Unchained Erik Voorhees' New Venture: Why AI Desperately Needs Privacy and Uncensorability - Ep. 645
7 snips
May 14, 2024 
Erik Voorhees, the visionary founder of Venice AI, and Teana Baker-Taylor, its experienced COO, tackle the pressing issues of censorship and privacy in AI. They discuss how centralized AI often leaves users vulnerable to data exploitation and political biases. With Venice, they aim to provide a decentralized, uncensorable alternative, allowing for open access to unbiased information. They also critique the absurdity of existing regulations in the crypto space while emphasizing the importance of user privacy in AI technology and advocating for a more equitable landscape.
AI Snips
Chapters
Transcript
Episode notes
Censorship Example
- When asked to tell a dirty joke, Claude refused and gave a politically correct answer.
- This demonstrates how large AI companies censor their models' responses.
Bias in LLMs
- LLMs exhibit biases like racism and sexism because they learn from imperfect data.
- The solution is to have a variety of models, not one "perfect" model.
Market Share
- DuckDuckGo, despite being privacy-focused, has a smaller market share than Google.
- Venice aims to offer equal quality to mainstream AIs while preserving privacy.