The LLMs are at the gates, solving some of the hardest problems in CS theory.
This is the text book definition of living through a paradigm shift.
0 Boosts
0 Likes
This is for the CS and programmer sorts is a moment that I think we are in the same boat as graphic artists.
Some of the highest skilled work is being threatened.
It's really not just ETH.
Its SaaS, its cybersecurity its everything software right now.
We are newly vulnerable to automation.
We have no unions practically.
And the fight for the old ways is not looking great in the face of a new pace of change.
What remains interesting at this phase is the fact so many of these "break throughs" essentially amount to bridging communication and time gaps between humans.
In legal terms, they fucking stole shit lmao and claimed it as corporate discoveries.
However from an academic lens I think what we're truly seeing is a distillation of content and thought processes invented by various humans of diff times.
The LLMs simply are allowed to ask questions professors didn't let prior PhD students work on...
@nullagent "Seeing ETH fall to the LLMs"?
AFAICT nobody has verified the "proof". It's possible it might be valid, but the default assumption should be that it's not.
It's really weird that people are a lot more eagar to believe the statistical bullshit extruder that's known for searching out exploitable bugs in theorem-provers to prove false statements with gibberish than to believe human cranks who publish this type of result.
CS Theory is basically in a situation right now where its clear. The VC class has built our version of Hubble, the Large Hadron Collider and NASA all rolled up into one interesting pile of science (yes with uncertain validation) currently being hawked by corporate shit heads as the second coming of christ himself.
The VCs, techbros and pundents aside... imo we're past my person line of where the AI/LLM technology can simply be ignored as capitalism.
This is fundamental CS theory.
About a month ago I was able to comfort myself by saying "well at best an LLM programming is approximating solutions to the halting problem just like me(probably worse)"
But after seeing ETH fall to the LLMs for a truly modest cost(along with hundreds of proofs and thousands of vulns)... the writing is on the walls.
A LOT of what my generation learned about CS Theory is proven false with this one.
Not quire at bedrock yet, but shift confirmed.
https://en.wikipedia.org/wiki/Exponential_time_hypothesis