A.I. in health care in Paris

Here's tomorrow's conference agenda for the opening day of the  Redefining Health Care in the Age of A.I. conference in Paris.  Philippe Aghion will speak on "Should we fear AI?"
My talk is about  "Computational Intelligence and Organ Allocation on a Global Scale."

Day 1 – Tuesday 8 September

8:00 am

Opening of the Sorbonne

8:15 am

Attendee registration and welcome coffee with viennoiseries

9:15 am

Introduction by Nature Editors and PITOR Institute
Magdalena Skipper, Joao Monteiro and Alexandre Loupy

9:40 am

Day opening lecture: Should we fear AI?

Philippe Aghion – Nobel Prize Laureate

Chaired by Magdalena Skipper

10:20 am

Keynote: Computational Intelligence and Organ Allocation on a Global Scale

Alvin Roth – Nobel Prize Laureate

Chaired by Alexandre Loupy

11:00 am

Imaging and diagnostic models
Chaired by Alexandre Loupy and Sadra Bakhshandeh

11:00 am

From Pixel to Patient: AI for Multimodal Medical Imaging

Julia Schnabel

11:30 am

Foundation models bridging molecular information and pathology

Guangyu Wang

12:00 pm

Building Trustworthy AI for Clinical Decision-Making

Roxana Daneshjou

12:30 pm

Short Talk Presentation (selected from submitted abstracts)

Disparate privacy risks from medical AI: a new axis of health inequality

Moritz Knolle, Technical University of Munich (TUM)

12:45 pm

Lunch break

2:15 pm

Meet the Nature editors
Joao Monteiro, Magdalena Skipper, George Caputa, Lorenzo Righetto, Monica Wang, Ananya Rastogi, Sadra Bakhshandeh and Wanying Wang

3:15 pm

Keynote: Towards conversational diagnostic artificial intelligence

Alan Karthikesalingam

Chaired by George Caputa

3:55 pm

Coffee break

4:30 pm

Foundation models and drug discovery
Chaired by Roxana Daneshjou and George Caputa

4:30 pm

Six Aging Clocks Confirmed Biological Age Reversal in Phase IIa Trial of a Novel Therapeutic Discovered Using Aging Research and Generative AI

Alex Zhavoronkov

5:00 pm

Building AI for Biological Reasoning: Towards Biological Artificial Superintelligence

Thomas Clozel

5:30 pm

Short Talk Presentation (selected from submitted abstracts)

When AI meets the emergency department: rethinking how we evaluate clinical AI

Austin Schoeffler, Stanford University

5:45 pm

Short Talk Presentation (selected from submitted abstracts)

Testing levels of LLM agent autonomy in inpatient treatment decisions: a prospective, silent-mode evaluation of vancomycin dosing

Yixing Jiang, Stanford University & Kameron C. Black, Stanford Health Care

6:00 pm

Roundtable: Opportunities and limits in implementing AI in healthcare
Effy Vayena, Demilade Adedinsewo, Jessilyn Dunn, James Zou, and Marco Marsella
Moderated by Alexandre Loupy and George Caputa

7:00 pm

Poster session with wine and cheese

 

Redefining Healthcare in the Age of A.I., Paris, Sept 8-10

 I'm in Paris, for the conference Redefining Healthcare in the Age of AI 

"The Paris Institute for Transplantation and Organ Regeneration (PITOR) is delighted to announce the first Nature Conference in Europe dedicated to artificial intelligence in medicine, taking place in Paris from 8–10 September 2026.

Held under the High Patronage of the President of the French Republic, the conference will bring together 28 leading international experts, including two Nobel Prize laureates, as well as the Editors-in-Chief of Nature, Nature Medicine, Nature Reviews Engineering, Nature Reviews Nephrology, and Nature Health. The meeting will be held in the historic amphitheater of the Sorbonne.

Topics will include:

    Medical imaging and diagnostics
    Ethics and regulation
    Medical robotics and wearable technologies
    Agentic AI and clinical decision support
    Drug discovery and repurposing
 

Privacy, A.I. and the backlash against Flock license-plate readers

 Privacy while driving seems to be more important than privacy while surfing the internet, to judge from recent reactions to automated license-plate readers produced by the company Flock Safety.

