@mathildeanita isn't just because the company is grossly overvalued in terms of current revenues? if you look at revenues or profit per employee, I suspect google is leaner.
@Noahpinion I don't think so: what's the cost of that operation? Who can afford it? Which model can they use? All OpenAI model would not do that work outside of OpenAI.
not obvious this will turn out well, it may be very rough early on.
@xlr8harder Also, what is the cost of this operation in terms of token. If they spent 15M to prove a math theorem, how much they spent and who can afford similar bills?
@giffmana First comment teaches how to remove yourself.
Second one is about getting spammed.
Third comment expresses hate...
More evidence:
https://t.co/peVL1CtRf7
People don't want to deal with AI, they should book via API. Also, I expect higher 'no shows' from AI bookings.
Today Zurich is AmericaMaXXing:
Knabenschiessen is a public holiday where 5000 teenagers go at the shooting range with assault rifles to identify the king and the queen of shooting.
@antirez A second order problem is that people will stop sharing exciting problems or progress, because the obvious outcome is being scooped by mathematicians in the AI labs with infinite compute.
@antirez The community was driven by the desire to discover and prove new math, and during such process they generated insights and understanding. Those two things were generally aligned. If AI solves top problems, talent lose interest (see recent Fields medal abandoning math).
@antirez they can definitely remove motivation and talent. if AI is solving all the hot problems, and you need immense capitals to advance, people will stop becoming mathematicians, because the main driver is the passion of solving a problem.
The long-term probability of success of any kind of detector in this space is zero, especially if you have an oracle available for testing (and even worse. for training).
Would you build an authentication service based on a selfie today?
You know how journalists try to find aligned "experts" to attack the subjects of their articles?
Wired found a professor who said he doesn't think AI detectors work.
He didn't even evaluate Pangram.
lol