ThursdAI - The top AI news from the past week

🔥 📅 ThursdAI - Sep 12 - OpenAI's 🍓 is called 01 and is HERE, reflecting on Reflection 70B, Google's new auto podcasts & more AI news from last week

18 snips
Sep 13, 2024
This week, the podcast features Devendra Chaplot, a research scientist at Mistral AI, known for his work on multimodal models, and Steven Johnson, editorial director at Google Labs and author, who contributed to NotebookLM's innovative developments. They dive into OpenAI's new reasoning models, O1 Preview and O1 Mini, and discuss the implications of accountability in AI evaluations. Exciting advancements in audio processing and enhancements in AI capabilities are highlighted, along with the impressive features of NotebookLM that revolutionize research and writing.
Ask episode
AI Snips
Chapters
Books
Transcript
Episode notes
ANECDOTE

Reflection 70B Fiasco

  • Matt Schumers' Reflection 70B model, lauded for beating GPT-4, failed to reproduce results.
  • The incident sparked community discussion on responsibility and trust in model releases.
ADVICE

Republishing News

  • Alex Wolkov advises exercising caution when republishing information about new model releases.
  • He emphasizes the need for verification, especially with limited time.
ANECDOTE

Mistral and Lama Issues

  • Wolfram mentions witnessing releases with issues like tokenizer misconfigurations, citing Mistral and Lama.
  • However, he highlights the importance of reacting responsibly to such mistakes.
Get the Snipd Podcast app to discover more snips from this episode
Get the app