Unplug the server and you’ll the see the amount of risk these things hold. They made an aid for software engineers (who already seemingly do nothing for work) to help them think even less on the job.
Big scary AI model we can’t let get out, HOWEVER if a company were to pay an enterprise cost for we’ll let them have it! It could bring about a dark age, which is why if you turn to page 3 in the pamphlet shareholders can see how it would affect returns! Bunch of bullshit, and anyone falling for this really just shows how out of touch they are.
X1989xx on
I think there’s been a bit of scare tactics going on with mythos that people outside cyber security, macklem likely included, are buying into. “A system that automatically identifies software vulnerabilities” does sound scary. But mythos is by far not the first ai model or even piece of software that can do that. We’ve had static analysis tools for decades that can read code and identify vulns. Current AI models are decent at it too. I haven’t used mythos yet, I’m sure it’s better than the current models. But this whole thing just screams Amodei’s classic “this is going wipe out x industry, change the world forever, it’s so powerful we don’t even know if we should’ve made it” spin cycle that blows over when people find out it’s just another incremental improvement on the existing models.
2 Comments
Unplug the server and you’ll the see the amount of risk these things hold. They made an aid for software engineers (who already seemingly do nothing for work) to help them think even less on the job.
Big scary AI model we can’t let get out, HOWEVER if a company were to pay an enterprise cost for we’ll let them have it! It could bring about a dark age, which is why if you turn to page 3 in the pamphlet shareholders can see how it would affect returns! Bunch of bullshit, and anyone falling for this really just shows how out of touch they are.
I think there’s been a bit of scare tactics going on with mythos that people outside cyber security, macklem likely included, are buying into. “A system that automatically identifies software vulnerabilities” does sound scary. But mythos is by far not the first ai model or even piece of software that can do that. We’ve had static analysis tools for decades that can read code and identify vulns. Current AI models are decent at it too. I haven’t used mythos yet, I’m sure it’s better than the current models. But this whole thing just screams Amodei’s classic “this is going wipe out x industry, change the world forever, it’s so powerful we don’t even know if we should’ve made it” spin cycle that blows over when people find out it’s just another incremental improvement on the existing models.