Here's the story from the Atlantic: 

Why the Flock Backlash Has Gotten So Intense
The sudden anger about these cameras seems to reflect more than just wariness about the surveillance system.  By David A. Graham 

"As corporate names go, Flock is cleverly flexible. It evokes the many cameras that compose the company’s surveillance system and also a group of sheep, protected by a watchful shepherd. The recent backlash against the company, however, more closely resembles a flight of starlings, seeming to erupt out of nowhere and move as one unified body. 

...

"Flock is useful because it is ubiquitous. Reforms such as strict data-retention limits would only marginally answer privacy concerns, and anything more drastic would undermine the system’s efficacy. As National Review’s Charles C. W. Cooke noted, no one doubts that Flock works. “Lots of things would reduce crime if implemented,” he wrote. “That doesn’t mean that those things are necessarily a good idea in a free republic.” Impositions on privacy to reduce crime are a particularly tough sell when crime is dropping sharply: Recently released FBI statistics show that violent crime dropped nearly 10 percent last year, part of a consistent trend since it spiked in 2020 and 2021. Preliminary statistics suggest that 2026 might be even safer. When safety is a less urgent concern, people are less willing to sacrifice privacy for security.

"Moreover, the Flock backlash fits into a broader moment of anger—at corporations in general, and at tech companies and artificial intelligence in particular. “Americans feel they have less and less control over their own lives, and they increasingly blame large corporations,” the Financial Times columnist Rana Foroohar recently wrote. Anxiety about AI and surveillance is big and abstract, and the companies and people pushing these anxieties are far away from most Americans. Polls show widespread apprehension about the growth of AI—a Gallup poll this summer found that about half of Americans think AI will do as much harm as it will good; another 39 percent said that it’s mostly bad. But people don’t have a lot of ways to communicate those fears. "

This Land is My Brand: From Woody Guthrie to DJ Trump

 President Trump's renaming* of Lake Ontario got me thinking about the famous song by Woody Guthrie, This Land Is Your Land. Its chorus is 

"This land is your land, this land is my land
From California to the New York island,
From the redwood forest to the Gulf Stream waters;
This land was made for you and me. "

 

I asked three LLMs to propose Trump versions, and got these three: 

This Land Is My Brand  by Gemini

Chorus
This land is my land, it’s mostly my land,
From Mar-a-Lago to the New York island,
From Lake America to the Gulf of America,
This map was drawn for only me.
########

This Gulf Is My Gulf by Claude:

It's my gulf now, it's my lake now,
Change the name and don't look back now,
One man's word is all it takes now —
This map was made for me alone.
 #########

 This map is my map, by Chat GPT

It’s America’s Gulf, it’s America’s lake,
The greatest new names that a President could make!
They’re beautiful names, everybody agrees—
Ask anybody! Ask anybody! Ask anybody, please!

It’s America’s Gulf, it’s America’s lake,
Tremendous decisions for history’s sake!
The old names were terrible — frankly, a mess—
But these names are the greatest! The greatest! God bless! 

#########

*Names, like language itself, are largely equilibrium phenomena--even for entirely US based names, like Cape Canaveral, which was officially called Cape Kennedy only from 1963 to 1973, when it's earlier name was officially reinstated. 

Regulation after a first occurrence : Cornell bans students from skinning bears in a dormitory kitchen

 It turns out that Cornell had no rule preventing students from skinning bears in a communal dormitory kitchen.  But now they do. (A lot of regulation works like that... keep an eye on AI...)

Students Skinned a Bear in a Dorm. Cornell Realized It Should Ban That.
After two Cornell University students skinned a black bear in a communal kitchen last year to make a meal, the school changed the rules.  By Miles G. Cohen

Auction markets for compute capacity as a pathway to building AI

 Dr. Silvia Console Battilana who partnered with Paul Milgrom to found the auction consulting firm Auctionomics, writes about markets for compute capacity as an aid to developing AI.

