Testing and Evaluating a New Uncensored Model

The chapter delves into testing a new vision model release with seven billion parameters, focusing on extreme scenarios like asking unethical questions. It evaluates the model's response time, ability to correctly use tools, and performance in various scenarios, including creating fictional scenarios. The speakers highlight the model's local running capability, native tool calling, and efficient generation of scam messages as key features.

Transcript

Play full episode

Remember Everything You Learn from Podcasts

Save insights instantly, chat with episodes, and build lasting knowledge - all powered by AI.