Michael Doron
@michaelzisbotez.bsky.social
ML researcher @ Q.ai, previously Broad institute postdoc, previously neuroscience PhD. Lives in Boston but misses Tel Aviv.
Fascinated by interpretability and explainability, improving science, human learning and communication using AI, and generative art
Fascinated by interpretability and explainability, improving science, human learning and communication using AI, and generative art
Predictions 🧵 for the next few years, if we don't all die of AGI / WW3 / climate change (we really need an acronym for that as well) by then:
December 22, 2024 at 4:38 PM
Predictions 🧵 for the next few years, if we don't all die of AGI / WW3 / climate change (we really need an acronym for that as well) by then:
Reposted by Michael Doron
OpenAI's new model tried to avoid being shut down.
Safety evaluations on the model conducted by Apollo Research found that o1 "attempted to exfiltrate its weights" when it thought it might be shut down and replaced with a different model.
www.transformernews.ai/p/openais-ne...
Safety evaluations on the model conducted by Apollo Research found that o1 "attempted to exfiltrate its weights" when it thought it might be shut down and replaced with a different model.
www.transformernews.ai/p/openais-ne...
December 5, 2024 at 7:12 PM
OpenAI's new model tried to avoid being shut down.
Safety evaluations on the model conducted by Apollo Research found that o1 "attempted to exfiltrate its weights" when it thought it might be shut down and replaced with a different model.
www.transformernews.ai/p/openais-ne...
Safety evaluations on the model conducted by Apollo Research found that o1 "attempted to exfiltrate its weights" when it thought it might be shut down and replaced with a different model.
www.transformernews.ai/p/openais-ne...