the actually sound rationale behind deacc (or, forms of deacc)
I am okay with Kimi 3.1 deliberately leaving out top-end offensive cyber capabilities from their training data.
Discriminating against use cases is MUCH less bad than discriminating against users. And the world does need all the time it can get to prepare, harden, check old code, (properly) formally verify new code, etc before the full weight of very powerful cyber attacks is upon us.
And if this first big wave of AI risk is not handled well, then we may well get a dystopia of highly locked down models, KYC gating on everything, highly intrusive surveillance on everything, a society where you may be a second or third or fourth-class citizen in terms of model access based on your credentials and nationality.
TIL that the Ledger implant was pretty much public knowledge before it got exploited
youtube.com/watch?v=WNpW52x-…
How come everybody is afraid of the AI security doom but frowns when you say you audit your stuff with Claude
but... but... it's more trendy to doom post?>>
hit the nail right on the head. The heavy lifting on any product starts after most of the coding is done. Always been that way
and it's not about marketing, it's about refining, understanding user needs, knowing what to build next, etc.
not refuting that cybersecurity is gonna be in the spotlight, but just as easily you can tell Claude to find vulnerabilities, you can tell it to patch vulnerabilities
And attack is always a bit harder than defense
from now on, whoever is an easy target is most likely lazy and not forward thinking
The remainder of the year is going to be all about cyber security
It's going to be really bad. Everyone will get impacted. Data will get leaked, money will be lost. Planes may get grounded, even.
Politicians will talk about it, make it their #1 issue.
But they can't really do anything about it
And then, like nothing, some time next year it'll all suddenly go away. Issues will stop appearing and it will be the most secure we have ever been
this is why people are in shambles
if you consider yourself a good programmer, you almost certainly enjoy AI, because you keep the creative/big brain responsibilities and tasks but lose the puzzle brain teasers that don't really matter
the big brain architecture/design stuff is almost always what makes a 100xer a 100xer. never writing code (although 100xers excel at that too)
dying on this hill
the ECDSA discovery is a clear sign we're living in the golden age of humanity, where good questions are rewarded almost for free
and a perfect reflection of how in shambles everybody is is that so many still misunderstand the discovery when all it takes is one prompt to understand it
Ivo 7702/acc retweeted
the first discovery from anthropic regarding elliptic curve cryptography shows that the scheme is SAFER than many people believed
bunker mode retards in disbelief
CLAUDE FOUND THE SEEDS
anthropic.com/research/ecdsa…
Ivo 7702/acc retweeted
bro this is literally good news that increases the confidence in the curves. a 48-bit prng is not a cryptographic hash. nobody broke any cryptography here, claude just reverse engineered how public parameters were generated 30 years ago. the original engineers didn't make any notes. the discovery strongly suggests the seeds don't introduce any weaknesses or backdoors, like many believed. the inputs turned out completely plausible and benign.
Ivo 7702/acc retweeted
Unreal. Satoshi prob chose secp256k1 instead of NSA's P-256 because of this mystery - its a lot easier to just trust the number 7 (coefficient in secp256k1) than the mysterious possibly backdoored monster coefficient in P-256
now we know where the monster came from! its benign!
Claude figured out the origin of the NIST curve parameters. Turns out to have been timestamps!
anthropic.com/research/ecdsa…
For consumer products, convincing the users that it’s useful/valuable was almost always harder than writing the code
For pet projects, this wasn’t the case, just the mere existence of code of something technologically impressive was enough
But now that anyone can conjure anything into existence, pet projects are treated like consumer projects. A repo isn’t impressive on its own