Staying Relevant in the Age of AI

I just got back from an exciting work trip to Michigan. We met with many clinic and hospital directors, and had an overwhelmingly positive reception to our startup’s services.

This has put a bit of extra work on my plate for the holiday week. I’m squeezing it in instead of gym time. Something has to give when you run your a business. Time for family, health, business, or sleep. Health is always the first to go, but I am learning tricks to minimize its impact by making movement and activity a part of family and business time.

I’m going to keep this email and next short because of the extra work, and because you also have better things to do during the Holidays. Like being with family or being a hero on-call so that other’s can be with their families.

I was shocked to hear OpenAI announce a new model yesterday. I shelved an email I was writing about how to create innovation in the clinical setting to cover my thoughts on this dizzying news.

On the final day of OpenAI’s “12 Days of OpenAI”, they unveiled their latest breakthrough: the o3 model. This announcement came just two weeks after releasing o1 (o2 was skipped due to trademark issues). With o3, we’ve reached a milestone where machines can match or exceed human intelligence on all fronts, including reasoning tasks essential to medicine.

For perspective, o3 achieved a record-breaking 87.5% on the ARC-AGI benchmark, a test designed to measure complex reasoning at a human level. Machines now rival us in fields that once felt untouchable. Soon, independently diagnosing, managing, and treating patients may become trivial for AI.

This leap forward has also sparked discussions about the limits of AI development. OpenAI’s co-founder, Ilya Sutskever, has raised critical points about the challenges ahead. Notably, he left OpenAI this past year because of concerns about the alignment of the company with its goals on keeping AI safe.

Ilya’s thoughts on the challenges:

• Data Limitations: Sutskever notes we’ve reached “peak data,” where the internet no longer provides the vast reserves needed for scaling AI.
This is good news for doctors because the human intuition of picking up a zebra is harder for a machine that hasn’t trained on it, in theory. In practice, you can solve for that by giving the machine access to the same medical literature we study to infer on those cases.

• New Training Paradigms: To overcome this, he envisions using synthetic data and developing “agentic AI”, autonomous systems capable of reasoning and human-like behavior.

• Unpredictable Systems: As AI evolves to reason, it will become less predictable, introducing new complexities and risks as it tries to reason through something new!

It’s startling to consider. Just decades ago, the idea that machines could play chess better than us, calculate faster than us, or communicate globally in seconds felt like science fiction.

Before that, there was a time where humans had to reconcile the idea that technology could help them travel faster than on foot or by horse. But then we built machines that let us touch the moon.

General intelligence was humanity’s last frontier. It was that intelligence that gave humans an advantage as a species to bend the world, all its resources, and all its creatures to our will. And now, we’re outsourcing that intelligence.

This isn’t a judgment on whether this progress is good or bad. It’s a reality check. The technology is here. Others will be using it. Competing with your raw intelligence against someone leveraging AI is like running a footrace against someone driving a car. You won’t win.

Ability to learn fast and adaptability are going to be the two most important skills in this upcoming rollercoaster. There's already whiplash from the speed of progress and we haven't seen any divergence yet. 

We’re entering a time of divergence. Hundreds of millions around the world will use tools like o3 at full speed, each applying AI in unique directions. It’s hard to imagine what the speed of progress will look like when more and more people build on top of the infrastructure being assembled by OpenAI, Google, Anthropic, etc.

For those interested in engaging with this transformation, OpenAI is offering a unique opportunity: you can apply to test the safety and capabilities of the upcoming o3 model before anyone else.

Apply here! This is an excellent chance for med students, academics, or any professional curious about the intersection of AI and medicine.

If you apply and get accepted, please shoot me an email. I’d love to hear about your experiences and what you discover.

Ask/Opportunities

I’m looking to hire a Psychiatrist licensed in Michigan as a remote consultant for our new clients. If you know anyone looking for flexible, part-time work, please introduce us!

Happy Holidays Everyone,

Mohammed