https://aimove37.com Essays and analysis on AI's move 37 moments. https://aimove37.com/alignment-tax/ The cost of making AI systems safe and how it impacts their raw creativity. https://aimove37.com/cicero/ How Meta's Cicero mastered the game of Diplomacy by combining natural language processing with strategic planning. https://aimove37.com/the-bitter-lesson/ Rich Sutton's famous essay proved right: compute scales better than human heuristics. https://aimove37.com/creativity-vs-computation/ Analyzing the philosophical and algorithmic boundary between calculating a result and creating something new. https://aimove37.com/gpt-4-sparks/ Analyzing the zero-shot reasoning capabilities of GPT-4 and the debate over emergent behavior in LLMs. https://aimove37.com/about/ The mission and philosophy behind documenting AI's emergent capabilities. https://aimove37.com/scaling-laws/ How predictable scaling laws govern the emergent capabilities of large language models. https://aimove37.com/dota-2/ How OpenAI conquered the complex, long-horizon strategy of Dota 2 using massive scale reinforcement learning. https://aimove37.com/methodology/ How aimove37.com evaluates and categorizes machine breakthroughs. https://aimove37.com/hallucination-or-invention/ When an LLM generates false information, is it failing, or is it exhibiting nascent creativity? https://aimove37.com/protein-folding/ How AI conquered a 50-year-old grand challenge in biology by predicting 3D protein structures. https://aimove37.com/alpha-go/ Analysis of AlphaGo's match against Lee Sedol and the true meaning of Move 37. https://aimove37.com/ Analyzing AI's Move 37 moments across history, from AlphaGo to GPT-4. https://aimove37.com/gran-turismo/ How GT Sophy mastered the highly complex physics and racing etiquette of Gran Turismo Sport. https://aimove37.com/objective-functions/ The danger of mathematical goals in AI and the paperclip maximizer thought experiment. https://aimove37.com/alphastar/ How DeepMind conquered StarCraft II, mastering imperfect information and real-time continuous spaces. https://aimove37.com/pluribus/ Analysis of Pluribus, the AI that defeated professional players in six-player No-Limit Texas Hold'em.