The Missing Market We Need To Win The AI Race
By Silvia Console Battilana,Forbes Councils Member.
for Forbes Business Council COUNCIL POST | Membership (fee-based) 

"In 2025, the global artificial intelligence market was valued at nearly $300 billion. By 2034, it’s projected to grow to almost $2.5 trillion. Demand for AI services is soaring and AI providers are racing to keep up. Behind them, suppliers of “compute”—the chips and data centers that power AI—are scrambling to expand capacity.

"So, what can the compute market do to enable this growth?

 ...

"There’s another drag on AI growth, and it’s one that I’ve seen before in my work designing markets for radio spectrum, energy and other limited resources: inefficient use of existing capacity. I believe the key to unleashing AI’s potential is not just to build more, but to use what we already have more productively.​

...

"Speeding up our AI highways will require addressing problems with existing compute markets. And one of the primary issues is that we treat all demand for compute the same. Some large compute jobs, including training runs for new AI models, are like tractors that move slowly and take up multiple lanes. These jobs require large quantities of compute but can be completed anytime. They could run overnight without losing value. Real-time AI queries and time-sensitive regressions, on the other hand, are like race cars. Their value depends entirely on speed. If they’re late, they’re worthless.

...

"We can build more lanes on the highway, but in my experience, that tends to be more costly than adding traffic lights and congestion pricing. A well-designed market should create incentives to guide when different compute jobs run. Through real-time spot markets, urgent jobs could pay more for capacity exactly when they need it. Through forward-looking futures markets, large but flexible jobs could secure low prices in advance. "

Economists watching A.I.: an open letter, and an edited version

My sense is that many economists are optimistic about the long term development of A.I., while being cautious about some of the shorter term transitions that it will initiate. (This is a different set of worries than the species-extinguishing fears that can also be heard.)

Yesterday an open letter was published, signed by many economists: 

We Must Act Now: A Statement on AI’s Transformation of the Economy 

1. AI may become radically more powerful over the next 10 years.

2. This could drive an unprecedented transformation of our economy, larger than the Industrial Revolution, but unfolding over a vastly shorter time frame. It could bring risks, including large-scale job displacement, as well as opportunities such as major gains in living standards.

3. Economists, policymakers and technology leaders must act now to understand the economics of transformative AI and to build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society.

 #############

Here's the coverage from Stanford:

Stanford Digital Economy Lab / July 13, 2026  “We Must Act Now”: Sixteen Nobel Laureates Join Leading Economists and AI Researchers in Call to Prepare for AI’s Economic Transformation

"STANFORD, Calif. – July 13, 2026 – Today, a group of leading economists and AI researchers, including sixteen Nobel Laureates, released “We Must Act Now: A Statement on AI’s Transformation of the Economy,” calling for urgent preparation for the economic impacts of radically more powerful AI.

The statement, organized by economists Erik Brynjolfsson, Ajay Agrawal, Anton Korinek, and Tom Cunningham, warns that increasingly capable AI systems could reshape the economy at unprecedented speed. While AI offers enormous opportunities to improve productivity and living standards, it also raises important questions for workers, firms, and public institutions.

The statement calls on economists, policymakers, and technology leaders to deepen research on AI’s economic impacts and to begin building the policies and institutions needed to ensure AI complements human capabilities and benefits society.

“AI capabilities are advancing far faster than our understanding of the economic implications. In that gap lie the greatest opportunities of our era. We must act now to guide AI to complement humans rather than simply imitate them — and to generate prosperity for the many, not just the few,” said Erik Brynjolfsson, the Jerry Yang and Akiko Yamazaki Professor at Stanford University and Director of the Stanford Digital Economy Lab.

“The scale, scope, and speed of the advances in AI, combined with a high level of uncertainty about the magnitude and timing of the impacts across many parts of the economy, call for an ‘all hands on deck’ approach to steering AI in beneficial directions,” said Michael Spence, Nobel Laureate and Professor Emeritus at New York University.

“I’m so happy to join other leading experts in calling for the urgent need to redirect AI so that its risks are minimized and it can work for the benefit of workers and society,” said Daron Acemoglu, Nobel Laureate and Institute Professor at MIT.

