TechTakes
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
ChatGPT's new search feature hasn't even launched and already its shitting the bed
The sneers are writing themselves
Just found out from a screenshot in a tweet that Marc Aandreesen considers Nick Land to be some kind of patron saint of "techno-optimism". Setting aside Land's ugly views about everything... I didn't think optimism was what he was known for. More like grasping, desperate, disgusted embrace of the onward march of capital.
Nick Land (...) “techno-optimism”
His thing is literally called DARK ENLIGHTENMENT, fucking Final Fantasy villain level of grandiose menacing naming, how on earth would that be "optimism".
Marcee do you even like know who your idols are?
after all this time, TIL that Roko pronounces his name "Rocko" https://www.youtube.com/watch?v=VIwJDnej7pg
I'm not a native english speaker, how else was he supposed to pronounce it?
AP takes down CF Vance apologia, not asserting things that they cannot evidence (that Vance did not have sexual relations with that couch)
before: https://archive.is/fXiMc after: https://archive.is/j3aot
This is the journalistic integrity we expect of AP News
Regarding that claimed breakthrough about AI winning the International Mathematical Olympiad: a reminder that a proof which hangs together logically is not necessarily a proof that makes sense.
Those formalized proofs are so incredibly ugly, it's amazing. Of course it doesn't much of a sensible indentation, but then there are single proof steps where I have no idea what it's even doing. [...] And then there are nonsense mathematical steps. The solution of problem 2 starts with induction, before introducing any variables. It applies induction to the number 12. And it write 12 as
(10)+2. Then it proceeds to do the whole proof in the base case of the induction, and notices that the induction step is trivial, since the goal is the same as the induction hypothesis (but instead of theassumptiontactic it usescongr 26).
Also, choice sneer in the comments:
AlphaProof is more "AlphaZero doing self play against Lean" and less "Gemeni reading human proofs"
Not a sneer, but a mildly interesting open letter:
A specification for those who want content searchable on search engines, but not used for machine learning.
The basic idea is effectively an extension of robots.txt which attempts to resolve the issue by providing a means to politely ask AI crawlers not to scrape your stuff.
Personally, I don't expect this to ever get off the ground or see much usage - this proposal is entirely reliant on trusting that AI bros/companies will respect people's wishes and avoid scraping shit without people's permission.
Between OpenAI publicly wiping their asses with robots.txt, Perplexity lying about user agents to steal people's work, and the fact a lot of people's work got stolen before anyone even had the opportunity to say "no", the trust necessary for this shit to see any public use is entirely gone, and likely has been for a while.