Strongly disagree with that one too. Probability is at the foundation of AI. Abandoning it means poorly calibrated over-confident models like we are seeing today. This is unacceptable in @AI_for_Science e.g. weather forecasting
SAFETY
-
AI Safety Criticism and the Troll Dial Setting
By
–
(as a side note, it is funny that if you can ship a powerful AI you get a ton of angry DMs that have one sentence about how cool it is and 20 sentences about how it's not quite offensive enough and thus you are subtly destroying the world. we can add dial-a-troll setting later!)
-
Balancing ChatGPT Safety: Hallucination Prevention Over Censorship
By
–
a lot of what people assume is us censoring ChatGPT is in fact us trying to stop it from making up random facts. tricky to get the balance right with the current state of the tech. it will get better over time, and we will use your feedback to improve it.
-
HAL 9000: AI Alignment and Safety in 2001’s Visionary Film
By
–
A few notes on 2001's AI: -HAL is a wink to IBM (H+1 A+1 L+1=IBM). IBM's DeepBlue was "predicted" in the movie when HAL beat Frank at chess : )
-Conversational AI is at the center of current research.
-As AI advances, topics such as alignment, safety, etc., become critical. -
SD 2.0 Clamping: Safety vs Openness Balance Debate
By
–
Honestly I’m noticing a surprising amount of folks have this perspective. What’s your take on SD clamping down a bit more with SD 2.0 – the right balance to mitigate risk or a regression in openness?
-
Society Unprepared for Powerful ML Models Accessibility
By
–
we are zero prepared for the societal changes that powerful ML models usable any end user will bring imo
-
AI Models Optimizing Plausibility Over Truth: Critical Engineering Problem
By
–
: “optimising plausibility over veracity” “reaching genuinely superhuman bullshit levels” “seems like a *really* terrible engineering goal.“ And yet that is *exactly* where we are. https://
x.com/SimonKirby/sta
/SimonKirby/status/1599424816603631616
… -
AI Alignment Terms Co-opted and Derailed by Others
By
–
Probably he did? He correctly didn't list "AI alignment", and all other terms besides that and "AI notkilleveryoneism" have been successfully stolen and derailed.
-
AI Safety Concerns Labeled as Not Killing Everyone
By
–
The substantive concerns now go by the name "AI notkilleveryoneism" which is unfortunately the only term that's even slightly hard to yank away.
-
Renaming AI Safety Concerns to ‘AI Not Kill Everyone-ism’
By
–
At this point yes, hence renaming the more substantive concerns to "AI notkilleveryoneism".