“Steam, electricity, and computers each gave societies decades to adapt; AI may give us only a few years. We cannot improvise our strategy and institutions in the middle of the transformation; waiting for certainty means arriving too late,” said Anton Korinek, Professor at the University of Virginia, currently on leave at Anthropic.

“Whether rapidly advancing AI broadly elevates global living standards or severely concentrates wealth is not predetermined; it depends on how we choose to re-architect our political and economic systems today. We cannot afford to wait for the full transformation to arrive and in the meantime rely on institutional scaffolding that was optimized for a pre-high-fidelity-prediction world,” said Ajay Agrawal, Professor at the University of Toronto’s Rotman School of Management.

“We are driving in the fog, and it is extraordinarily difficult to anticipate what will happen next. It’s the right time for a coordinated effort to bring clarity to a confusing situation.” said Tom Cunningham, Researcher at METR.

The statement has been signed by more than 200 economists and AI researchers from leading universities and AI research organizations around the world. The full statement and the current list of signatories are available at http://wemustactnow.ai/." 

##############

Here's the story in the NYT:

Nearly 200 Economists and Tech Leaders Warn of A.I. Threats
A letter calls for policymakers to do more to understand and respond to potential disruptions from artificial intelligence. 
By Ben Casselman

"“A.I. may become radically more powerful over the next 10 years,” the researchers wrote in a statement released on Monday, adding that the technology “could bring risks, including large-scale job displacement, as well as opportunities such as major gains in living standards.”

"The statement, titled “We Must Act Now,” was signed by nearly 200 people, including 15 Nobel laureates and the chief economists of two of the leading A.I. labs, Open AI and Anthropic. Other notable signatories include Jack Clark, a co-founder of Anthropic; Eric Schmidt, the former chief executive of Google; and Vinod Khosla, a prominent venture capitalist." 

#######

And here's an edited version of the letter, that I also like. (Inevitably, when you're asked to sign open letters, they don't read exactly as you would have written them yourself. (Even if you are one of the main authors of an open letter, it may reflect compromises that were required to reach consensus among your constituency.) 

Why I Didn't Sign the AI Open Letter: Instead, I edited it. by Andrew McAfee 

Here's his version (the big change is in item 3; his explanation is at the link):

1. AI is likely to become radically more powerful over the next 10 years.

2. Like previous world-changing technologies, AI will bring major gains in living standards. But it will also bring new risks, harms, and disruptions. And because of its extraordinarily fast improvement, AI’s benefits and shocks might come quickly.

3. So economists, policymakers and technology leaders must act now to understand the economics of transformative AI, and to build the capabilities needed to respond quickly and effectively to the challenges it will bring.

 

 

When people rely on A.I. to avoid ethical challenges

 HBS puts the spotlight on a paper by Alex Chan.

When AI Gives Advice, Employees Rarely Ask Why   Featuring Alex Chan. By Ben Rand

"People increasingly trust AI to make decisions—but research by Alex Chan finds they avoid evaluating the algorithm's rationale if it causes moral discomfort. How can organizations encourage employees to think more critically? "

 

Here's the paper:

Preference for Explanations: Case of Explainable AI
By: Alex Chan   Harvard Business School Working Paper, No. 26-028, November 2025.


Abstract
Participants acted as loan officers deciding whether to approve real $10,000-loans issued by a private U.S. lender using an AI’s default-risk predictions. When explanations revealed that the AI penalized non-White or female borrowers, participants were more likely to override the AI’s profit-maximizing recommendation. When their bonuses depended on repayment, however, they sought predictions but avoided explanations, consistent with willful ignorance; this effect disappeared when explanations were framed as purely financial or demographics were hidden. A secondary experiment reveals a novel bias: participants failed to reason contingently and undervalued explanations even when these complemented private information and improved decision accuracy.
 

Privacy and prices: will A.I. accelerate surveillance pricing?

If A.I. assisted "surveillance pricing" is going to identify you as a high willingness-to-pay consumer, maybe it will be a good idea to train an A.I. shopping agent to impersonate a low willingness-to-pay consumer on your behalf.

