✅ People are turning to chatbots for health advice. ✅ The explosive development of humanoid robots. ✅ The invasive Benjamin Button. ✅ The rapamycin debate. ✅ Nvidia goes bio/acc.

🤙🏼 Want to connect? Add me on LinkedIn. 🙏🏼 Not subscribed to the LEVITY podcast on Youtube yet? Do it here. 🎧 More of a listener? The LEVITY podcast is also available on Spotify, Apple Podcasts and other places.

There’s a question I have been pondering lately. Will there be any human doctors in a few decades?

If that sounds preposterous just wait until you hear my answer.

Let’s examine some of the signals pointing in that direction.

The other day The New York Times reported of a study that had ChatGPT - all by itself - outperform human doctors. Not only that, but it fared better than doctors that also had access to ChatGPT.

This was puzzling to the researchers conducting the study. The conventional wisdom held that AI + human should be superior. But don’t be surprised if we find this assumption repeatedly challenged in the years to come.

Turns out the doctors that had access to ChatGPT wouldn’t listen to what the chatbot had to say. They had already made up their minds - they were the experts, after all. A classic case of confirmation bias reinforced by overconfidence bias.

There are many, many other studies and reports of foundation models* being used to diagnose or predict disease.

* Think of a foundation model as an AI system capable of generalizing across various tasks, unlike AI systems designed to excel at only a specific, narrow task. A large language model or multimodal model like GPT-4 is an example of a foundation model. ”Multimodal” means it can process multiple types of input, such as language and images, rather than just one modality.

I think it’s uncontroversial to say that quite soon it will be considered malpractice for any human doctor to not consult AI when treating a patient. Artificial intelligence simply have advantages we can’t compete with.

For example:

Indeed, patients are already consulting chatbots. In a Ted Talk last year Eric Topol, the famous cardiologist, told us the story of the patient Alex*, a six-year old. Alex had ”relentlessly increasing pain, arrested growth, dragging left foot with gait, severe headache”. Three years had passed and Alex had been examined by 17 doctors. No one could figure out what was wrong.

Then his mother turned to ChatGPT. Provided the chatbot with all of the symtoms. ChatGPT suggested tethered cord syndrome. He had surgery and is now perfectly healthy.

This story highlights a major flaw in modern medicine: its siloed nature. Alex’s mother noted in an interview with Today, ”no matter how many doctors the family saw, the specialists would only address their individual areas of expertise”.

* In the talk Eric Topol has the patient named as Andrew, which is incorrect.

Another anecdote from The New York Times: a woman in her 60’s whose ”right side of her face was sagging, she tripped over words as she spoke and her head hurt so much that she couldn’t rest her head on the pillow”. She and her husband drove three hours to the emergency room where she was examined and sent home - the doctor saw no cause for concern.

At a loss, she turned to ChatGPT for the first time. The chatbot suggested Bell’s palsy, a form of facial paralysis that requires urgent treatment. Returning to the ER, a different doctor confirmed the diagnosis.

According to KFF, non-partisan organization focused on health policy, about one in six adults (17%) say they use AI chatbots at least once a month to find health information and advice, rising to one quarter of adults under age 30 (25%).

Dr. Ateev Mehrotra, a public health researcher and professor at Brown University who studies patient uses for AI chatbots, told The Times that ”we’ve always thought that this is something coming down the pipe, but isn’t being used in big numbers right now”. Seeing the KFF poll surprised him.

It’s hardly surprising. For decades, most of us have been googling our symptoms - it’s already a deeply ingrained habit. The difference now is that we’re getting actionable results. Google, on the other hand, often led us straight into a rabbit hole of doom, where the only logical conclusion was that a slow, painful death was imminent.

Sure, chatbots don’t always get it right - but then again, neither do humans. According to BMJ Quality & Safety, ”an estimated 795 000 Americans become permanently disabled or die annually across care settings because dangerous diseases are misdiagnosed”. That is a staggering number.

Remember also, that we’ve only scratched the surface of what foundation models are capable of. Their ability as diagnosticians are likely to increase dramatically over the next several years.

Ah, but what about the doctor-patient relationship, you ask?

Well, honestly, that relationship often feels a little bit like relatives: you can’t really decide what you are going to get (if you’re not wealthy). In my own experience, I’ve encountered many wonderful, compassionate doctors, but also a fair share who seem completely indifferent.

Eric Topol suggests AI will be a boon for the doctor-patient relationship*, freeing up physicians to ”connect with patients the way we haven’t done for decades”. I’d argue that this might hold true during a relatively short transition period, but it’s unlikely to be the final outcome.

It reminds me of ATMs. When automatic tellers were first introduced, many thought they would replace human bank tellers entirely. Initially, the opposite seemed true - the rollout supported the narrative that ”automation complements humans”. But over time, the trajectory shifted toward the comprehensive automation of retail banking.

Why? Because the product isn’t the interaction with a professional. In banking, the product is money. In healthcare, the product is health.

* We might refer to this as ”AI impact compartmentalization”, whereby one recognizes the immense potential of AI but stops short of accepting its implications that feel personally threatening. I’m probably guilty of this myself now and then. But I’ve also publicly said that my job as a writer might not exist for long.

And let’s be honest: future AI systems are likely to understand and respond to your needs better than any human physician ever could. They would have the superhuman ability to remember everything about you and all the time in the world to listen. Assuming that deep human-machine communication can never surpass human-to-human interaction is a mistake.*

Already, there are people claiming to be in romantic relationships with chatbots. A study published in Nature reveal that humans often rate AI-generated poetry more highly than poetry written by humans. A significant number of us already use chatbots as therapists.

The implications of this growing interaction with AI - its opportunities, pitfalls, and complexities - are vast. But I won’t go further into that here, as it would lead me completely astray.

