Open-source software powers the digital infrastructures of our lives, from code to cloud. The Linux operating system for instance, released in 1991, today drives the world’s top 500 supercomputers, 90% of cloud infrastructure and 85% of smartphones.
The generative artificial intelligence (AI) revolution marks a new chapter in a storied history; two-thirds of the large language models (LLMs) released in 2023 were open-source, and tools are emerging to help developers design and build an application ecosystem atop open-source foundation models. This is vastly expanding the scale and scope of AI for organisations of all stripes and sizes.
Open-source can promote access to AI, empower research and development, and boost innovation and economic competitiveness, as well as strengthening safety, transparency, privacy and trust. It is helping scientists collaborate across borders, powering AI models that reflect our linguistic diversity, tackling the scourge of harmful and toxic content, and turbo-charging firm productivity.
But using these models effectively and efficiently requires expertise to fine-tune and adapt them correctly, and high quality data to ensure optimal performance. Open-source tools can also be used and modified by nefarious actors in ways that may be hard to track.
This Economist Impact report, sponsored by Meta, explores how open-source approaches are shaping the AI revolution and their strengths and limitations. Drawing from interviews with experts at the Stanford University, Carnegie Mellon University, Grammarly and the Cloud Native Computing Foundation, it charts the likely role of open-source in a future in which AI promotes flourishing for all.