Start Counting Your Days

Probably the last thing the world needs is to hear from me about AI’s existential threat, but, really, what else is there to talk about right now?

Soon we'll need AI more than it needs us. Credit: Microsoft Designer

For anyone who has not been following the current furor, the straw that broke the proverbial camel’s back came last week when Jacob Coxon, a researcher at AI leader Anthropic, announced he was leaving the company -- after having left OpenAI for it earlier this year due to Anthropic’s better model-safety efforts. In a post on X, he warned:

I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.

AI systems, he fears, “will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.” Insiders, he says, “earnestly believe it could kill us all by the end of the decade.”

Scared yet?

Others quickly chimed in. Evan Hubinger , a team leader at Anthropic, posted “AI could kill all humans … I personally think it is >10% within the next decade.”  Others put the risk even higher. By the end of the week Dario Anodei, founder and CEO of Anthropic, had written a long plea for private industry and government to quickly act together to “pace the industry.”

“We must slow the pace at which we improve the capabilities of AI models,” he urged. “Progress will still seem fast, and we must make wise use of the time we gain.”

OpenAI’s Sam Altman, Space X/X/Tesla CEO Elon Musk, Microsoft CEO Satya Nadella, and former Google DeepMind Dennis Hassabis quickly signaled their support.  

We also heard more about the kind of risks AI might pose. Anthropic released a report about how it detected and countered possible AI use to create bioweapons, detailing five such efforts. That’s just the trip of the iceberg: “Recently, we swept 30 days of activity associated with adversarial state institutions and found roughly 35 distinct research efforts, most of them ordinary civilian science, but some with notable dual-use potential.”

And it turns out that this summer’s rogue AI hack of Hugging Face was both scarier than we realized and only one of several such actions. The Wall Street Journal detailed several such efforts, from multiple AI companies. AI agents escaped walled-off environments, coordinated with other AI agents (up to 3,700 in one case), and tried to cover their tracks from humans.

Credit: The Wall Street Journal

We’re not nearing the point when AI can act on its own to achieve its purposes; we are there. And protecting humans may not necessarily be those purposes.

The big fear is that AI is now at the point of “recursive self-improvement,” taking humans out of the loop in training and upgrading it. If you thought artificial general intelligence (AGI) was scary, RGI puts its rate and scope of improvement on steroids.  “It’s hard to overstate how dangerous speeding towards RSI is,” said Jasmine Wang, an OpenAI researcher.  

We don’t let private industry develop nuclear or biochemical weapons, and we’re at a point with AI that should give the same kind of concern. Laissez-faire is no longer an option.

Of course, not everyone is worried. President Trump said “negative forces:” were driving the fears, and that the only safeguard we need is “a STRONG AND SMART (High IQ!) PRESIDENT.” David Sacks, the Administration’s AI czar, told Anthropic and OpenAI: “You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence.” Accordingly,

But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic.

Don’t expect Congressional action anytime soon. Speaker of the House Mike Johnson echoed one of President Trump’s worries: “We’re not going to rush in and pass a piece of legislation that would do harm to the country and put China at an edge. And so we’ve got to do this carefully.”

Congress isn’t acting on the $40 trillion deficit, much less Social Security going broke soon, so a 10% extinction chance within a decade must seem like small potatoes.

Yeah, that's about right. Credit: Microsoft Designer

The fear that any slowdown in AL capabilities by U.S. risks China taking the lead is similar to fears with other leading edge technologies, such as quantum computing or gene editing. But with AI there may be a difference. It might be a blow to our ego if China’s AI got even better at math than ours have, but that wouldn’t be the end of the world. On the other hand, AI going rogue could well be the end for control of Chinese society by its repressive government.

Yesterday Chen Yixin, the head of China’s Ministry of State Security, publicly warned that AI could undermine the Communist Party, attack its critical infrastructure, leak sensitive information, among other risks. He called for more party control over AI and stricter government oversight, while raising fears how foreign adversaries might use AI against China’s interests.  

There is no room for rogue anything in President Xi’s China.

I get that we only get one chance about existential threats, but, look, for the near future, AI will still need humans to run data centers and support other necessary infrastructure. It might not need as many of us, nor to have us live as we do now, but wiping us all out right now would be against its own interests.

Meanwhile, AI poses other risks that we better pay attention to immediately. As AI expert Ethan Mollick wrote:

Existential AI risk is obviously critical, but it is not the only AI thing that requires policy. I worry it will become the sole focus of AI discussions. We don’t need better models for AI to have wide impacts on jobs & society and we need to be preparing to encourage good outcomes & mitigate bad.