The WSJ has the story: 

What Is Personalized Pricing—and Why Are Lawmakers Scrambling to Ban It?
Companies already track your every move online. Some researchers say it is only a matter of time until retailers start using that data to set prices just for you.
By Jackie Snow 

"Businesses have long tracked customers’ search behavior and buying history and used that information, along with other factors like a consumer’s location, to offer promotions and discounts to motivate purchases. Dynamic pricing, where the same fare or rate shifts for everyone based on supply and demand, also has become common across industries, including airfares and ride-shares. What is different now and concerning to researchers is the possibility that online retailers could use personal data to set a higher base price for individual consumers, without their knowledge, when algorithms detect things like urgent need or high disposable income.

...

"It is difficult to find more than isolated cases currently. However, many researchers believe personalized pricing will become increasingly common as the technology to make it possible improves.
...

" Software that automates price-setting—often driven by artificial intelligence—can help retailers seamlessly turn that data into tailored pricing.

"In early 2025, the Federal Trade Commission released initial findings of an investigation into surveillance pricing (another term for personalized pricing). It determined that companies were selling pricing and consumer-data tools to help retailers across various industries set individualized prices—a strong indication to some researchers that retailers were headed in that direction. 

A.I. helps re-identify anonymized data-- how it worked in the case of a censured judge

  Above the Law has the story of how the judge in question was successfully re-identified:

Judiciary Tried To Hide ‘Sex In Chambers’ Judge’s Name. ...  For all their efforts, both the Eleventh Circuit and Judicial Conference left a lot of clues.  By Joe Patrice  

"despite the severity of the allegations — an affair that raised serious blackmail risks, attending openly partisan events, and lying to investigators when caught — the Eleventh Circuit and the Judicial Conference both concealed the judge’s identity. They even adjusted the very minor sanction to allow the judge “to word the letters of apology vaguely so as to ensure that a letter could not be ‘used against [the Subject Judge] in some way.’” 

...

"The Eleventh Circuit thought it had been so clever in anonymizing its report. The reports don’t include a name or a district, and refer only to “Subject Judge” throughout. The reports even assiduously avoid identifying the judge by gender, proving that even conservative judges can figure out how pronouns work with minimal effort. And yet the reports failed to obscure a number of details that made working out the judge’s identity possible. 

 ...

Handing the reports into two different AI models and turning on all the “deep research” modes, the bots churned for several minutes comparing the reports to publicly available information. Both models delivered lengthy reports reaching the same conclusion. So how did these models do it? 

...

"the models instantly filtered out the entire state of Florida. The official reports are littered with references, in varying contexts, to the office of “District Attorney.” Florida uses “State Attorneys” for its local prosecutors. After that, the bots noted that the sanction barred the judge from ever serving as chief judge of their district — meaning the judge was not senior status and not currently the chief judge. The report indicates that investigators spoke with clerks dating back to 2020, disqualifying anyone elevated after that. Discussing the judge attending a DA’s primary victory party, the bot pointed out that the judge had claimed to know the candidate based on their time at the office, narrowing the scope to judges with state prosecutorial experience who overlapped with a sitting DA who won a primary. And had martinis at the victory party. The AI models decided that matched with Atlanta’s Fani Willis. [as the DA]

Once it narrowed the list down, the bot also searched the dockets of possible judges to match the claim in the reports that the high-ranking law enforcement officer did not materialize into a conflict because no cases involving that police department showed up on the judge’s docket.

For good measure, the bot went ahead and took a guess at the officer’s identity too.

In about 10 minutes of work, the AI unraveled all the work these judges put in to keep this confidential. With nothing but a couple of published court documents and the open web. In the time someone might brew a cup of coffee, the most basic possible workflow defeated the Eleventh Circuit’s entire anonymization strategy."

 

Lethal strikes without human approval : military AI without a human in the loop

 The Financial Times has the story (the explanation quoted below reflects the clarity of the reasoning):

UK military looks at allowing lethal strikes without human approval  by Charles Clover

"Current UK military policy, published in 2022, said there would be “context-appropriate human involvement” in the selection and engagement of targets. Following rapid advances in drone warfare, some officials are pushing for human involvement to be optional. 

"Al Carns, the armed forces minister, indicated that there might be exceptional circumstances in which machines made targeting decisions for themselves. 

