Oh did they call all these traders after the market closed and ask them? Are the traders in the room with you right now?
0 Boosts
0 Likes
@fullrange You can't abandon something that wasn't possible in the first place. Massive electrician shortage, even bigger shortage of electricity let alone other materials
I would be in if it was for something useful, but first of all it is a lousy inefficient way of programming. Remember that it took 330 Bytes of RAM to fly to moon…
@fullrange I don't know that's a fair comparison. I've seen things built in 24 hours that couldn't be built in over 10 years.
Now if we're actually talking about the amount of energy (electricity) being used:
- electricity is only going to get cheaper. We're just behind on infrastructure upgrades expansion in the USA vs China
- as compute gets cheaper we always get wasteful with it. Same as any resource really.
- if you want to compare compute to human developers, do the math on the raw energy inputs used to feed them. Nothing else, just food during the extended time it takes to build something comparable. Turns out compute is cheaper. By far.
There's a lot to criticize about LLMs, but you have to completely change your workflow and habits to get the best results. And that's hard, I struggle with it too. My friends who don't struggle make me feel so embarrassed with how good their results are. With full test coverage. With full types coverage. Without spaghetti code and duplicated functions/logic.
It's really possible. But people set in their ways and who like their tooling and processes will struggle.
I don't endorse the wasteful giant datacenters and lying to shareholders, etc etc.
Now if we're actually talking about the amount of energy (electricity) being used:
- electricity is only going to get cheaper. We're just behind on infrastructure upgrades expansion in the USA vs China
- as compute gets cheaper we always get wasteful with it. Same as any resource really.
- if you want to compare compute to human developers, do the math on the raw energy inputs used to feed them. Nothing else, just food during the extended time it takes to build something comparable. Turns out compute is cheaper. By far.
There's a lot to criticize about LLMs, but you have to completely change your workflow and habits to get the best results. And that's hard, I struggle with it too. My friends who don't struggle make me feel so embarrassed with how good their results are. With full test coverage. With full types coverage. Without spaghetti code and duplicated functions/logic.
It's really possible. But people set in their ways and who like their tooling and processes will struggle.
I don't endorse the wasteful giant datacenters and lying to shareholders, etc etc.
Thank you for that long answer. Of course it is not a fair comparison. However 2nm structures are just around the corner that again doubles the parts per area count and rumors say 1nm is pioneered in the lab…
Finally I want to see computers to lift the workload off humans to free them into art, but now hordes of programmers are working on computers painting and composing?? Honestly, I don’t get it.
What about abandoning DC buildout?