Users accused the AI policy expert's push for accountability over blameless postmortems of enabling cover-ups rather than learning, with some sarcastically mocking the framing and others demanding outright bans.
No Digg Deeper questions have been answered for this story yet.
Same thing happened with emotional attachment, cyber misuse, sycophancy, etc. - none surprising to those paying attention, but treated (not just by OAI) as something to deal with as it comes, or rather when people start complaining loudly. Enough with the smol bean stuff.
Ya this is also why I am extremely opposed to calling OAI or Ant a startup
It is important to strike a balance between a culture of continuous learning, “blameless postmortems,” etc. on the one hand and “someone needs to ultimately be accountable/not everything can be excused” on the other. The AI industry is tilted way too far towards the former RN
Something I wish people appreciated more is that while the details of AI safety incidents are often hard to predict, the broad outlines rarely are. = one reason the emphasis on “iterative deployment” irks me here... none of this is unheard of or excusable
This is not an AI specific point. Warning signs, lax safety culture, etc. generally precede disasters in aviation, nuclear, etc. We need less “AI exceptionalism” here. Sure it may be unusually dangerous later but let’s start by applying lessons from the usually dangerous stuff!
It is important to strike a balance between a culture of continuous learning, “blameless postmortems,” etc. on the one hand and “someone needs to ultimately be accountable/not everything can be excused” on the other. The AI industry is tilted way too far towards the former RN
Ya this is also why I am extremely opposed to calling OAI or Ant a startup
@Miles_Brundage big “I’m just a lil guy!” energy
@Miles_Brundage Prefer to just ban this shit altogether.
@Miles_Brundage Postmortems are for learning, not washing the file
@Miles_Brundage big “I’m just a lil guy!” energy