Progress on AI safety is happening. Governments, companies, academics and civil society are coming together to create clear commitments and secure a better, safer roll out of AI.
POLICY
-
Balanced Analysis of Executive Order on AI Needed
By
–
FWIW I do think there are over-reactions to the EO, and I'd be interested to hear a thoughtful analysis of it. I also think it's not entirely without problems either.
-
Google CEO Testifies Before Congress on AI Regulation
By
–
Wish I’d been in Washington today for Google’s CEO’s testimony.
-
Red-teaming large models: fascism claims examined
By
–
Have you seen thoughtful people that you wish to engage with claiming that asking for redteaming of >$100M models is fascism? If not, engage with what those people actually say.
-
AI Regulation Loophole: Bitwise Operations Escape Rules
By
–
The AI Regulation says nothing about Bitwise operations — only floating or integer.
Bitwise aficionados go party for now. -
AI Regulation Unclear on Fine-tuning Reporting Requirements
By
–
The AI Regulation ill-specifies fine-tuning. In this ill-defined state, its not clear if you train a model for 1e25 flops, release it, and someone else trains it further for 1e25 flops, if any reporting has to occur.
-
AI Model Size Regulation and Training Compute Requirements
By
–
Regulation starts at roughly two orders of magnitude larger than a ~70B Transformer trained on 2T tokens — which is ~5e24. Note: increasing the size of the dataset OR the size of the transformer increases training flops. The (rumored) size of GPT-4 is regulated.
-
Meta on Content Moderation and Government Regulation Responsibility
By
–
My name is not Mark Zuckerberg.
This was about asking governments for definitions of what constitutes acceptable or unacceptable content online.
Meta doesn't see itself as having the legitimacy to do so.
But in the absence of regulation, it has to do so.
It's actually the exact -
Open Source AI Ban Won’t Stop Rogue Nation Threats
By
–
If you are asking:
"Should democracies protect themselves against threats by bad actors and rogue nations?"
The answer is obviously yes.
But the real question is:
"Will a ban on open source AI reduce the threat from rogue nations and bad actors?"
The answer is a resounding no. In -
AI-powered models improve hospital workflows with safety safeguards
By
–
AI-powered models can improve hospital workflows and guide treatment decisions, but we need a nationwide plan to make sure the technology does not harm patients. @StanfordMed