274: AI testing AI? A look at CriticGPT
Aug 20, 2024•15 min
Listen to this episode in Metacast mobile app
Don't just listen to podcasts. Learn from them with transcripts, summaries, and chapters for every episode. Skim, search, and bookmark insights. Learn more
Episode description
In this episode, we speak with Rob Whiteley, CEO of Coder, about OpenAI's recent announcement of CriticGPT, a new AI model that provides critiques of ChatGPT responses in order to help the humans training GPT models better evaluate outputs during reinforcement learning from human feedback (RLFH). According to OpenAI, CriticGPT isn't perfect, but it does help trainers catch more problems than they do on their own.
Key talking points include:
- The downsides of having AI testing the quality of other AI models
- Why it's important to be specific about what types of errors the model is allowed to look for
- Is this another example of rushing into AI?
For the best experience, listen in Metacast app for iOS or Android