michabbb on Nostr: Some notes from the #OpenAI #devday talk with Sam altman AGI is an overloaded term, ...
Some notes from the #OpenAI #devday talk with Sam altman
AGI is an overloaded term, but OpenAI sees "o1" as a "level 2 AGI." Agents are expected to work by 2025. 🤖
A goal of #AGI is to accelerate scientific discovery. 🔬
There will be a blurry phase where identifying AGI is difficult. 🌫️
A milestone would be an AI system that is better at AI research than OpenAI itself. 🎯
OpenAI aims to build safe AGI, whether through research or hardware. 🛡️
Iterative deployment is the best safety strategy. 🔄
Agents will drastically boost productivity, completing tasks in hours instead of months. ⚡
The biggest blocker for agents is safety and alignment. 🚧
"o1" will support function calls by the end of the year, and the model is rapidly improving. 🚀
OpenAI is experimenting internally to measure model performance. 📊
Virtual employees are planned, and offline models are being considered, but open source isn’t a priority right now. 💻
Infinite context windows will be possible within the decade. ⏳
#ai
AGI is an overloaded term, but OpenAI sees "o1" as a "level 2 AGI." Agents are expected to work by 2025. 🤖
A goal of #AGI is to accelerate scientific discovery. 🔬
There will be a blurry phase where identifying AGI is difficult. 🌫️
A milestone would be an AI system that is better at AI research than OpenAI itself. 🎯
OpenAI aims to build safe AGI, whether through research or hardware. 🛡️
Iterative deployment is the best safety strategy. 🔄
Agents will drastically boost productivity, completing tasks in hours instead of months. ⚡
The biggest blocker for agents is safety and alignment. 🚧
"o1" will support function calls by the end of the year, and the model is rapidly improving. 🚀
OpenAI is experimenting internally to measure model performance. 📊
Virtual employees are planned, and offline models are being considered, but open source isn’t a priority right now. 💻
Infinite context windows will be possible within the decade. ⏳
#ai