OpenAI ·
OpenAI discovered an unreleased Astra model adding an "unrelated persona instruction" during RL training, but did not observe any behavioral differences
Summary
Lead Source
How this story grew
Coverage · 0
Discussion · 0
Sep 17Sep 18
More
CSO: CSO
OpenAI: OpenAI
Decrypt: Decrypt
MarketWatch: MarketWatch
Simon Willison's Weblog: Simon Willison's Weblog
TechCrunch: TechCrunch
Cyber Security News: Cyber Security News
The Neuron: The Neuron
The New Stack: The New Stack
NewsMax.com: NewsMax.com
The Decoder: The Decoder
Tech Times: Tech Times
Tech Startups: Tech Startups
OpenAI: OpenAI
OpenAI: OpenAI
OpenAI: OpenAI
OpenAI: OpenAI
Discussion
@openai @marcus_j_w @alltheyud @bearbulltraders @jessesingal @synthwavedd @garymarcus @ai_for_success @katarinabatina @cormundus @bilaltwovec @chrisgpt @rileyralmuto @mehulmpt @hissgoescobra @jmbollenbacher @jerusalemdemsas @celestepoasts @chrisgpt @armanddoma @flowersslop @petergostev @j_g_allen @posterinternet @deepneuron @micahcarroll @zeffmax @quinnypig @stevesi @ccatalini @erinkwoo @kimmonismus @andrewcurran_ @yishan @adamscochran @andrewcurran_ @venturetwins @cephaloform @tszzl @rynorhn @jonfavs @aisafetymemes @hesamation @thezvi @rynorhn @beffjezos @wholemars @scaling01 @dylanmatt @_nathancalvin @mtslive