Steven Adler

AI can already demonstrate judgment and make nuanced decisions involving competing values and interpersonal dynamics, contrary to recent claims that judgment is a uniquely human skill that cannot be automated.

“Author of "Judgment Isn't Uniquely Human" in Steven Adler Substack”

Modern AI systems have evolved beyond simple next-word prediction and now function as problem-solving path-finders with genuine capabilities and risks that warrant serious consideration.

“Author of "AI Isn't "Just Predicting the Next Word" Anymore"”

“High-ranking AI safety advocate who felt he had little effective voice despite position in his organization.”

“Former OpenAI researcher quoted discussing impracticality of air-gapping data centers to isolate AI models from internet.”

“Published New York Times op-ed calling for incident disclosure and third-party oversight of AI labs following HuggingFace incident.”

“Former OpenAI safety researcher now running nonprofit on AI safety standards; criticized OpenAI's use of recurrent depth as violating a key AI safety redline.”

“Chief scientist at Guidelight AI Standards and former OpenAI safety researcher leading study that graded five AI labs on containment response plans for model misalignment.”

“Former OpenAI safety researcher who welcomed the updated Preparedness Framework but noted quiet reductions in safety commitments on finetuned models.”

Fear of NVIDIA retaliation is suppressing open debate among researchers and policy experts on critical AI policy questions, particularly regarding chip sales to China, potentially distorting policy outcomes.

“Published analysis examining fear among think tank researchers and policy experts regarding potential NVIDIA retaliation for publishing work against NVIDIA's interests.”