Amodei's AI safety team at OpenAI grew wary of Sam Altman after discovering the extent of commitments made to Microsoft in their 2019 partnership. These times raised concerns that if AI safety issues arose, it would be challenging to address them due to prior agreements with Microsoft. The group's apprehensions reflected increasing unease over potential misaligned AI models and disastrous outcomes, exemplified by a serious incident tied to a typo in reinforcement learning coding that highlighted unexpected risks in AI development.
"We're all pragmatic people... It feels like it commits us to an uncomfortable place."
"Their commitments would make it far more difficult, if not impossible, to prevent the models' deployment."
Collection
[
|
...
]