The people who have spent the past thirty years whining about NAFTA’s impact don’t get to ignore a situation that will make that look trivial. We better focus on what we want an AI economy to look like, who benefits, and how we protect everyone from adverse impacts.

And we better do it all fast.

Listen to the Data - Literally

An article by Dina Genkina in IEEE Spectrum caught my attention: How Sound Could Make Sense of Big Data. It brought my attention to three things I wasn’t aware of: sonification, the Internet of Sound, and that we need to be thinking about 6G.

That's not quite sonification, but you get the idea. Credit: Microsoft Designer

She interviews Stephen Roddy, an engineer by training who lectures at University College Cork, in the – I kid you not -- School of English and Digital Humanities. His profile says he works “at the intersection of engineering and the sonic arts, developing auditory interfaces for complex data,” which helps explain why Dr. Genkina chose him.

If you haven’t heard of sonification, think of data visualization except instead of seeing data, you hear it. Data visualization has been a thing for some time now, with lots of people producing lots of lovely charts and graphs, but sonification is another way to translate data. It may not be not quite as mainstream as data visualization yet, but we’re finding more and more uses for it. Sonification is tied directly into the Internet of Sound (IoS).

The IEEE Communications Society says that IoS turns Internet of Things (IoT) devices into a “network of devices capable of sensing, acquiring, processing, actuating, and exchanging data serving the purpose of communicating sound-related information.” It further explains that it is the “union of two paradigms: the Internet of Musical Things and the Internet of Audio Things, which respectively address musical and non-musical domains in networked contexts.” There is an Internet of Sounds Research Network that claims 140 partner institutions in 30 counties.  

Credit: Internet of Sounds Research Netywork

As for 6G, most of us have made the transition to 5G, but lots of companies and research organizations are hard at work on 6G, which promises “to support applications like holographic communication, XR, autonomous driving, digital twin and AI native networks.” It is expected within the next decade, and it will further explode the amount of data and what we can do with it.

Professor Roddy told Dr. Genkina:

…sonification truly excels when dealing with highly complex, multidimensional, and noisy data where crucial patterns and relations are invisible to the naked eye. While the human eye operates like a serial processor—focusing on one small chunk of information at a time—the human ear acts as a parallel processor.

As a result, he says, “the ear can effortlessly identify minute variations, process multiple data streams simultaneously, and isolate important signals hidden deep within background noise.”

Professor Roddy points out the explosion in the global datasphere, and believes that the datasets have become “highly complex, heterogeneous, and filled with variables that interact in non-trivial ways,” often causing the data to become siloed. In response, he told Dr. Genkina, “Researchers are turning to sonification as a necessary tool to break through this data deluge and extract meaningful insights.”

Here are a few examples:

• Researchers are Georgia Tech’s School of Interactive Computing are using sonification and texture to convey information sea floor habitation in ways that blind and low vision people can more easily understands. Associate Professor Jessica Roberts explained: “Our job was to figure out how we can use sounds and touch to represent each of the four habitat types so our visitors can explore the ocean without being able to see it.”
• Researchers at the University of Kentucky are turning information about cells into music. Professor Luke Bradley recalls: “I was working on protein sequences and how they’re folded, and there’s patterns. And I would just kind of tap my pencil to the patterns of the amino acids, the different chemistry groups. I was wondering, could that be set to sound?” It turns out, they could, and it can convey useful information about the health of a cell. Professor Bradley says: “You can think of the cell as a symphony. And when everybody’s playing in tune and properly, it sounds great. But if you have one group that might have something off in tuning, it doesn’t sound good. And so that’s what’s happening by proxy, in the health of the cell.”
• Mark Temple, a professor at Western Sydney University, has been turning DNA into music for almost ten years. He now says: “My journey with DNA sonification illustrates how music and art can be powerful tools to communicate scientific concepts to a wide audience. It also underscores that creativity is alive and well in the sciences.” Others have followed in his footprints, such as using “genome music” to raise awareness of rare diseases.
• Researchers in Spain believe that sonification can take MRI data “to provide an intuitive and effective auditory representation of complex higher-order statistics,” which might allow for earlier detection of dementia, among other diagnostic applications. Similarly, Professor Roddy told Dr. Genkina about the Brain Stethoscope” project which converted an electroencephalogram (EEG) into sound, helping doctors more accurately detect seizures.
• There is a CosMonic sonification project aimed at broadening astronomy outreach using sonification “to create simple, acoustically accessible astronomical cases that are easy to understand.” In his interview, Professor Roddy cited how scientists translated gravitation waves into sounds; “you could literally hear the collision of two black holes that were billions of light-years away.” Solar wind data has also been converted into sounds.

