Josh

trillions

an easy signal that what you're seeing is pure doomerism is that people assume AI is this thing that is only capable of destruction but somehow never any solution it can simultaneously both break anything but apparently fix nothing fascinating

was surprised to see so much panic today, likely made worse by other things like rates and oil that distorted the actual substance in the AI/math talks tl;dr — you're all gonna be fine. the leading chains of today have ample means to be able to respond to something like this in the unlikely scenario of it happening longer: i) is AI getting better at math? yes does that imply that it will suddenly shatter foundational cryptography concepts? no "computer smart. therefore the past work of all mathematicians is probably wrong." is not a logical chain with basis sure, AI will find flaws in past human work in many areas -- in the same way humans themselves find flaws from other humans in many areas. this is how science works if we find issues, we discuss them, fix them, and evolve it takes large leaps of complete doomerism to assume that all teams, individuals, and ecosystems will fail at literally everything ranging from planning to prediction to detection to mitigation on all time horizons simultaneously further, in these scenarios, AI is exclusively seen as a tool *only* for max destruction but never as a thing capable of a solution so, it can apparently break everything but not fix anything. i hope you can see the bias here for one, we already know this to be false given recent scaling of formal verification via ai it is of course easier for AI to make progress in fields and topics that are relatively less looked at vs something like foundational cryptography which all intelligence agencies, nation states, and high-level mathematicians are heavily incentivized to break, all the time ii) the most unhinged doomsday scenarios assume an instantaneous attack with no lead time by a random person outside of law and jurisdiction, which I would posit is simply not possible even ignoring the fact that the labs would get there first (and hence the government) it's not as if one can just break all cryptography on all chains via their beefed up cloud hosts over a few hours or days or even months that not only assumes an algorithmic advancement with unfathomable magnitudes of improvement but also completely ignores the gargantuan operational complexity and opportunity cost of doing a global heist (vs the trillion other more important things you can do with an ECC break) it would be vastly impressive if a slight speedup could be found, but the speedups required to fulfill the doomsday fantasies of people on here are not grounded in anything iii) importantly, assume I am completely wrong for all of the above well even in that case, upgrading the systems in a world where ECC breaks is entirely possible and feasible (and has been worked on for decades) in fact, there are many ways of doing it that we know of depending on the existing constraints of the system! imo, the biggest "risk" in all this is that a frontier lab publishes something about a slight speedup in the future, and then everyone extrapolates that to default to panic which then leads to a selloff like today I don't think this is all bad because it never really hurts to think about and plan for these things on the protocol level, but the "retail" panic was certainly a huge self-own by the people pushing this stuff doomerism is not going to win in crypto