Stanford Report Reveals Massive Growth and Costs in 2023 AI Landscape
Stanford HAI releases its 2023 AI Index, showing that large language models grow exponentially in size and cost while significantly outperforming traditional benchmarks. The report also highlights the severe environmental impact of training these advanced systems.
Stanford University's Institute for Human-Centered Artificial Intelligence (HAI) releases its 2023 AI Index, providing a detailed snapshot of the current state of artificial intelligence across research, economics, and policy. The independent report tracks and visualizes data to help decision-makers advance AI responsibly. Through 14 detailed charts, the index illustrates exactly how rapidly the technology evolves in the modern era.
Large language models scale up at an unprecedented rate, growing significantly in both size and training expenses. For example, the 2019 GPT-2 model costs an estimated $50,000 to train, while the 2022 PaLM model costs around $8 million. Because these advanced systems consistently beat existing performance benchmarks, researchers now face the challenge of creating more difficult tests to accurately measure AI capabilities.
This massive scaling brings serious environmental consequences, as training large models requires enormous amounts of electricity. The BLOOM model alone consumes 433 megawatt-hours of power, which equals the energy needed to power an average American home for 41 years. These high carbon emissions highlight the growing need for sustainable practices as the AI industry continues its rapid expansion.