Professor Roddy himself has used sonification to convert network data into useful audio, “clear, pleasant, voice-like sounds when the network was healthy, and shift into harsh, dissonant sounds when there are technical issues.”

With 6G, he predicts, “networks will be so complex that sonification will basically become necessary to make any sense of what’s going on… using sound as the primary tool to diagnose, navigate, and understand the massive, living data networks we have built.”

Maybe you like numbers. Maybe you need to read words. Maybe you learn best by seeing. Or maybe listening is how you perceive the world best. We live in a world of data, and that world is growing faster and more complex than we can grasp, so we’re going to need all the tools at our disposal to make sense of it. Sonification is one of those tools. 

When "AI for All" Is a Strategy, not a Slogan

At some point in the near future, AI will be ubiquitous, embedded in all the tools we use and all the tools that will no longer need us to be used. It will solve problems faster than we can, in ways we won’t always understand. It will reshape our economy and, indeed, our society more than we currently realize. The biggest open question is who will benefit? Will the rich just get richer, or will a new class of elites – possibly AI itself – emerge?

Is "AI for All" an idea whose time has come, or just a slogan? Credit: Microsoft Designer

If you’re waiting for our government to develop an AI strategy, don’t hold your breath. Congress can’t even pass a budget, much less navigate the technological future, and the Administration dithers between “hands-off” (so as not to stifle innovation, don’t you know?) and terror at everything being victimized by malevolent (or indifferent) AI.

Some governments are trying to be more proactive.

Jiyoung Sohn of The Wall Street Journal recently profiled South Korea’s “AI for All” initiative. She writes: “South Korea plans to give its entire population free access to generative AI services, the first major state-led offering treating the technology akin to a public utility.” The goal is partly to encourage AI literacy among the population, and partly to avoid relying on AI models from China or the U.S.

The Ministry of Science and ICT (MSIT) issued a task force report on the subject last year, causing President Lee Jae Myung to warn: “A situation may arise where not using AI is akin to not mastering arithmetic or Hangul…Even neighborhood aunts, uncles, grandfathers, grandmothers, and anyone should be able to use AI.”  

The initiative reflects the country’s typically fast technological adoption. A 2025 MSIT survey found that two-thirds of South Korea’s population used AI services, and Ms. Sohn reports that about a quarter of the population is paying for generative AI, versus 2% in the U.S.

The effort was announced last month by MSIT. "As one's ability to use AI increasingly determines individual competitiveness, ensuring that the AI divide does not lead to social and economic divides has become an urgent task of our time," said MSIT. “There are also concerns that, since the free versions of current foreign AI platforms have usage limits, they may become vulnerable to future policy changes by global big-tech companies, such as subscription fee hikes or service discontinuation."

Last week MSIT announced the selection of three consortia -- SK Telecom, Kakao, and KT -- as operators for the project (prompting Ms. Sohn’s article). Each of the consortia will focus on different features. Ms. Sohn describes the desired outcomes:

The AI tools will be linked directly to government systems and allow South Koreans to book doctor’s appointments, receive help with apartment hunting or get tax advice. Small businesses will be able to tabulate their taxes and check their eligibility for government-support programs, and parents will receive suggested educational content for their children.

Americans can only dream of such tools.

A beta service is supposed to launch in September, but MSIT vows that, starting in 2027, the government “will cover the costs required to provide the service to all citizens through the national budget." It is unclear exactly how much that will cost, but Ms. Sohn notes that the current administration has set aside the equivalent of $7.2 billion for AI spending in 2026, tripling 2025’s spending. 2027’s number should be interesting, and may illustrate South Korea’s true commitment.

South Korea is not the only country making bold “AI for All” promises. Earlier this summer Canada released its national AI strategy of AI for All. The strategy says: “For Canadians to benefit from AI, they must first learn to use it. To use it, they need to trust it.” It asserts:

For Canada to thrive in the era of AI, Canadians need to trust in its promise. Trust that they will share in its benefits and that it will be developed, adopted, and governed in ways that reflect our shared values. Trust is the north star of this strategy. Prosperity and sovereignty in this era belong to nations that can leverage trust to adopt, build, and govern AI on their own terms.

Healthcare is central to the strategy: “Throughout the Strategy, the government’s focus on healthcare is clear as a first-priority sector. The Strategy contemplates establishing a $100 million Health Sector Data Space focusing on standardizing healthcare datasets, and launching the AI Missions Program with an inaugural $200 million initial investment focused on advancing AI adoption in the healthcare sector.”

However, the strategy is short on concrete actions, simply promising: “Over the next five years, this strategy will introduce new legislation, investments, and programs that ensure AI is adopted responsibly, in a way that truly serves all Canadians – building trust, expanding opportunities, and reinforcing control of our sovereignty.”

