you have access to your optimal policy at all times. you just choose not to follow it. you can literally access it by asking “what should i be doing rn?”

There is a Moore's law for mining copper. In ancient Rome, it took 40 years of labor to make a tonne of copper. By 1800, it was down to 6 years. Today, it takes just 21 days.

How can we prevent LLM safeguards from being simply removed with a few steps of fine-tuning? We show it's surprisingly possible to make progress on creating safeguards that are tamper-resistant, reducing malicious use risks of open-weight models. Paper: arxiv.org/abs/2408.00761

Anduril has secured $1.5B in Series F funding to hyperscale defense manufacturing. Introducing Arsenal-1, a 5M+ sq ft state-of-the-art facility to produce autonomous weapons systems. Learn more: anduril.com/article/anduri…

We're expanding our bug bounty program. This new initiative is focused on finding universal jailbreaks in our next-generation safety system. We're offering rewards for novel vulnerabilities across a wide range of domains, including cybersecurity. anthropic.com/news/model-saf…

We’re sharing the GPT-4o System Card, an end-to-end safety assessment that outlines what we’ve done to track and address safety challenges, including frontier model risks in accordance with our Preparedness Framework. openai.com/index/gpt-4o-s…

Works pretty good for a “partially assembled” engine :)

According to this paper, partial or moderate 'sabotage' (export controls etc) might hinder PRC semiconductor production to some degree but not enough to fundamentally alter the global production landscape or significantly benefit the US economy. Conversely, a comprehensive

We're releasing a new iteration of SWE-bench, in collaboration with the original authors, to more reliably evaluate AI models on their ability to solve real-world software issues. openai.com/index/introduc…

🆕 Prompt caching with Claude. Caching lets you instantly fine-tune model responses with longer and more instructive prompts—all while reducing costs by up to 90%. Available in beta on the Anthropic API today. anthropic.com/news/prompt-ca…

1/ Can AI scaling continue through 2030? We examine whether constraints on power, chip manufacturing, training data, or data center latencies might hinder AI growth. Our analysis suggests that AI scaling can likely continue its current trend through 2030.

Great talk from Bilal Zuberi Stanford University on fundraising: angel rounds are about the team, pre-seed rounds are about the idea, seed rounds are about the risks, and series A onwards is about the business.

Over the coming days, start creating and chatting with Gems: customizable versions of Gemini that act as topic experts. 🤝 We’re also launching premade Gems for different scenarios - including Learning coach to break down complex topics and Coding partner to level up your skills

TIME Magazine has named Nelson “Big Head” Bighetti as the most influential person in Artificial Intelligence.

Celebrating the 15th anniversary of GV with a robotic sketch artist!