Nearly 20 years ago I met a doctor who had built a diagnosis system for brain related conditions. When he showed it to me I was completely blown away. You would enter some symptoms and the system would display a list of possible diagnoses. If you then entered test results it narrowed down the diagnoses – in fact it was able to tell you which test would most help you at this differential step! The whole thing ran on a laptop. So what was the catch? Well it turned out that it had taken him years to compile the underlying data and even within the domain of brain related diseases there were still parts missing. Extending this to all conditions would seem to take forever.

Three years ago it occurred to me that a better way to do this would be to get many people to contribute structured knowledge and thus massively speed up the process. So I set out and wrote a blog post with the somewhat wonky title of “The Next Frontier for Peer Production: Open Machine Learning Services,” with peer production meaning many contributors and open referring to the idea that the knowledge would not be locked up but accessible. The blog post resulted in someone emailing me and introducing me to the team from Human Dx. It turned out that this was exactly what they were working on for medicine and we wound up investing later that year.

There were many difficult problems to solve around how to structure the data, how to get contributors excited about participating and how to create value for everyone in the process. I am thrilled with the progress that Human Dx has made on all of these. And as of a few days ago they have started to publish out the early results of the amazing human - machine collaboration system they are building. Right from the Human Dx homepage you can now navigate this beautiful diagram:

When you click through on a condition, such as Myocardial Infarction, you can see the detailed knowledge that is accumulating in the system including related diagnoses

and who has contributed knowledge

On the homepage and on every detail page there is also a search box you can use to find diganoses and symptoms.

I am extremely excited about these visualizations as they make progress along Human Dx’s mission and approach much more tangible. If you are inspired by this also, you can either apply to join the team, or start contributing cases.

Nearly 20 years ago I met a doctor who had built a diagnosis system for brain related conditions. When he showed it to me I was completely blown away. You would enter some symptoms and the system would display a list of possible diagnoses. If you then entered test results it narrowed down the diagnoses – in fact it was able to tell you which test would most help you at this differential step! The whole thing ran on a laptop. So what was the catch? Well it turned out that it had taken him years to compile the underlying data and even within the domain of brain related diseases there were still parts missing. Extending this to all conditions would seem to take forever.

Three years ago it occurred to me that a better way to do this would be to get many people to contribute structured knowledge and thus massively speed up the process. So I set out and wrote a blog post with the somewhat wonky title of “The Next Frontier for Peer Production: Open Machine Learning Services,” with peer production meaning many contributors and open referring to the idea that the knowledge would not be locked up but accessible. The blog post resulted in someone emailing me and introducing me to the team from Human Dx. It turned out that this was exactly what they were working on for medicine and we wound up investing later that year.

There were many difficult problems to solve around how to structure the data, how to get contributors excited about participating and how to create value for everyone in the process. I am thrilled with the progress that Human Dx has made on all of these. And as of a few days ago they have started to publish out the early results of the amazing human - machine collaboration system they are building. Right from the Human Dx homepage you can now navigate this beautiful diagram:

When you click through on a condition, such as Myocardial Infarction, you can see the detailed knowledge that is accumulating in the system including related diagnoses

and who has contributed knowledge

On the homepage and on every detail page there is also a search box you can use to find diganoses and symptoms.

I am extremely excited about these visualizations as they make progress along Human Dx’s mission and approach much more tangible. If you are inspired by this also, you can either apply to join the team, or start contributing cases.