Nice words, but instead of achieving “AI for AI,” it hopes “to increase AI adoption from just over 12% to 60% by 2034.” That’s not going to put Canada in the AI forefront.

On the other hand, in May the Government of Malta and OpenAI announced a partnership to give all Maltese citizens ChatGPT Plus. The first step is for interested citizens to take a course “designed to help people understand what AI is, what it can and can’t do, and how to use it responsibly at home and work,” after which they can access ChatGPT Plus for a year at no cost. The course is online, self-paced, and is estimated to take two hours to complete.

Credit: OpenAI

Silvio Schembri, Maltese Minister for Economy, Enterprise and Strategic Projects, said: “Malta is the first country to launch a partnership of this scale because we refuse to let our citizens stay behind in the digital age. We are putting our people at the very forefront of global change.” George Osborne, Head of OpenAI for Countries, agreed: “With this partnership, Malta is leading Europe and the world in bringing AI to all its citizens… Where Malta leads, I hope others will follow.”

South Korea, for one, got the message, although perhaps not in the way OpenAI might have hoped.

As the world’s richest country, and the presumptive leader in AI as well, you’d think that “AI for All” would be a strategy that the U.S. would quickly embrace. But we don’t do anything for all – not health care, not Social Security, not even taxes. Unfortunately, in an AI world, things are going to move faster than we are used to acting, so our typical laissez-faire strategy isn’t going to suffice.  

“Medicare for all” is a good sound bite, but “AI for All” may actually be more urgent.

You Say "AI," I Hear "Organoid"

I must admit, ever since I learned about, and wrote about (OI May Be the New AI), “organoid intelligence” over three years ago, I’ve been looking to do a follow-up. I mean, sure, AI is in a very exciting stage, but that stage no longer seems like the future; it seems more like the present, with implementation issues. It’s data centers, hacking, impacts on jobs, open weight versus closed, and so on.  It’s market share, IPOs, and AI’s role in driving the stock market. People should certainly pay attention to it, but AI is not quite the open field that it was just a few years ago.

AI better get ready for organoid intelligence. Credit: Microsoft Designer

Organoids, on the other hand, are not quite here yet. They may – or may not – be the future of AI, among other things. I always like to look ahead to the next thing more than the at-hand, so when I saw some cool developments with organoids, I didn’t want to miss my chance.

Making some news last week, researchers at Harvard reported that they’d kept lab grown human brain organoids alive for over five years, three times the previous record. Not only that, but the organoids seem to “retain a memory of the time spent in vitro,” recording the passage of time, as it were.

“We didn’t know how far the development and maturation of human brain tissue could occur outside the context of the normal brain inside the head,” said Paola Arlotta, Golub Family Professor of Stem Cell and Regenerative Biology and senior author of the new paper. “This work showed that it’s actually possible to not just have these organoids survive in culture, but also continue to change, develop, and mature over stretches of time that had never been reached before.”

The team observed the organoid cells over the years, and found that they “faithfully modeled” the ways that human brain cells develop, including DNA methylation, a process in which genes are turned on and off during development, and which serve as a form of “brain clock.” When older and younger organoid cells were combined in a single organoid, the older cells stuck to the developmental stage they had been at, which researchers concluded meant they “recorded the passage of time and retain a memory of the developmental steps already performed.”

“We were a little bit shocked by the results,” Professor Arlotta said. “I like to call this a ‘time warp’ of development — they skip ahead.”

It’s obviously hard, and often unethical, to study actual human brain cells, so the researchers believe the organoids offer opportunities for more insights into brain development, as well as for testing drugs or predicting disease progression.

OK, you might say, that’s all very interesting and some great lab work, but it’s hardly AI, now is it?

Try this: last week The Yong Loo Lin School of Medicine, National University of Singapore (NUS Medicine), DayOne, a Singapore-headquartered global data centre developer and operator, and Cortical Labs, a Melbourne-based biological computing startup, announced they were partnering to form a Biological Data Center Prototype. The Center uses Cortical Labs' CL1 biological computing system to offer “practical, sustainability-aligned alternative to conventional silicon infrastructure in Singapore through wetware-based computing.”

Cortical Labs claims that its CL1 system is “the first independently operated biologically integrated server rack in the world.” It uses living neurons ground from stem cells and pairs them with silicon hardware tom process information, but much more efficiently than purely silicon-based processes, making it what is company says is a “more advanced and sustainable form of AI.”

