I was hit hard by a week of poor health which meant I missed a policy gathering in Helena today which I was really excited to attend. One of the topics was autonomy and choice in medical care and health.
And with any unexpected change of plans I try to see the upside. Because I was bed resting I was able to catch up on a keynote speech by engineer and technical communicating savant Andrei Karpathy’s talk to YC’s Startup School.
He is an excellent public speaker and has a rare gift for clarity which benefits the entire software ecosystem. And we are an industry who disproportionately see the value of sharing in real time the changes we are seeing as we build. This generation built the networks and seeded the data with our content that enabled these models.
I saw in the talk the long lineage of technical cycles, access expansion and autonomy expanding that I have been a part of since my childhood. I’ve seen a few development and deployment cycles to use the theories of Carlotta Perez
Each cycle granted more power to sharing. The excess value generation of making our tools open to more external use has proven itself. And that has generally made for cycles of innovation that are shared mid deployment by the people as it happens.
And yet we still struggle with the right way of interacting with the tools. Math is fairly abstract. Your average human doesn’t much care for conditionals. We developed mathematics over such odd timeframes that it’s somehow easier to think it’s not in tandem with a culture and a commercial environment.
Maybe some only look at the industrial or military applications for tools and they care little about how they were made. The level of autonomy and control and abstraction that is enabled by software baffles. The more accessible something becomes the more we need to think of the user of the tool. Specialists can use special tooling and need not be so accessible. When it becomes a tool for masses things change. And we are in a changing moment for software as a tool just as the world has the highest expectations for them. Because we are perhaps at the edge of the great buildout.
Karpathy said that working with LLMs can feel like using the command line. It’s an intuitive framing for many programmers. He believes we have not yet found the graphical user interface for this era even as we are perhaps building new operating systems.
The GUI or graphical user interface was a mind blowing shift for the personal computing revolution. It allowed in a world of new users including you to use the benefits of computing. Which wasn’t just calculating missile trajectory. The commercial possibilities were as endless as the personal and aesthetic.
That change in access built enormous businesses and was the stuff of nasty backstabbing in the commercialization processes and the competition was very sharp in personal computing era. My father sold software through an old school reseller called Ingram and I gather it was a pretty wild time in the eighties.
But the fresh paradigm is always beyond reach. It’s there waiting to upend your entire world.
To quote Neal Stephenson “ in the beginning there was the command” essay
We were all off the Batch, and on the Command Line, interface now—my very first shift in operating system paradigms, if only I’d known it.
We are in an operating system shift now and we don’t know what to think about it it’s structure. It’s modeled on humans so it has all the same problems we have. It has cognitive deficits just as humans do. This annoys normies who don’t understand how it’s built.
We are interfacing with a new kind of compute output and it will slowly change everything around it as the abstraction layers bring more people into the effort.
We don’t really know what it looks like at this order of magnitude but the change is here and we get to make it. It frankly seems exhausting to ponder and a much much much harder problem set for power than generalized intelligence.
How does this relate to medicine and autonomy? Well, it’s become clear that medicine will be one of the areas that benefits from new access.
I care about the way we develop tools for the entire stack of medicine from pharmaceuticals to patient data. I don’t want another era of regulatory capture. The way we build applications affects how much autonomy and freedom we can give both doctors and patients. I know don’t want to be stuck with what we’ve got. More people should benefit from the changes ahead.
