Madrona reposted this
I had a fun conversation about the future of Open Source AI with Laurens van der Maaten of Meta, Jonathan Frankle of Databricks, and Ali Farhadi of Ai2. These folks and others are investing heavily to put more control in the hands of AI developers. And that creates a counterweight to closed models, a bit like Android did for iOS. YouTube linked below, but first - here’s a glimpse into our conversation: 🌟 Motivations Behind Open-Source AI: Each speaker shared the unique reasons their lab has chosen to release open-source AI models. A common theme was the importance of decentralization. But why is decentralization important, and what's in it for them? 🤔 Future of Open-Source AI Models: We asked our panel if they believe open-source AI models will meet or surpass the best closed models for general purposes within the next 36 months. 💰💡 Investment and Innovation: With the next wave of capabilities projected to be incredibly expensive (around $10B or more), we explored whether open-source AI might need to slow down, or continue to thrive through direct investment or efficient new techniques. 🤝🖥️ Community Contributions: While community contributions have always been a cornerstone of open-source software, open-source AI presents unique challenges due to the need for significant GPU resources. We spoke about how the community can effectively contribute to open-source AI. 📈🔧 AI-Enabling Software vs. Core Models: Lastly, we discussed the role of AI-enabling software as a complement to open-source AI models. How does open-source software fit into the evolving story of open-source AI? Really enjoyed this discussion. Thanks to the group, and to Eric Newcomer for having us! YouTube Link: https://lnkd.in/gbpDnvX3
Founder of Newcomer, a startups and venture capital media and events business
3dLoved the panel! Thank you so much for moderating for us. Not many VCs we would trust to moderate a panel...