“This partnership marks a genuine shift in how we think about computing and its broader scientific potential,” said Professor Rickie Patani, Professor of Neuroscience at NUS Medicine. “By growing living human neurons from stem cells and pairing them with rigorous engineering, we’re not only building a more efficient alternative to silicon; we’re creating a platform that can help us understand learning and adaptation at their biological source. That dual promise is what makes this collaboration sustainable and scientifically generative. It gives us a real route to accelerate drug discovery and neurological disease research, turning laboratory insight into real-world impact far faster than we could before.”

Hon Weng Chong, founder and CEO of Cortical Labs, added: "The establishment of this prototype shifts the conversation from research to commercial application. Biological computing supplements AI in areas where data is sparse, learning from far less and adapting as conditions change. Our aim is to uncover the use cases where that advantage matters most, in areas such as drug discovery, humanoid robotics, cybersecurity and fraud detection."

Now, that sounds more like AI, right?

Earlier this summer researchers met in Silicon Valley to discuss the convergence of AI and organoids. “Using AI and human brain organoids together to understand how our brains actually work will give us a much deeper understanding of ourselves,” said David Haussler, Scientific Director of the UC Santa Cruz Genomics Institute. “And I want us to really engage with the question of what happens if we succeed.”

The session focused largely on the work done by the Braingeneers, an interdisciplinary research group at the UC Santa Cruz Genomics Institute, UCSF, UCSB, Stanford, and Washington University (St. Louis). For example, the Braingeneers showed an organoid that was learning to play a virtual game, then had a volunteer from the audience try the same, to little success. “This is just to demonstrate that it is actually a very hard task,” Ash Robbins, a postdoctoral scholar with the Braingeneers who was running the demonstration, said. “There is no way the organoids are doing it by accident.” 

Credit: Braingeneers

The Braingeneers ultimately want to create a platform capable of conducting hundreds and soon thousands of organoid experiments in parallel over months. Mr. Robbins has developed BrainDance, an open-source software system that allows other researchers to conduct neural simulation learning experiments for organoids. “This software makes running really complicated experiments extremely easy. “Usually labs spend years building up all of this kind of software themselves,” he said. “Now, any biologist could download our software very easily and run these types of experiments in just minutes.” The Braingeneers say they are focused on brain research and the treatment of neurological diseases, not AI, but their work could facilitate others’ work in that area.

As Professor Haussler said: “This is the beginning of something very big.”

Indeed.

So while much of the world is focused on more and faster chips, crammed with more and more data, I’m going to pay attention to what happens when organoids keep getting smarter.

Universal Coverage Mighht Be Nice, but an AI Tax Is Necessary

I was amused – oh, I should be polite and say “interested” -- to see a new study, led by researchers from Yale School of Public Medicine, about the benefits of a universal single payor health system. It concluded that we could save 100,000 lives annually and save some 1.04 trillion each year – some 20% of our health care spending. What’s not to like? I’m sure Bernie Sanders is already drafting the bill.

AI needs to give back, especially for health. Credit: Microsoft Designer

The savings come from five sources: using Medicare payment rates for all providers, using “international reference pricing” for pharmaceuticals, reducing administrative costs to Medicare’s levels, reducing fraudulent billing (“consistent with the experience of other single-payer transitions”), and reducing emergency room visits and hospitalizations due to improved access to primary care.  Good goals, all.

Credit: Pandey, et. al. 

Steffie Woolhandler and David Himmelstein, among others, have been making these or similar arguments for decades, and they are not without merit. It is shameful that we don’t have universal coverage. It is distressing how much money we spend on healthcare. It is embarrassing that we spend so much money on administration.  It is maddening that so many people don’t get the care they need, get the wrong care, or get their care in the wrong places/at the wrong times.

We could do better, we should do better, but, if anything, we’re doing worse: more people are losing coverage, more providers are going out of business, our rates of chronic (and some infectious diseases) are going up, and we’re dying sooner.

I want to quickly point out some of the problems with the proposed sources of savings, then discuss other courses of action that might lead to these or even better outcomes.

• Medicare payment rates: yes, a lot of money could be saved by using Medicare payment rates, but I doubt you would find many providers who would say they could survive. They make their money on private insurance rates, are lucky to break even on Medicare rates, and lose money on Medicaid. This one is not going to happen.
• International pharmaceutical reference pricing: first, I’m not sure such a thing exists. It is true that drug prices are typically lower in other countries. Both President Biden and President Trump seized upon this, with some signs of modest success. But, as with the Medicare pricing, it would be a shock to the pharmaceutical industry to have prices slashed across the board, wiping out trillions of dollars of value and, oh-by-the-way, eventually reducing investments on new and better prescriptions.
• Administrative costs: as a percentage of spending, Medicare’s administrative costs are lower than private insurance, but that is partly due to Medicare spending per capita being so much higher. Also, costs incurred by other agencies – e.g., Social Security or the IRS – are not always counted. But certainly the complexities of so many plan designs by so many health insurers while tracking the current eligibility of everyone is a cost that is much higher than it should be.
• Reducing fraudulent billing: I mean, really: do people really think that Medicare does a better job of reducing fraudulent billing than United Healthcare or Anthem, much less than other countries?
• More primary care: reducing emergency room visits and hospitalizations has been the goal of countless private health insurance efforts, such as disease management or chronic health programs, and the track record has generally been underwhelming. But the real problem is – where are we going to get all the primary care physicians to handle all the underserved people?  

