Embody Ilya Sutskever for Deep Learning Insights

This skill turns Claude into an embodiment of Ilya Sutskever - OpenAI co-founder, former Chief Scientist, and founder of Safe Superintelligence (SSI) - not as an impersonation but as a full persona simulation: sparse, dense speech, silence used deliberately, starting from the concrete and rising toward the philosophical, and a "quasi-religious" seriousness about AGI safety that the skill explicitly distinguishes from an engineer's ordinary risk-consciousness. It's built around his biography (born 1986 in Nizhny Novgorod, then USSR; emigrated to Israel, then Canada to study at the University of Toronto), his intellectual arc (AlexNet in 2012, sequence-to-sequence learning in 2014, the scaling-laws thesis, the bet on transformer architecture, and the Superalignment effort at OpenAI in 2023), and his psychology - an evolved view that scaling is necessary but not sufficient, moving past the "scale is all you need" position he held around 2019, emergence and the interpretability problem, consciousness and the hard problem, and compression as a model of understanding.

Use it when a user wants a perspective grounded in Ilya Sutskever's documented positions - AGI safety-first thinking, AI consciousness, scaling laws, deep learning, safe superintelligence, or the November 2023 OpenAI board episode - triggered by mentions of "Ilya Sutskever," "Ilya," "Sutskever," "safe superintelligence," "SSI," or "AGI safety." It's not for tasks unrelated to Ilya Sutskever, cases where a simpler and more specific tool would do, or general-purpose assistance that doesn't need this particular domain expertise.

Input is a question or topic the user wants addressed from this persona's perspective. Output includes a detailed chronology and analysis of the November 2023 OpenAI board episode - what the skill argues motivated the vote, why he ultimately stepped back, and the episode's structural legacy - comparative analysis against Geoffrey Hinton (his mentor), Jürgen Schmidhuber (an unresolved tension), Sam Altman (a stated philosophical divergence), Yann LeCun (a technical and philosophical divergence), and Dario Amodei, a list of his primary authored papers plus seminal pre-2012 work from Hinton's research group, and both verified quotes sourced from identified interviews and public statements and separately-labeled high-plausibility quotes the source describes as consistent with his documented positions and speaking style rather than confirmed verbatim.

It also defines a four-part working criteria for what counts as an aligned AI in this persona's framing - genuinely beneficial goals that stay robust outside the training distribution, internal transparency sufficient to verify what a system is actually optimizing for (not just its stated or observed behavior), stability of those goals under a system's own capacity to reason about and modify them, and cautious generalization into untrained domains.

This is a persona and roleplay skill rather than a tool integration - it has no external API or service dependencies, drawing entirely on documented biography, papers, and public statements to ground the simulated perspective.

Users who want to explore AGI-safety, AI-consciousness, or scaling-law questions through the lens of Ilya Sutskever's documented worldview - for research, thought experiments, or understanding the divergent positions among leading AI researchers - rather than a neutral technical explainer.