openAI o1 and "o1-mini" (aka strawberry)

 openAI o1 and "o1-mini" (aka strawberry) (as of end Sept '24)
  • in preview, rate limited "50 queries per week for o1-preview and 50 queries per day for o1-mini."
    • 'ChatGPT Plus" and "Team" users will be able to access o1 models in ChatGPT already.
  • not in copilot yet, only a matter of time I assume

notes:
  • openAi trained these models to spend more time thinking through problems before they respond ...like a person would. 
    • learned to refine their thinking process, try different strategies, and recognize their mistakes.
    • o1 model update performs similarly to PhD students on challenging benchmark tasks in physics, chemistry, and biology. 

    • enhanced reasoning capabilities may be particularly useful if you’re tackling complex problems in science, coding, math, and similar field

  • for complex reasoning tasks this is a significant advancement and represents a new level of AI capability. 

  • it doesn't just predict, it thinks
  • o1 uses advanced reasoning techniques like Chain of Thought (CoT) to solve complex problems step by step
    • chain of thought is an AI technique which mirrors human reasoning and breaks down problems into logical steps

  • such is the advance that is o1, that openAi reset the counter back to 1
  • o1 series excels at accurately generating and debugging complex code. 
  • openAI "o1-mini", a faster, cheaper reasoning model that is particularly effective at coding
    • 80% cheaper than o1-preview, making it a powerful, cost-effective model for applications that require reasoning but not broad world knowledge. 

sources: openAI, copilot and reddit, ibm etc.

Comments

Popular posts from this blog

My Reading Lists

angular js protractor e2e cheatsheet

deep dive into Material UI TextField built by mui