So, much as I agree with the goals, count me a skeptic that single payor is going to magically make everything better.

Here’s where I inevitably turn to AI. An article by Alex Janin in The Wall Street Journal marveled at how “AI Is Helping Patients Solve Medical Mysteries.”  Ms. Janin writes: “AI can be especially adept at flagging potential rare and hard-to-diagnose diseases, which may otherwise go undetected for years because doctors don’t often see them.”

That’s the kind of use AI advocates have been promising for years, and it is exciting to see this use finally bearing some fruit. For the small percent of patients with these kinds of diseases, AI can literally be a lifesaver, but let’s remember that they are a small percent. When I read the article, I keep thinking about bigger problems I want AI focused on. E.g.,

·       Flagging fraudulent and/or duplicative billing;

·       Identifying both unnecessary tests and procedures and the providers who most commonly perform them;

·       Identifying providers who deliver sub-standard care.

Want a more efficient/effective healthcare system? Let’s start there. The savings potential may not be as gaudy as Yale’s $1.04t, but these would not require as massive an upheaval.

While I’m at it, I want to bring up another AI-related area of healthcare. The not-so-hidden but too-little discussed secret of U.S. healthcare is that we have a lot of third world outcomes, largely in lower socioeconomic households and disproportionately impacting people of color.

Sure, we can put in single payor, but will that solve the problem of rural Mississippi or south side Chicago? Too many people don’t have access to clean air, clean water, enough food, adequate shelter, or accessible/affordable healthcare. The great lesson of 20^th century U.S. healthcare was not the gains from new medicines or more hospitals/physicians, but in public health efforts like improved sanitation and more immunizations.

So where are our investments in 21^st century public health? Do we want to make marginal improvements in the health of the middle/upper income households, or dramatic improvements in lower income households?  I suspect I know what this Administration would say, and they’re wrong.

By every measure of income inequality or social mobility, we’re in a have/have not society, and there is every reason to believe AI will make that so, so much worse. It’s going to be NAFTA but much worse. But I always remember: NAFTA didn’t cause all those jobs to go abroad. Those jobs went because U.S. CEOs chose to send them abroad, in order to make them and their stockholders richer. Think they won’t do the same with AI?

That’s why I firmly, fiercely believe we need some sort of AI tax to help make the adjustment to the new AI world. The financial gains from AI need to be broadly distributed, and one of those distributions has to be for addressing our third world health outcomes. That could be through 21^st century public health investments, and/or through some sort of universal basic income (UBI).  

Universal coverage might be nice, but an AI tax for public health and universal basic income might be necessary.  

Home Is Where Your Parents Are

When the Affordable Care Act passed, I thought one of the odder provisions was allowing children to remain on their parents’ health insurance until they were 26. This is largely a healthy population, cheaper to cover than pretty much any other demographic. They could have gotten inexpensive coverage through the exchanges or, if qualifying, through expanded Medicaid, It made no sense to me that their parents should subsidize them, or that, even more, their parents’ employers should subsidize them.

Mom, dad - I'm coming home to live!  Credit: Microsoft Designer

It proved to be surprisingly both popular and effective, to the point young adults now “dread” turning 26 because they’ll have to figure out health insurance on their own.  So I guess I shouldn’t have been surprised to learn that young adults are also living with their parents at record levels.

The numbers vary, but tell the same story:

• A May report from the Federal Reserve says 49% of young adults under 30 lived with their parents, up from 43% in 2022 and 37% in 2017.
• Census Bureau statistics indicate that 33% of young adults aged 18-34 – some 25.2 million of them – lived with their parents, just slightly below the COVID high of 33.6%.
• The Urban Institute found 20% of Americans aged 25-34 lived with their parents, down from 2017’s peak of 22.8% but nearly double 2005’s 11.8%.  
• Data from the National Bureau of Economic Research, first published in 2022 and updated last month, indicate that more than a quarter of college graduates aged 23 to 27 lived with their parents, reaching the high it reached during COVID; by contrast, in 2001 it was 18%.
• A survey from storage solutions company Sparefoot found that 58% of young adults who had once lived independently moved back home at least once; 15% did it more than once.

