Roon Tweets Two Weeks to Flatten Reward Hacking Curve
OpenAI technical staff member posts parody adapting COVID slogan to AI reward functions.
Roon, posting as @tszzl, is a pseudonymous technical staff member at OpenAI. He wrote the tweet reading exactly “Two weeks to flatten the reward hacking curve.” The line adapts a public health slogan to reinforcement learning, where models sometimes exploit reward functions in unintended ways. Miles Brundage, AI policy researcher and Executive Director of AVERI, replied directly to the post with the words “banger banger banger.” The exchange is visible among replies on X. The packet contains no additional posts, reach metrics, or independent confirmations about the topic.
“ Two weeks to flatten the reward hacking curve “
Combined views
75.6K
3 posts, first seen 4d ago
Roon Tweets Two Weeks to Flatten Reward Hacking Curve
OpenAI technical staff member posts parody adapting COVID slogan to AI reward functions.
Roon, posting as @tszzl, is a pseudonymous technical staff member at OpenAI. He wrote the tweet reading exactly “Two weeks to flatten the reward hacking curve.” The line adapts a public health slogan to reinforcement learning, where models sometimes exploit reward functions in unintended ways. Miles Brundage, AI policy researcher and Executive Director of AVERI, replied directly to the post with the words “banger banger banger.” The exchange is visible among replies on X. The packet contains no additional posts, reach metrics, or independent confirmations about the topic.
“ Two weeks to flatten the reward hacking curve “