Who is the true inventor? The one who came up with a great idea, or the one who managed to secure funding for it? On one hand, ideas are worth nothing — you could shovel out a whole wagon of them, and many would closely resemble what others are thinking up and building at the exact same time. What really matters is the team.

Of course, you can go to the other extreme — build a great team working on a “so-so” idea that isn’t innovative at all. But that’s basically like launching a new brand of soap in new packaging — it’s all been done before. It might even be successful, but it’s not innovative.

Truly breakthrough technologies are rare.

There’s this idea that there are no innovations other than combinatorial ones — they come from mixing existing elements in new ways. Humanity has created so much already that we could keep generating combinations for thousands of years. There’s a game that illustrates this well — I always use it as an example. It’s called Alchemist (Doodle God). You start with a few basic elements — earth, water, fire. By combining them, you get new ones. Then you combine those with others, and so on. Eventually, the number of combinations becomes massive — many are non-obvious or even funny.

And the thing is, humanity has so many of these building blocks now that even if every person on Earth created a new combination, there would still be countless more left to discover.

Innovation is about finding ideas that are as far apart as possible, and combining them in ways that produce surprisingly useful meaning. Ideas that are already semantically close — people have already stumbled upon them for themselves, so they’re no longer interesting.

I’ve got this sense that humanity has hit a wall when it comes to fundamental innovation, and to regain the speed we once had, we need help.

On one hand, the number of elements we can combine for innovation is off the charts. But on the other hand, I think science is in a bit of a crisis — at the cutting edge, the “nutrient medium” for invention is drying up. I’ll write a separate post about science, but for now I’ll just say: in my opinion, there are a lot of research papers that are physically hard to access, and many of them are unverifiable. That’s a strong claim, so I’ll back it up with some numbers in that future post.

And because what’s on the surface these days is mostly hype, and it’s hard for the average person or inventor to find something truly new, access to data and, just as importantly, automation, could make a big difference.

New ideas and hypotheses can definitely be generated by ChatGPT or similar language models. The transformer model is quite good at creating something out of already existing material — whether it’s text, poetry, or images. But validating ideas is a different story — it can’t yet be done outside of the real world. Though if we had some kind of simulated environment, partial automation might be possible. Still, its connection to reality would be pretty limited.

Automating hypothesis validation is a tough problem — but it always has been. This is where public or governmental institutions providing the needed infrastructure would play a key role. The country that can make this happen will reap major rewards.

——Translated with ChatGPT from original version—–