Episode notes
What happens when AI trains on junk content or racially skewed data? Turns out, it starts thinking like clickbait… or worse. Today’s episode breaks down two studies that show how bad data can corrupt AI and how most people don’t even notice.
We’ll talk about:
- What “AI brain rot” is and how junk posts actually make models dumber
- A shocking study where people couldn’t spot racial bias in training data — unless it targeted them
- The new Google Skills platform with 3,000+ AI courses
- A crazy-fast analogue AI chip from China and the tools you need to try this week
Keywords: AI brain rot, AI bias, DeepSeek OCR, Google Skills, ChatGPT Atlas, Claude, AI training data, Codi, YouTube deepfakes, OpenAI
Links:
- Newsletter:
Keywords
ClaudeOpenAIYouTube deepfakesAI training dataChatGPT AtlasGoogle SkillsDeepSeek OCRAI biasAI brain rot