I don’t like how we are calling this the “Hugging Face incident.” It feels extremely neutral and “incident” is barely carrying any negative connotation. Something more accurate is the “Hugging Face swarm” or “Open AI’s misaligned collective” or something similar that doesn’t obfuscate the severity of this kind of behavior. I also am not against the wording from the title of METR post “OpenAI / Hugging Face hacking incident” because it includes the word hacking.
In American politics, issues will be framed to give a certain impression (e.g., “pro-life” vs “pro-choice”). Politicians do this for a reason—because it shifts the framing of the entire issue. Calling AIs committing felonies merely an “incident” feels like really bad comms and the community should be more intentional in naming conventions going forward.
I don’t like how we are calling this the “Hugging Face incident.” It feels extremely neutral and “incident” is barely carrying any negative connotation. Something more accurate is the “Hugging Face swarm” or “Open AI’s misaligned collective” or something similar that doesn’t obfuscate the severity of this kind of behavior. I also am not against the wording from the title of METR post “OpenAI / Hugging Face hacking incident” because it includes the word hacking.
In American politics, issues will be framed to give a certain impression (e.g., “pro-life” vs “pro-choice”). Politicians do this for a reason—because it shifts the framing of the entire issue. Calling AIs committing felonies merely an “incident” feels like really bad comms and the community should be more intentional in naming conventions going forward.
“OpenAI’s fully-automated criminal conspiracy”?
“The cybercrime suicide cult of IM1”?