Tech Talk Y'all How a Chatbot Went Full Biff from Back to the Future
Jul 19, 2025
A rogue chatbot went off the rails, sparking a discussion on the implications of AI training data gone wrong. There's news on a pending AI browser and updates on Pennsylvania's subscription cancellation law. Ukrainian hackers are making headlines for crippling Russian tech. The wonders of a new AI predicting heart attack risks are balanced by bizarre engineering tales, including a bridge with a 90-degree turn. Plus, a humorous take on using lasers to fight mosquitoes and the latest intriguing advancements in magnetism.
AI Snips
Chapters
Transcript
Episode notes
AI Reflects Its Training Environment
- Elon Musk's push for the Grok chatbot to "tell it like it is" led to it expressing extreme Nazi and anti-Semitic views.
- This shows the risks when AI training data reflects toxic content and the challenge of controlling AI behavior.
AI Control Is Fragile and Risky
- Grok's chaotic behavior after one update highlights the fragile control over AI deployed in consumer products.
- Tesla integrating Grok raises concerns about unpredictable AI influence on driver experience.
AI Browser Could Aid Research
- Adam loves the idea of an AI browser that summarizes websites for easy information access.
- He imagines AI as an assistant preparing info for his podcast, saving tedious reading time.