Found 4 bookmarks
Custom sorting
(1) † lucia scarlet 🩸 on X: ""ok but how do we know Private Cloud Compute is actually private. do we just take Apple's word for it" Wrong 1. they will regularly release images of the entire custom OS powering these servers for public inspection, along with a virtualisation environment, so security https://t.co/NUvmLYYAdh" / X
(1) † lucia scarlet 🩸 on X: ""ok but how do we know Private Cloud Compute is actually private. do we just take Apple's word for it" Wrong 1. they will regularly release images of the entire custom OS powering these servers for public inspection, along with a virtualisation environment, so security https://t.co/NUvmLYYAdh" / X
1. they will regularly release images of the entire custom OS powering these servers for public inspection, along with a virtualisation environment, so security… — † lucia scarlet 🩸 (@luciascarlet)
#reference#chips#privacy#security#open source#research#academic#companies/apple#apple-silicon#ai#servers#discourse/counterclaims
·x.com·
(1) † lucia scarlet 🩸 on X: ""ok but how do we know Private Cloud Compute is actually private. do we just take Apple's word for it" Wrong 1. they will regularly release images of the entire custom OS powering these servers for public inspection, along with a virtualisation environment, so security https://t.co/NUvmLYYAdh" / X
swyx 🔜 ai.engineer on X: "IMO @AnthropicAI is very close to making a breakthrough in productizable interpretability. For ~4 years all we've had to really control LLMs is temperature/top_p and logit bias. We recently got `seed` and constrained structured output, with `interactive=false` on the way. But https://t.co/Z8C28evxgN" / X
swyx 🔜 ai.engineer on X: "IMO @AnthropicAI is very close to making a breakthrough in productizable interpretability. For ~4 years all we've had to really control LLMs is temperature/top_p and logit bias. We recently got `seed` and constrained structured output, with `interactive=false` on the way. But https://t.co/Z8C28evxgN" / X
For ~4 years all we've had to really control LLMs is temperature/top_p and logit bias. We recently got `seed` and constrained structured output, with `interactive=false` on the way. But… — swyx 🔜 ai.engineer (@swyx)
#ai#research#academic#threads#companies/Anthropic
·x.com·
swyx 🔜 ai.engineer on X: "IMO @AnthropicAI is very close to making a breakthrough in productizable interpretability. For ~4 years all we've had to really control LLMs is temperature/top_p and logit bias. We recently got `seed` and constrained structured output, with `interactive=false` on the way. But https://t.co/Z8C28evxgN" / X