# Local models & open-source AI

- Channel: [ZuBerlin](https://streameth.org/zuberlin)
- Date: 2024-06-18
- Duration: 39:44
- Topics: cryptocurrency, ethereum, bitcoin, machine learning, AI
- Watch: https://streameth.org/watch/6671665907f92b086c3b16e4
- Download: https://vod-cdn.lp-playback.studio/raw/jxf4iblf6wlsyor6526t4tcmtmqa/catalyst-vod-com/hls/b29eemudvr9rwqyn/1080p0.mp4

## Description

Clip In this video, Anna, the co-founder and CEO of Vana, discusses her background in machine learning research and crypto, her journey from mining Ethereum to dropping out of school to start a machine learning company, and her work at Vana focusing on self-sovereign technologies for AI and data. She explores the concept of decentralized AI, emphasizing that it doesn't necessarily mean putting everything on the blockchain but can be achieved through local open-source models. Anna contrasts crypto's early fringe use cases with those of AI, such as generating explicit images using private local models. She explains the process of training diffusion models, tweaking model parameters for desired outcomes, and the importance of open-source models for maintaining privacy and control over AI.

Anna highlights community-driven projects where contributors iteratively improve models, like in open-source software development. She also touches on funding challenges faced by open-source AI due to difficulties in monetizing freely available models and how this reliance has shifted towards big tech companies. Addressing data privacy concerns, she discusses Vana's approach to secure data contribution without risking leaks.

The video concludes with a Q&A session where Anna addresses various topics including the potential risks and benefits of open-source AI models, decentralized approaches to model funding and training, the importance of keeping frontier models open source for