Today I sit with the same amazement, taken aback again, appreciating how ridiculously awesome this is. Congratulations to the winners and everyone involved!
The fact I have a computer writing flowery alt text descriptions of my photos with unnerving accuracy is something I would not have predicted for another 20 years. But, here we are...
State-of-the-art machine learning architectures aren't actually that complex. Diffusion models and transformers can be explained to a bright high schooler. I'm sure Archimedes and Euclid would have no problem understanding them.
What they might have a problem understanding (or even imagining) is the mind-boggling amount of computation required to make those systems do anything useful. Getting Llama to produce a single token of text takes more calculations than all of humanity did by hand during all of Classical Antiquity.
Imagine all the stuff... transistors, Turing/Von Neumann machines, lithography, theoretical computer science, OS and compilers, the Internet... and lastly there's modern day machine learning that builds on top of all the above.
The base level stuff isn't exactly protons and electrons, but given the nanometer scale of our chips, it's not that far away from the truth, and we (humanity) has somehow built amazing stuff on top of that.
Smart "intellectual" people would certainly be willing to challenge basically everything they assume about nature, but I don't think your run of the mill farmer would be able to do that.
It would be like explaining a baking recipe just talking about wheat and flour and heat. The point is that from first principles, ML is a huge jump. From first principles, baking is not.
https://github.com/mlang/tracktales
The fact I have a computer generate spoken narration for my MPD playlist with descriptions of album art included just blows my mind. 2023 was indeed a fucking milestone.
My grandma was blind, and I just spent 6 months looking after a guy who was blind (but has now had surgery and beat me in the eye test at the doctor's!), so I think about blindness a lot when I design.
And this isn't only about the monetary value itself, but also the fact that a large cash prize attached to a challenge boosts the prestige of finding a solution. Nobel Prizes come with about a million bucks on top of them, after all.
I'm quite confident that if someone offered $100 million for deciphering the Voynich manuscript or Linear A, we'd have a solution within 3 years.
Make it $8m or $12m and FAANG employees can actually start to justify working on it seriously from a money perspective
And yes, $1 million is very substantial for an individual. And the cool thing about offering it as a prize (from the point of view of the organizers, that is) is they only have to pay one person or team, although potentially thousands ultimately contribute to the solution, directly or indirectly.
But it's not clear what current tech could help with. Machine learning can't be applied to something you don't have training data for.