"What we saw made us realize: the next time we train a really big model, we should expect it to have these capabilities sort of inherent to it, rather than ones that we need to elicit."
"This is not a special model. There will be other systems just like this in a few months from other companies, and then a year to a year and a half later, there'll be open-weight models from China that have these capabilities."
"Project Glasswing, which grants certain companies selective access to Mythos for further safety and capability testing, will help reveal more about the AI's capabilities."
Anthropic's new AI model, Mythos, was developed with a focus on cybersecurity capabilities. Co-founder Jack Clark noted that Mythos has already surpassed benchmarks in detecting vulnerabilities. He anticipates that other companies will soon release similar models, emphasizing the need for preparedness for more advanced AI systems. Project Glasswing allows selective access to Mythos for testing. Clark's remarks coincide with ongoing legal issues between Anthropic and the Pentagon regarding the use of its models for surveillance and autonomous weapons.
Read at Nextgov.com
Unable to calculate read time
Collection
[
|
...
]