Research indicates AI may hit a bottleneck in intelligence as companies like OpenAI are depleting available human-written training data. With diminishing new content and lawsuits over data use, AI advancement could stall. Larger models like GPT-4 may exhaust data sources by 2026, while training on AI-generated data could lead to quality erosion, creating an 'autophagous loop.' Despite challenges, efficiency improvements in AI could negate the need for vast training data.