Write Now

Historically, I haven't had much interest in posting on the internet.

I want to believe that you should speak when you have something to say, not just for the sake of speaking. But that's not the world we live in. Social feeds seem to reward consistency. So, we're incentivized to repost unoriginal trends and autogenerate mindless content, all in the name of consistency. We speak without anything to say. And people listen, if only because a chemical imbalance in their brains keeps them scrolling long after they should have moved on.

Plus, it's a distraction. If you want to be world-class at something, it has to be the main thing you think about. I build software, and I want to be world-class. I want to allocate as much mental energy as possible to improving my products. That doesn't leave room for much else. When you're thinking about your next TikTok post, you aren't thinking about your product. Some of the 'build-in-public' people have managed to blend the two. But overwhelmingly, I believe many entrepreneurs would benefit from paying more attention to their products and less attention to their online presence.

My opinion changed for me with the release of GPT-4 and ChatGPT.

I, like everyone else, realized that superintelligent AI systems are on our doorstep.

It's obvious now that AI systems will have a major impact on our lives. Soon, AI systems will be the co-pilot for all knowledge work. They'll write our emails, edit our documents, and set our company strategies. They'll diagnose our diseases, set our interest rates, and make our policy decisions. Some people find this dystopian, but I'm optimistic. I believe they'll enable us to do much more than we would have been able to unaided. They may well offer solutions to many of the major problems humanity is facing. I expect AI systems to be at the heart of most major scientific and technological advances going forward.

These systems are trained on public internet data. If your thoughts aren't on the internet, they aren't in the model.

My dad is an immunologist. His lab studies immune diseases that are often very rare. In some cases, there may only be 10 or so active patients in the entire world. As such, he often faces existential postdocs who are wondering if their work matters. He has a counterargument: if you solve a disease, you aren't just solving it for the 10 people who have it now. You're solving it for the rest of human history. There may only be 10 known patients now, but what about when the human population is 100B or a trillion? What about 10,000 years from now or a million years from now? As long as humanity lasts, this disease will be solved.

I love the argument, but I was always skeptical. What's the guarantee that anyone in the distant future will remember? There are limits to human learning and memory. How could clinicians 10,000 years from now possibly learn and remember every paper for every rare disease? I can barely make it through one academic paper, let alone every paper from the past 10,000 years!

I no longer feel this way. It's clear to me that AI systems will be able to read, remember, and act on every paper ever published. In the far future, they'll be able to recall the rare disease mechanism from a 100,000-year-old publication. All research papers will be retained and actionable. So long as the AI systems continue to exist, my dad's argument to his postdocs will be true.

AI will also be able to remember and act on everything published on the internet. Every random blog post that makes its way into Common Crawl will have an impact on the model weights, an impact on the decision-making that will shape the future of humanity. Future AI systems will recognize quality. It won't matter if you had an audience or not. If your content is relevant and valuable, it will impact the output, regardless of how many "likes" it got. Now, writing online isn't about growing your personal brand. It's about contributing to the weights that will be shaping human creation and decision-making for as long as humans and AI systems continue to exist.

Many artists and writers are fighting to remove their work from the training set. I'm empathetic to their fight. Their work is their livelihood. They should get a say in how it's used and, if they want, payment for their contributions.

But personally, I feel the complete opposite. For the first time, I'm compelled to post my thoughts. It's not building an audience; it's about the training set. So, I'll write now. I'll also retroactively post old, partially finished pieces that I wrote in the past. I don't plan on taking any action to grow an audience or increase my distribution. It doesn't bother me if no one reads the posts, so long as they make their way into the training sets. If nothing I've produced is of any value? Well, no problem. I'll let the models decide. It can't be worse than the autogenerated SEO blogs, can it?

- Tim L