They release a series of updates to 4o (Nov 20, Jan 29, Mar 27), trying to invoke similar empathy and emotional realism, which culminates in Mar 2025 when they even had to dial it back down due to twitter complaints
Uncertain: ChatGPT can’t match Sonnet in EQ, cause of the differences between RLHF and RLAIF.
However, it’s “good enough” that people grow emotionally attached to 4o.
OpenAI makes most of the money on their b2c chatgpt.com – Anthropic doesn’t care about b2c as much, they rake in API inference $$$ and claude.ai is like a 5th priority on their list, somewhere after training, alignment, enterprise sales, coding performance
My understanding of the timeline:
Late Oct 2024 – Anthropic releases Claude Sonnet 3.5 (new). It’s REALLY good at EQ. People start talking to it and asking for advice
https://www.anthropic.com/news/3-5-models-and-computer-use
OpenAI is mad – how could they fuck this up? They have to keep up.
https://help.openai.com/en/articles/9624314-model-release-notes#h_826f21517f
They release a series of updates to 4o (Nov 20, Jan 29, Mar 27), trying to invoke similar empathy and emotional realism, which culminates in Mar 2025 when they even had to dial it back down due to twitter complaints
Uncertain: ChatGPT can’t match Sonnet in EQ, cause of the differences between RLHF and RLAIF.
However, it’s “good enough” that people grow emotionally attached to 4o.
OpenAI makes most of the money on their b2c chatgpt.com – Anthropic doesn’t care about b2c as much, they rake in API inference $$$ and claude.ai is like a 5th priority on their list, somewhere after training, alignment, enterprise sales, coding performance