A new paper co-authored by NVIDIA, Cambridge University, and other institutions, titled "Red Queen Gödel Machine," has sparked debate over a potential 2028 timeline for Artificial Superintelligence (ASI). The paper describes a mechanism where AI autonomously develops new learning algorithms, tests them in a sandbox, discards failures, and retains successes for further self-evolution. The AI then actively evolves its own examiner, creating stricter judges to evaluate its own code, locking it into an endless cycle of self-iteration. This approach builds on the 2003 concept of the "Gödel Machine," which required mathematical proof before self-modification—a task requiring "black hole"-level computational power. The new Darwinian Gödel Machine (DGM) and Huxleyan Gödel Machine (HGM) abandon mathematical proofs for evolution, generating mutated code variants. Anthropic co-founder Jack Clark predicts a 60% probability that a highly autonomous, self-evolving AI will emerge by the end of 2028.