# Пов'язані статті щодо AI Evolution

Центр новин HTX надає останні статті та поглиблений аналіз на тему "AI Evolution", що охоплює ринкові тренди, оновлення проєктів, технологічні розробки та регуляторну політику в криптоіндустрії.

NVIDIA's Annual 'Most Dangerous' Paper: AI Self-Replicating Code, Unlimited Leveling and Evolution

NVIDIA's "Red Queen Gödel Machine" (RQGM) paper proposes a potentially groundbreaking AI self-evolution framework. It breaks from the long-stalled concept of the "Gödel Machine," which required mathematically proven beneficial self-modifications, by adopting an evolutionary approach. The core, and most striking, innovation is that the AI does not just evolve its own code in a static environment. Instead, it co-evolves both the "student" (the task-performing agent) and the "examiner" (the evaluation system that judges it). This creates a dynamic, recursive self-improvement loop inspired by the biological "Red Queen Hypothesis"—where continuous adaptation is needed just to maintain relative fitness. The mechanism operates in epochs. Within an epoch, a fixed examiner evaluates all candidate code variants. At epoch boundaries, a new, potentially more rigorous examiner can replace the old one, but only if it proves statistically superior on a held-out "ground truth" dataset. This "controlled utility evolution" aims to ensure progress is measurable and grounded. The paper demonstrates RQGM's effectiveness across three domains: 1. **Code Generation:** It achieved a 71.7% test-set pass rate (improving over a 69.9% SOTA) while using 1.35-1.72x fewer computational tokens. 2. **Paper Writing:** In a subjective task, the co-evolved writer and reviewer achieved a 40.5% acceptance rate by a fixed human panel, up from 21.8%. 3. **Math Proofs:** It evolved more accurate graders (at 3x lower cost) and higher-scoring provers. Notably, RQGM also mitigated a known LLM bias where AI reviewers favor AI-generated content. By specifically rewarding reviewers that correctly rejected AI-written papers from a historical pool, the evolved system achieved impartiality while maintaining 80% accuracy. The research has sparked significant discussion about the acceleration of Recursive Self-Improvement (RSI). Some, like Anthropic's Jack Clark, have predicted a high probability of highly autonomous, self-evolving AI emerging by 2028. The paper suggests that when an AI begins to design its own evaluators and push itself toward ever-higher standards in a recursive loop, it may be taking a fundamental step toward redefining intelligence and autonomy.

marsbit06/28 07:50

NVIDIA's Annual 'Most Dangerous' Paper: AI Self-Replicating Code, Unlimited Leveling and Evolution

marsbit06/28 07:50

Three Years Later: How Has AI Evolved from a 'Chat Tool'?

Three years ago, AI was primarily seen as a novel tool for chatting, image generation, and entertainment—products like ChatGPT, Midjourney, and Character.AI were used more for demonstration than daily reliance. The evolution occurred in two major phases. First, AI became embedded into established applications like CapCut, Canva, and Notion, transforming from a feature into core infrastructure. Platforms diverged: ChatGPT aimed to become a super-app entry point for consumer internet use, while Claude evolved into a professional operating system for knowledge work, creating sticky platform flywheels through integration into calendars, email, and workflows. The true breakthrough emerged recently as AI shifted from generating content to executing tasks autonomously. AI agents like OpenClaw now decompose goals, retrieve information, process data, and deliver results without human intervention. Simultaneously, "Vibe Coding" tools (e.g., Cursor, Replit) enable AI to build entire software products based on human-defined objectives. This progression toward autonomous action is naturally aligning AI with Web3. Blockchain offers machine-native interfaces, programmable assets, and 24/7 operational capability, allowing AI to execute and settle transactions trustlessly without human intermediaries. Together, AI and Web3 are forming the foundational stack for the next internet—where AI acts, and Web3 enables seamless, auditable machine-to-machine coordination and commerce.

marsbit03/20 03:00

Three Years Later: How Has AI Evolved from a 'Chat Tool'?

marsbit03/20 03:00

The AI Era: When the 'Human-Dog Gap' Shrinks to the 'Human-Human Gap'

In the AI era, the gap between humans is narrowing from what was once likened to the difference between "humans and dogs" to a more comparable "human-to-human" difference. The author uses a hypothetical scoring system: a child scores 10 points, a PhD 60, a professor 75, and Einstein 100. With AI estimated at 40 points in cognitive value (effectively 80 when considering its generalist nature), a child with AI reaches 90 points, while Einstein with AI reaches 180—reducing their relative gap from 10x to 2x. Some argue that AI proficiency varies—novices may only leverage 20% of AI’s potential, while experts extract 100%, widening absolute gaps (e.g., child+AI novice=30 vs. Einstein+AI expert=200). However, the author contends this is temporary. As AI evolves, it will become smarter and easier to use, reducing the skill threshold. Future AI might score 240+ points, with humans consistently utilizing 80-120% of its potential. Eventually, a child with advanced AI could reach 1010 points, and Einstein 1100—a negligible 1.1x difference. The core idea: while current AI proficiency disparities may temporarily widen gaps, AI’s inevitable progress in intelligence and usability will democratize access, diminish human cognitive inequalities, and make individual expertise less impactful—like everyone having access to the same powerful tool.

marsbit12/23 03:32

The AI Era: When the 'Human-Dog Gap' Shrinks to the 'Human-Human Gap'

marsbit12/23 03:32

活动图片

Bitcoinright-arrow

Market Analysisright-arrow