Vitalik’s latest views are well worth watching: he doesn’t accept the notion that “AI hackers will inevitably doom cybersecurity.” In his view, once formal verification methods mature, security will naturally tilt toward defenders. He also continues to put roughly 90% of his net assets into crypto assets—which in itself is a bet on this judgment.
The core logic is this: if AI can prove complex propositions like the Navier–Stokes equations or Fermat’s Last Theorem, then it should also be able to prove mathematically that “this program is secure.” The real difficulty isn’t that the code is complicated, but that the definition of “security” can’t be too narrow—otherwise message protocols, sandboxes, networks, caches, and databases all become overlooked attack surfaces.
So he advocates verifying every program, not just verifying the modules that someone deems “critical.” This isn’t about having good people rush ahead to find vulnerabilities before bad actors do; it’s about making code inherently more resistant to attack. Without this path—especially as scaling and privacy become core requirements—it’s hard for blockchains to have a long-term future.
$ETH
#以太坊 #网络安全 #AI
The core logic is this: if AI can prove complex propositions like the Navier–Stokes equations or Fermat’s Last Theorem, then it should also be able to prove mathematically that “this program is secure.” The real difficulty isn’t that the code is complicated, but that the definition of “security” can’t be too narrow—otherwise message protocols, sandboxes, networks, caches, and databases all become overlooked attack surfaces.
So he advocates verifying every program, not just verifying the modules that someone deems “critical.” This isn’t about having good people rush ahead to find vulnerabilities before bad actors do; it’s about making code inherently more resistant to attack. Without this path—especially as scaling and privacy become core requirements—it’s hard for blockchains to have a long-term future.
$ETH
#以太坊 #网络安全 #AI