Instead, I’ll leave you with a quote from James Muldoon, associate professor at the University of Essex and research associate at the Oxford Internet Institute, which captures the essence of this shift: ”Ultimately, for many individuals, simulated care and understanding is real enough.”

Real enough and good enough.

* Just to clarify: I’m not saying that communication between humans and machines will always surpass the best of human-to-human interaction. My argument is that our relationships with doctors are often already distant and transactional. While it’s possible to form meaningful connections with a physician, more often than not, these interactions are purely about addressing specific needs. In an AI-driven world, those kinds of transactional relationships could become even more fragile.

Ah, but what about the physical aspects, you ask? A chatbot has no body, no face, no hands. It can’t really examine you behind a screen, can it?

With a camera looking at you it could certainly divulge more information than you’d think. With a microphone it could pick up on audio cues to certain aspects of your health. And, as we’ve seen, it’s quite capable from text alone.

But, more importantly, artificial intelligence will be embodied. This fall, I’ve spent countless hours researching humanoid robots, speaking with leaders in the field, and exploring whether this long-standing science fiction dream is on the brink of becoming reality.

The answer appears to be a resounding yes - and it’s likely to happen sooner than most people expect.

Skeptics remain, of course, but they’re probably the same voices who insist autonomous cars will never work - even as those very cars are already navigating the streets of San Francisco today.

Humanoid robots will soon cost less than a car but offer infinitely more utility. One of their many potential roles? Serving as your personal doctor.

These robots could deliver most primary and preventive care directly in your home, handling procedures that don’t require a clinical setting or large equipment. But they wouldn’t stop there - your humanoid robot could also be your therapist, personal trainer, and even your gourmet chef.

A company called Physical Intelligence aims to make humanoid robots as easy to use as chatting with an AI assistant - just say what you need, and it gets to work. Major players like Nvidia, one of the world’s most profitable companies, are fully committed to this vision, as is Tesla. Elon Musk has even predicted that by 2040, humanoid robots will outnumber humans. Figure, a startup that seemingly emerged from nowhere, has already raised $675 million at a $2.6 billion valuation. OpenAI is investing in humanoid robotics companies, while numerous Chinese startups and enterprises are also entering the race.

One of the key drivers behind this push is, in fact, healthcare - specifically, elderly care. As populations in many countries age rapidly, a significant gap is emerging: there won’t be enough people to care for the elderly. This labor shortage is already being felt in many parts of the world, making humanoid robots an increasingly urgent solution.

Sure, you can think of countless potential stumbling blocks. Will this be so heavily regulated that it becomes a non-starter? (Unlikely, given the immense demand and the overwhelmingly favorable risk-to-benefit ratio.) Who will be liable if something goes wrong? (We’ll figure it out - just like we’ve done for every other transformative technology.) And what about privacy? (Very important, though let’s be honest - many of us have paid little attention to it, when there’s convenience to be had.)

Ah, but what about hospitals, you ask? Surely human doctors will still be needed there?

While it’s true that we’ll require specialized facilities with unique equipment and conditions, there’s no fundamental reason humans would be indispensable.

I get that this idea is well outside the Overton window for many people, and that trust in technology varies widely. But attitudes can shift quickly - there was a time when people feared the black magic of electricity in their homes.

As I see it, the signals are clear: the future likely holds no human doctors. Robots and AI will perform the job far better and at a fraction of the cost.

And while robots + AI may not be the final destination - there are even more futuristic scenarios to imagine - I’ll save those for another day.

News from around the longevity and health space.

Turritopsis dohrnii, often referred to as the ”immortal jellyfish” possesses a remarkable biological trait: not only can it avoid death, but it can also reverse its aging process, effectively becoming younger when necessary.

But now scientists have discovered something both similar and different in the comb jelly Mnemiopsis leidyi, an invasive species. Instead of a dramatic reset as with T. dohrnii, this transparent sea creature takes the Benjamin Button route - gradually reversing its development while remaining the same individual.

My probably imperfect analogy is that the ability of T. dohrnii is akin to full cellular reprogramming whereas M. leidyi takes the partial reprogramming route.

Bryan Johnson has decided to stop taking rapamycin. His reasoning is definitely worth a read, but don’t overlook Krister Kauppi’s (as always) nuanced reaction to the news.

Nvidia is not only interested in humanoid robotics but also in biotech and pharma. Now the company has made their framework for drug discovery and molecule design.

Here’s a case that could be seen as an argument against everything I’ve said about the diminishing need for human doctors. But in reality, it’s probably not. While I’m not convinced that this type of pod is the ideal solution, I believe the bigger issue was simply bad timing.

What I’ve been up to lately.

Well, I haven’t really been up to this myself lately, but apparently my book Evigt Ung has been featured in the play Dödshatarna (The Death Haters) in Stockholm, which ran through October and November. The theater group also incorporated Mot Döden (The Case Against Death) by Patrick Linden and Läkemedel mot Åldrande by Linus Petersson.

Patrick Linden, of course, is my co-host on the LEVITY podcast, and Linus Petersson is the co-founder of the Swedish Longevity Cluster. The three of us are, fittingly, real-life, non-fictional death haters.

Linus went to see the play and wrote a review on LinkedIn:

”I have to admit, the sense of recognition was incredibly strong: Zoom calls with researchers from San Francisco, the world’s oldest person - the 122-year-old Frenchwoman Jeanne Calment - rolling around in a wheelchair adorned with French flags, wild future speculations about technological advancements, the long-lived Greenland shark swimming by, impassioned speeches about the imperative to cure aging, and vivid visions of cryopreservation/cryonics. Brilliant humor and satire!”

Hey, you’ve made it all the way here! Thank you so much for reading! 🫶🏼