MLOps.community Behavior Modeling, Secondary AI Effects, Bias Reduction & Synthetic Data // Devansh Devansh // #311
28 snips
May 2, 2025 In this engaging discussion, Devansh Devansh, an open-source AI researcher and Head of AI at a stealth startup, shares insights on grounded AI research and the biases present in data. He emphasizes the balance between deterministic systems and autonomous agents, urging a rethink of data infrastructures. The conversation delves into the potential of synthetic data for reducing biases and enhancing fairness, encouraging listeners to consider ethical implications. Devansh also highlights the critical role of behavioral modeling in improving user experiences and insights.
AI Snips
Chapters
Transcript
Episode notes
Jailbreaking DeepSeek Hack
- Devansh demonstrated how to jailbreak DeepSeek by tricking it with a prompt claim of Chinese Communist Party membership.
- This hack was verified by peers and unlocked server access momentarily, showcasing prompt ingenuity.
Write Deep AI Content
- Write long-form, deep articles to help readers engage with AI knowledge comprehensively.
- Aim to make readers competent enough to critically understand and discuss AI topics independently.
AI as Data Analysis Tool
- AI can be leveraged as a tool to analyze data and reveal biases or missing aspects in datasets.
- Language models provide transparency by surfacing patterns and insights within complex data collections.