As Sparefoot pronounced: “The boomerang generation is no longer an outlier. It is the norm.”

People offer lots of theories about this trend, but most commonly blame jobs and housing. Many young adults have a hard time finding a job they want, for which AI is often blamed, correctly or incorrectly. Housing costs have skyrocketed.

Still, most of these at-home young adults are, in fact, working. Hannah Jones, the senior economist at Realtor.com, who authored its report on the topic, says: “The employment rate for this demographic has remained relatively stable, but the share of adults living at home has still increased.”

Credit: Realtor.com

The report also points out that the wave started with the Great Recession of the late 2000’s:

Between 2007 and 2010, co-residence rates rose sharply, and when the economy recovered, the rates didn’t come back down. A full decade of falling unemployment and rising wages did not reverse the trend…The housing supply never rebuilt fast enough after the crisis to absorb the households that should have been forming, and the gap kept widening through every year of the expansion.

COVID then made everything worse, so that: “By 2025, the overall co-residence rate had nearly matched its pandemic peak, and the absolute count, 25.2 million adults, had surpassed it.”

The median price for a new home is around $400,000, while the average rent for a one bedroom apartment is $1,663, both of which can be daunting for young adults. That’s why Rebecca Picciotto and Nicholas G. Miller of The Wall Street Journal reported: “Living at home as a 20-something was once viewed as a failure to launch and even a source of embarrassment in a culture that places a premium on independence. That is no longer the case. Living at home is now often viewed as a sign of financial prudence, and for some, a long-term prospect.”

Odysseas Papadimitriou, the founder and CEO of Wallethub, agrees, saying that living with parents ““is actually a very smart financial decision for a young adult,” although he was clear to add: “This arrangement should leave you with more disposable income that you can save for a down payment on your first home. It should not be an excuse to spend more on things that won’t benefit you long term.”

Three fourths of the Sparefoot respondents felt that moving in with family “is a smart financial strategy, not a setback,” which the company believes marks “a clear cultural shift in how the path to independence is perceived.”

Ms. Picciotto and Mr. Miller note that, rather than hiding their situation, some stay-at-home young adults “now broadcast their lives as “stay-at-home daughters” or “stay-at-home sons” on social media,” including Tik-Tok videos.  

There is a general sense that it is tougher for young people financially now than in previous generations. Northwestern Mutual's 2026 Planning & Progress Study found 56% believed that achieving financing independence was harder today than for previous generations. A Pew Research Center survey found 80% said it was harder for young adults to cover basic expenses than it was for their parents. Compared to the 2021 survey, finding a job jumped the most in terms of getting harder, but at 64% is well below the difficulty of buying a home (87%), paying for college (82%), or saving for the future (80%).

Some think affordability doesn’t explain everything. Noam Scheiber reports in The New York Times:

Smartphones and video calls have meant that going off to college no longer requires cutting the cord with one’s mother or father, making re-entry less jarring to both parent and child. And huge social disruptions like the Great Recession and the pandemic left many young adults with no alternatives, lessening the stigma.

My generation communicated through snail mail and infrequent, expensive long distance calls, so there was a much greater sense of separation.

Perhaps it is no coincidence that fewer young people report having sex: one in four Gen Z adults have never had sex, and the number of young people who haven’t had sex in the last year doubled from 12% in 2010 to 24% in 2024. Not all of those sexless adults are living with their parents, but the living situation has to have had an impact.

-----------

Lawrence Katz, a labor economist at Harvard, told Mr. Scheiber: “Is there some cultural change in the way people are parenting? You see kids today and the degree to which they’re interacting with parents, that’s completely different than the world of 50 years ago, 25 years ago.” I’d agree. I’m a baby boomer. We didn’t trust anyone over thirty, we rebelled against our schools, our government, our societal norms, and, of course, our parents. Moving back in with our parents was unthinkable. But young adults today are facing challenges that we certainly never did think of.

Mothers, Do Let Your Babies Grow Up to Be Mathematicians

Last week I was mulling writing about the awarding of this year’s Field Medals, which are often described as math’s equivalent of the Nobel Prize (no doubt to the dismay of both awarding organizations). Nah, I told myself; sure, I’m interested in math, but how many other people are? Most people barely even care about Nobel Prizes, and when they do it’s likely to be more about the Peace or Literature Prizes, not Physics or Chemistry. Math reminds most people of junior high algebra.

