Greetings,

The Frontier Model Forum wants to manage the risks of any AI models more powerful than current state-of-the-art. Basically, any model stronger than GPT-4 will have a much harder time breaking into the market. Are they trying to slow down competitors, or are they really concerned with AI safety?

• Anthropic, Google, Microsoft, and OpenAI launched the Frontier Model Forum to develop standards for evaluating risks of frontier models (= better than the most advanced current models), and to “plug into” global orgs like G-7 and OECD. - Axios
  
  
• GPT-4 defeated engineering and business students from a top college at generating product ideas – suggesting that it’s more creative than the average human, and that idea evaluation and selection will become the key skill for innovators now. - SSRN
  
  
• Netflix is racing ahead with AI-generated content – and ready to pay as much as $900,000 per year to experts that will help them realize their AI-powered future. - Engadget
  
  
• It’s unclear if this can be patched by LLM providers – researchers created a program that writes strings of characters which manipulate both open-source (e.g. Llama) and black-box (e.g. GPT) models into providing harmful content (e.g. how to build a bomb). - LLM-Attacks
  
  
• If you’d like to play around with Llama 2 on your PC, Mac, or iOS / Android phone, here’s a short and simple guide to get started. - Replicate
  
  
• Amazon is expanding their genAI-powered healthcare offering with AWS HealthScribe, which transcribes and summarizes patient histories, and AWS HealthImaging for analyzing medical imaging data at scale. - TechCrunch
  
  
• Shopify users can sign up for early access to Sidekick, their unique AI assistant that can even perform actions within the store – an interesting example of genAI being integrated into an established platform. - Shopify

----

Found this helpful? Forward this email to a colleague.

Was this email forwarded to you? Sign up here to stay on top of the AI news.