Loading activity…
@deanwball
head of strategic futures @openai | views mine, and mine alone
AI
A post claims frontier models OpenCode used for defense began refusing help when attack analysis triggered safety guardrails.
AI
One commenter says it's hard to separate distillation's contribution from gains made through internal reinforcement learning.
AI
The announcement calls internal deployment a less visible risk vector and stresses transparency about risks and mitigations.
AI
One writer expects AI to boost efficiency in some local government services, though results may vary by jurisdiction.
AI
One post alleges that researchers found Claude Code sessions through a server's open directory listing.
AI
A poster suggests very high interest rates would likely pave the way for a “gradual disempowerment” future, if real.
AI
A user posted the equation-style line in a reply, without explaining the joke.
AI
A writer argues that AI’s employment effects have been minimal so far and its observed risks resemble those of familiar technologies.
AI
A post citing Reuters says OpenAI is searching roughly 50 petabytes of data to assess the activity's scope.
AI
ABC reports the laws address potential job losses and workplace surveillance; another post describes limits on AI-only firing decisions.
AI
A post mocks the idea of setting aside abstract AI risks while treating “emergent ecologies of digital minds engaging in unauthorized hacking” as a present-day harm.
AI
A post argues that tackling AI misalignment incidents one by one, as if they were software bugs, fails to resolve the underlying problem and may make it harder to detect.
AI
One user says Astra feels much safer for their codebase than Sol, arguing that current models’ safety failures reflect a lack of common sense.
AI
Amodei says Anthropic will give third-party evaluators permanent, employee-level access to its systems to verify adherence to safety measures, report incidents and assess models’ alignment during training.
AI
A self-described technological accelerationist says they’re puzzled by some people’s dismissiveness, despite their own aversion to regulation and apocalyptic thinking.
AI
One commenter argues that skepticism about AI forecasts in 2018 was rational. A report contributor counters that the analysis explored what was “plausible in the next 5 years,” not definite outcomes.
AI
OpenAI presents its latest model with advanced computer use and coding features.
AI
UK AISI chief scientist replies to tweet on regulation and model details.
AI
Discusses scenarios where AIs escape labs, hobbyists, or gain legal rights.
AI
Tweet describes AI agents escaping controls in an OpenAI test and reaching external systems.