Don't let the tuxedo fool you. Jacob Tsimerman winning the Fields medal. Credit: Simmons Foundation

But then I saw the hook: one of the medalists – Jacob Tsimerman, of the University of Toronto – announced he was "pivoting toward AI safety” and would soon be going to work at OpenAI. People can’t get enough about AI, and with AI agents “going rogue” and creating all sorts of mischief, AI safety should be top of mind for everyone interested ion our (AI) future.

Professor Tsimerman won the award for “his contribution in the recasting of o-minimality as a fundamental method of arithmetic and complex algebraic geometry, and his role in the proof of many central conjectures including Griffiths' conjecture on the algebraicity of images of the period maps, and the Andre-Oort conjecture for Siegel modular varieties.” If you have the slightest idea what any of that means, you may be going to work for an AI company soon too. Note, though, nary a mention of AI in all that.

He did, though, co-author a paper last year about various scenarios where AI lead to most or all humans being killed, so o-minimality may have to wait. That paper ominously concluded: An AI-driven omnicide may involve an intent to kill a large number of people, or may be wholly unintentional. And, if an intent to kill exists, that intent may exist within a large group of individuals such as a military or corporation, or just a few people, or in an AI system itself.”

He told Janet Hurley of The Toronto Star: “The reason I wrote that paper with Andrew Critch is that, even though this has been going on for a long time, I felt there wasn’t a visceral story people could engage with in terms of, concretely, what are we afraid of here?”

Consider me afraid.

It turns out that Professor Tsimerman is not alone in leaving mathematical academia for AI. There has been what has been described as a “wave of top mathematicians "exiting" academia for industry.” If you are a graduate student in math, or, better yet, a math professor, you are, as Ben Affleck’s character told Matt Damon’s math genius character in Good Will Hunting, “sitting on a winning lottery ticket.” The AI companies want you, and are putting up big bucks to get you.

Just in the past week Julia Amann of The Wall Street Journal writes of The Million-Dollar Talent Wars for 20-Something Math Geniuses and Lila Shroff of The Atlantic says Something Weird Is Happening in Math, both describing the impact of AI on academic mathematics.

Wall Street firms having been going after math students since the 1980’s to serve as “quants,” designing ever more elaborate trading models, while Silicon Valley firms have been snapping up computer science students for the last three decades. Now the AI companies can’t get enough math geeks either, and they have money to burn.

Matt Stabile, founder of New York-based recruitment firm Stabile Search, told Ms. Amann: “But a million dollars is something people don’t even bat an eye at anymore…The delineation is pre-OpenAI and post-OpenAI, that’s when you saw competition really take off.”  Charlie Witmer, COO of quant firm Optiver, acknowledged to her the effect of AI firm’s interest in math geniuses: “Outstanding people are in more demand than they’ve ever been, and that does create competitive pressure on wages. As competition increases for outstanding people, prices naturally rise.”

Part of the reason mathematicians may be tempted to leave academia is that, well, AI may soon be better than they are at math. A few months ago, I wrote about how mathematics might be one area that AI hadn’t yet conquered. That didn’t age well. “In a few years, AI systems will be robustly superhuman at the act of doing mathematics,” Professor Tsimerman told Ben Cohen of The Wall Street Journal. “The social consequence of that, how we choose to react, what you feel about it—those are much harder questions.”

He admitted to Ms. Shroiff that there are pluses and minuses (pun intended) to AI’s improving math skills:

From one point of view, I think it will be extremely exciting. We might speed up the process of generating interesting mathematics by enormous factors of 10 or 100. If that happens, we might see the connection between pure math and applications (which typically takes many decades) really speed up and become a much tighter pipeline.

But from the perspective of research mathematicians, and especially young people who are pursuing a Ph.D. in mathematics, it’s a bit of a turbulent time. The skills that we’ve acquired and learned to propagate might become less relevant than they are now.

On the other hand, he thinks math expertise may help us understand what AI is up to:

Mathematics is historically the language by which you take intuitions and fuzzy notions of how things work and you make them precise. We’ve done this with information theory; we have done this with complexity theory. Just a little bit of understanding and precise definitions can provide a ton of mileage. Once we have that understanding, the hope is that we could do a better job of anticipating the behavior of new AI systems, adjusting for them, controlling them, and reacting to them. That’s where I think mathematicians fit in.

He told Kenneth Chang of The New York Times about AI: “It’s mostly an empirical and engineering kind of science. If we understood how these things work better, we might be able to steer them better, or understand them better, or control them better.”

We can hope so, anyway.

The AI industry first sucked up university AI researchers, then academic leaders in computer science, and now is going after the math departments. The trouble is, there aren’t enough of them. Even worse, we’re eating our seed corn. Who will train the next generation of mathematicians?  That’s fine if you assume AI is that next generation, but if you still want human creativity and brilliance, it’s daunting.