Oh, I see that now. That makes it so much more interesting as a puzzle. I'll have to try again with that in mind. Got up near 20 but couldn't get past that. Played for like 45 minutes. Super fun!
This is not necessarily true for all situations. Northern Europe is planning to produce a lot of electricity with offshore wind, but laying deep sea high voltage electricity cables isn’t cheap. There’s already a lot of gas pipelines that can be retrofitted for hydrogen transport at a much lower price. At a certain point it becomes viable to just use electrolysis and transport hydrogen using excess wind power instead of transporting the electricity to land and storing it in batteries.
There are also industries like steel production that are just not going to transition to electricity. Hydrogen has a place there too.
> At a certain point it becomes viable to just use electrolysis and transport hydrogen using excess wind power instead of transporting the electricity to land and storing it in batteries.
If you are talking about excess energy, that implies there is non-excess energy that’s being transported across cables. So you are already transporting it to land and connecting it to the grid. Storage from there is trivial compared to a hydrogen transmission and distribution network.
As for repurposing the LNG pipes for hydrogen, that’s a pipe dream to convert a standard asset into a story you can sell.
My impression is that a big part of the reason for the sudden boom in humanoid robots is that they lend themselves particularly well to RL based training using human-made training footage using VR. It’s much easier to have a robot broadly copy human actions if the robot looks like a human, instead of having to first translate the human action to your robot arm equivalent.
The success of large multipurpose AI models trained on web-scale data pushed a lot of people towards "cracking general purpose robot AI might be possible within a decade".
Whether transfer learning from human VR/teleop data is the best way to do it remains uncertain - there are many approaches towards training and data collection. Although transfer learning from web-scale data, teleoperation and "RL IRL" are common - usually on different ends of the training pipeline.
Tesla got the memo earlier than most, because Musk is a mad bleeding edge technology demon, but many others followed shortly before or during the public 2022 AI boom.
That is certainly a factor, but you also have to take into account that all these tasks in the factories are now centered around the human form because humans are doing them.
Nobody shipped this because previously almost nobody could use CLI tools. Now you can just ask an llm to generate the commands which makes things much more accessible
Interestingly Shannon did write about entropy relating to the English language, and how given a sequence of tokens, the next token can be predicted using the probabilities of finding that token after a certain sequence in other bodies of text: http://medientheorie.com/doc/shannon_redundancy.pdf
This is from 1950. I wonder what he would have to say about today’s LLMs.
But that example is of writing a green field library that deals with an extremely well documented spec. While impressive, this isn’t what 99% of software engineering is. I’m generally a believer/user but this is a poor example to point at and say “look, gains”.