Artificial Intelligence Needs Community Intelligence - Sriram Raghavan, IBM Research AI

Editor's Notes

• #2: Making AI truly open is the center of many lively debates. Even the phrase “Open AI” has been the center of much controversy. The skeptics say … “Open will always be behind” Skeptics argue the expertise and knowledge for innovation doesn’t exist in the open They argue open can’t compete with the corporate investments being made in infrastructure and data
• #3: Making AI truly open is in fact hard. But where we differ, is whether it is possible. That’s because we remember the history of open source. We didn’t always know how to do this. The early days of open source were painful, and it took decades to achieve progress. It started with the GNU project, which led to Linux. The Apache 2.0 license made things easier, followed by Git, and then even more streamlined with pull requests. But ultimately, we got there. It was possible. It was only because of the collective genius of the community, coming up with these necessary innovations (technical and non-technical), that we could pull it off.
• #4: The AI community is already solving the impossible Innovations have progressed AI from raw models, to customizing models to power assistants, to getting advanced capabilities and reasoning from models in agents. No signs of slowing. The community is driving more breakthroughs, at a faster pace, than ever before.
• #5: IBM is doing its part to make all parts of AI more open. From open models – truly designed for everyone- with lightweight models that don’t require expensive or hard to get GPUs, with an Apache 2.0 license To open tuning – making it possible for anyone to contribute additional knowledge to LLMs, without needing expert-level LLM training skills To open agents – BeeAI allows you to not just use its own agents, but also plug in agents from those other technologies To open collaboration – we contribute back to key projects that enable a highly efficient and scalable AI stack that we and others use
• #6: IBM is contributing 3 of our most popular AI open source proejcts to the Linux Foundation With the intent to contribute them to LF – AI & Data This signals our commitment to building AI in the open, with and for the community ---Data prep kit to helps clean, transform and enrich unstructured data (accounts for 90% of all enterprise generated data) --Docling, which has been extraordinarily popular, streamlines the process of turning unstructured documents into JSON and Markdown files --BeeAI which allows you to not just use its own agents, but also plug in agents from those other technologies This is so we can expand the reach and attract new contributors and consumers Now, people can invest their resources- pathway to more influence and the peace of mind of stability of project at a neutral foundation
• #7: Community offers inventions – new technology. Community must also build innovations – apply technology to meet a need. Be a creator, not just a consumer. Examples: Optional sidebar (indirect reference to open communities winning out over OpenAI): Commercial entities can’t solve this problem- they can’t compete with the entire community on ideas and experimentation. Historically, greatest innovations were community not commercial entites (internet, genome, GPS, new forms of energy)