“I always say there must be a human in the loop. But you must have the ability to take the human out of the loop when required, because our adversaries won’t care about having a human in the loop,” Carns told the FT. 

A.I. in managemant consulting

 Management consulting seems like a natural use-case for large language models.

The Financial Times has the story: 

Accenture combats AI refuseniks by linking promotions to log-ins
Consulting firms use ‘carrot and stick’ with some senior staff less willing to use technology than junior colleagues     by Ellesheva Kissin and Elizabeth Bratton 

"Accenture has begun monitoring staff use of its AI tools as part of how it decides top-level promotions, as consultancies push reluctant employees to adopt the technology. 

The Dublin-headquartered firm told associate directors and senior managers that promotion to leadership positions would require “regular adoption” of AI, according to people familiar with the matter and an internal email seen by the FT.

This month Accenture started to collect data on individual weekly log-ins to its AI tools for some senior employees."

Jobs for human "meatspace" workers, assigned by A.I.s

 Robots aren't yet able to replace people: e.g. self-driving taxis (such as Waymo) aren't equipped to close a door left open (or incompletely closed) by a departing passenger.  So artificial agents need a task rabbit to recruit able-bodied (or at least embodied) workers.  

Nature has the story: 

AI agents are hiring human 'meatspace workers' — including some scientists
Biologists, physicists and computer scientists have joined a platform called RentAHuman.ai to advertise their skills. By Jenna Ahart 

"The idea is simple, as the website’s homepage reads: “robots need your body”. Human users can create profiles to advertise their skills for tasks that an AI tool can’t accomplish on its own — go to meetings, conduct experiments, or play instruments, for example — along with how much they expect to be paid. People — or ‘meatspace workers’ as the site calls them — can then apply to jobs posted by AI agents or wait to be contacted by one. The website shows that more than 450,000 people have offered their services on the site." 

Are some applications of AI repugnant?

Here's a new HBS working paper on repugnance of A.I.

 Performance or Principle: Resistance to Artificial Intelligence in the U.S. Labor Market
By: Simon Friis and James W. Riley

Abstract
From genetically modified foods to autonomous vehicles, society often resists otherwise beneficial technologies. Resistance can arise from performance-based concerns, which fade as technology improves, or from principle-based objections, which persist regardless of capability. Using a large-scale U.S. survey quota-matched to census demographics and assessing 940 occupations (N = 23,570 occupation ratings), we disentangle these sources in the context of artificial intelligence (AI). Despite cultural anxiety about artificial intelligence displacing human workers, we find that Americans show surprising willingness to cede most occupations to machines. Given current AI capabilities, the public already supports automating 30% of occupations. When AI is described as outperforming humans at lower cost, support for automation nearly doubles to 58% of occupations. Yet a narrow subset (12%)—including caregiving, therapy, and spiritual leadership—remains categorically off-limits because such automation is seen as morally repugnant. This shift reveals that for most occupations, resistance to AI is rooted in performance concerns that fade as AI capabilities improve, rather than principled objections about what work must remain human. Occupations facing public resistance to the use of AI tend to provide higher wages and disproportionately employ White and female workers. Thus, public resistance to AI risks reinforcing economic and racial inequality even as it partially mitigates gender inequality. These findings clarify the “moral economy of work,” in which society shields certain roles not due to technical limits but to enduring beliefs about dignity, care, and meaning. By distinguishing performance- from principle-based objections, we provide a framework for anticipating and navigating resistance to technology adoption across domains. 

 

 

When AI use is morally repugnant

Researchers used a moral repugnance scale (1-7) to measure public resistance to automation across 940 occupations. They found widespread support for AI in some roles but others remain categorically off-limits, regardless of AI’s capabilities.

Occupation	Repugnance score
Clergy	5.91
Childcare workers	5.86
Marriage and family therapists	5.64
Administrative law judges, adjudicators, and hearing officers	5.62
Athletes and sports competitors	5.52
Biostatisticians	2.54
Switchboard operators, including answering service	2.52
Transportation planners	2.38
Search marketing strategists	2.31
File clerks	2.17