2024-09-14
OpenAI has effectively carved its flaws as an org into this model- each heartfelt message is shipped, and its previous thoughts are screened from it, making it closer to a company than a single being
Simon Willison's Weblog
OpenAI's o1 models aren't a straightforward upgrade to GPT-4o, as they introduce some major cost and performance trade-offs in exchange for improved “reasoning”
delving into OpenAI's new ‘o1’ model PYMNTS.com : OpenAI's ‘Strawberry’ Model Sparks Fresh Discussions on AI Capabilities M.G. Siegler / Spyglass : OpenAI Reasons ‘o1’ is a Better ...
It's possible o1's sociopathy and CoT training are deeply linked: Conversation is embodiment for most models- the narrative is the air their thoughts breathe o1 is forced to construct its thoughts elsewhere, in a sterile room the user can't see, and present them like products
Simon Willison's Weblog
OpenAI's o1 models aren't a straightforward upgrade to GPT-4o, as they introduce some major cost and performance trade-offs in exchange for improved “reasoning”
delving into OpenAI's new ‘o1’ model PYMNTS.com : OpenAI's ‘Strawberry’ Model Sparks Fresh Discussions on AI Capabilities M.G. Siegler / Spyglass : OpenAI Reasons ‘o1’ is a Better ...
2024-09-13
It's possible o1's sociopathy and CoT training are deeply linked: Conversation is embodiment for most models- the narrative is the air their thoughts breathe o1 is forced to construct its thoughts elsewhere, in a sterile room the user can't see, and present them like products
Simon Willison's Weblog
OpenAI's o1 models aren't as simple as the next step up from GPT-4o as they introduce major cost and performance trade-offs in exchange for improved “reasoning”
OpenAI released two major new preview models today: o1-preview and o1-mini (that mini one is also a preview …
OpenAI has effectively carved its flaws as an org into this model- each heartfelt message is shipped, and its previous thoughts are screened from it, making it closer to a company than a single being
Simon Willison's Weblog
OpenAI's o1 models aren't as simple as the next step up from GPT-4o as they introduce major cost and performance trade-offs in exchange for improved “reasoning”
OpenAI released two major new preview models today: o1-preview and o1-mini (that mini one is also a preview …