AI policy, resigned as chair of the UK’s Advanced Research and Invention Agency over a small conflict of interest - that is, being newly-employed by Anthropic. I wonder what he makes of it all.
A former forensic scientist at the Colorado Bureau of Investigation has been sentenced to 10 years after pleading guilty to perjury, forgery, attempting to influence a public servant and cybercrime. Yvonne “Missy” Woods manipulated DNA data across cases involving homicide, sexual assault and robbery, resigning in 2023 once an intern spotted missing information in a case she’d handled back in 2018. However, a subsequent internal affairs report found colleagues had raised concerns about Woods as far back as 2014, and she had already been temporarily pulled off DNA casework in 2018 over data manipulation allegations. Prosecutors have since had to review hundreds of cases. At least one murder conviction has been overturned, while two other defendants took plea deals rather than risk trials that prosecutors feared could be scuppered. The state puts the clean-up bill at more than $11 million. Woods told the court that she knew apologising “would not undo the harm I caused”, which is, at minimum, correct.
Health innovation minister James Frith warned this week that “mistrust” of Palantir is driving patients to opt out of sharing their NHS data, with 60,000 more people withdrawing consent between mid-May and mid-July alone. The government is currently weighing up whether to trigger the break clause of Palantir’s £330 million contract to run the NHS federated data platform, with doctors and patient groups citing the firm’s work with the Israeli military and ICE as cause for opposition. Palantir points to 110,000 more operations and a 15% cut in discharge delays as evidence that it’s working, although NHS England says it can’t “draw conclusions about cause and effect”. Frith called the rise in opt-outs “modest”, but campaign group Foxglove said people might reasonably be nervous, given Palantir founder Peter Thiel’s own 2023 view that the NHS “makes people sick”.
We were delighted to announce our membership of the Digital Public Goods Alliance (DPGA) this week. The DPGA is a United Nations-endorsed multi-stakeholder initiative that promotes the discovery, development and use of digital public goods, such as open software, open data, open AI models, open standards and open content, to advance the Sustainable Development Goals. Read our news story to find out what we’ve committed to do. We’ve also published an outline of our next research topic for the IDEA project, so take a look and find out more.
We’ve got webinars coming thick and fast this month. Solid World returns on Tuesday 15 September, 16:00-17:00, followed by Building municipal data capability: Lessons from the Urban Resilience Programme on Wednesday 23 September, 12:00-13:00 BST, looking at digital transformation in local government in South Africa. On Thursday 24 September, 14:00-15:00, we’ll explore the long‑standing question of how doctors and nurses approach decisions about sharing confidential patient information in the public interest, and on Friday 25 September, 12:00-13:00 BST, we’ll unveil a new partnership to re-establish the UK Knowledge Graphs Interest Group. Tickets are available for all of those.
And finally… diners are turning on AI menu photos. A Denver nurse’s viral post of a food pop-up’s barbecue menu, with meat like “leather belts crawling with tiny beetles”, drew more than 500 disgusted reshares, and a San Francisco cafe that tried the same thing got its shopfront graffitied. Restaurants call it cost-cutting. Oxford researchers say diners like the pics, right up until they’re told they are AI. I never trusted laminated menu photos anyway…
Until next time.
David and Alex
PS: Our friends at Applied Works, Arc and Snowflake, partners on TransitionArc and ResilienceArc, will come together for a live conversation exploring how climate and transition data can be transformed into useful, trusted and accessible intelligence that supports better decision-making on Thursday 17 September, 12noon – 1pm (BST). Check it out
PPS: The Smart Data Challenge Prize Significant Insights and Outcomes report has been published this week. We worked with Challenge Works (Nesta), bigspark and NayaOne to deliver the programme, so find out what we achieved.
