Source item
Open original source ↗OpenAI GPT-6 Astra benchmarks got released. Looks both AGI and the Singularity are almost here.
The leap in eval performance:
From 78.5% -> 100% on ExploitBench to 22.4% -> 64.6% on the Science benchmark is enormous.
And the most striking part is the VP of Research comment around:
"Astra was also OpenAI's first model that involved previous AI models very heavily in the training process, with the older generations helping train the new one."
So we also get the beginnings of recursive self-improvement as a bonus.
I don't think Bernie Sanders will be too happy about this?
Quoted context
These are the quoted author’s words.
@aleabitoreddit Could you post your thoughts on the new Astra 6 benchmarks from open AI ? Seems like we are getting closer and closer to AGI
Wolfex_Yeat ↗