"Did you realize that we live in a reality where SciHub is illegal, and OpenAI is not?"
"Did you realize that we live in a reality where SciHub is illegal, and OpenAI is not?"
Did you realize that we live in a reality where SciHub is illegal, and OpenAI is not?
You're viewing a single thread.
If this ends with LLMs getting shutdown to some degree, I wonder if it's going to result in something like a Pirate Bai.
It'll result in the industry moving to nations with more permissive scraping laws (like Japan) or less respect for Western copyright (Russia, China).
Yeah, realistically what will happen is china will get far ahead in natural language computing which will benefit it's economy, everyone who demanded chat GPT be stopped because they're scared of change will demand the government do something to catch up and they'll write exemptions into the law.
More likely they'll realise this is the obvious way things will go and the only legislation will make it harder for open source and community run ai, but hopefully not significantly.
The next round of ai gen will start reaching consumer space soon. CAD, especially for electronics and structural design (E.g. creating the right amount of supports to hold a given load). They'll scare a few corporations into pushing for legislation but I think the utility will be far clearer.
Are there any historical examples of a technology as valuable as LLMs being effectively shut down?
Nuclear energy.
Sort of joking.
Great question, but unfortunately it's out of both my knowledge and will-to-research base.
Stem cell research
Nice example. I think embryotic is the asterisk to it though.
Not to be confused with, "Pirate Bae", the pirate dating site for those endowed with abundant doubloons.
Doubt it. GenAI requires a shitton of resources, both in storage and for processing. Training a GenAI requires clusters upon clusters of NPUs and/or GPUs, even more than crypto miners and 3D renderers. The full storage requirements are proportional to the amount of training data you give, so expect them at least to be dozens of gigabytes long.
I doubt AI companies do it "for science" (yeah, right) so if they're shit down by a court of law they'll just shut the thing down. They can upload the code somewhere, but without training data their engine is useless.