New Open AI Leak! (And Other AI News)

New Open AI Leak! (And Other AI News)

🎙 Matt Wolfe 👥 1.0M 📅 December 15, 2023 ⏱ 27 min 👁 113K 📄 news review 🧭 2026-08-28
Available in: English (current) Français

Keywords

GPT-4.5GeminiMixtralAI newsMidjourney

Summary

In this weekly AI news video, Matt Wolfe covers a wide range of recent developments in the AI industry. He starts with a leaked screenshot suggesting a GPT-4.5 model with multimodal capabilities, though unconfirmed. He then discusses a hypothesis that GPT-4 performs worse in winter due to training data, and the partnership between Axel Springer and OpenAI for journalism. Other topics include the reopening of ChatGPT Plus subscriptions, a demo showing GPT-4 Vision can replicate Google’s Gemini demo, and the launch of NotebookLM, Gemini API, MusicFX, and Imagen 2. The video also covers Midjourney’s alpha website and upcoming V6, Stable Zero123 for 3D generation, Stability AI’s new membership model, Claude for Sheets, Mixtral of Experts, Meta’s Audiobox and smart glasses, Google’s SMERF research, Instagram and Snapchat AI features, and Tesla’s Optimus Gen 2 robot. The video is a comprehensive roundup with practical demonstrations and links to resources.

147 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides high value as a curated summary of AI news, with practical demonstrations and links to resources. The argumentation is generally solid, with the creator clearly distinguishing between confirmed facts, leaks, and speculation. He also provides context, such as the irony of Axel Springer partnering with OpenAI after previously considering a lawsuit. However, some segments are based on unverified information (e.g., the GPT-4.5 leak) and personal experience (e.g., issues with Claude for Sheets), which limits the strength of the argumentation.

Scientific Rigor, Source Quality, Title Accuracy

The video demonstrates good scientific rigor by citing sources for most claims, including links in the description to official announcements and social media posts. The creator also acknowledges uncertainties and provides corrections in the pinned comment. The title accurately reflects the content, focusing on the OpenAI leak and other AI news. The video’s adéquation titre/contenu is strong, as it delivers exactly what the title promises.

162 words

Title / Content Match

The title accurately reflects the content, which focuses on a leaked OpenAI model and other AI news from the week.

Quality & Reliability

7/10

The video is a weekly news roundup, presenting information from various sources with clear attribution. The creator acknowledges uncertainties (e.g., the GPT-4.5 leak) and provides corrections in the pinned comment. However, some claims are based on unverified leaks and social media posts, and the creator's personal experience with certain tools (e.g., Claude for Sheets) is anecdotal.

Chapters

Cited Sources

Concurring Sources

  • Gemini API Release — Google's official announcement of Gemini API, consistent with the video's coverage.
  • Mixtral — Mistral AI's official release of Mixtral, consistent with the video's description.
  • SMERF Introduction — Research page for SMERF, consistent with the video's coverage.

Dissenting Sources

  • GPT-4.5 Leak Details — The leak is unconfirmed and Sam Altman replied 'Nah' when asked if it was real, as noted in the pinned comment.

External References

Contribution & Novelties

The video provides a comprehensive and up-to-date overview of AI news, with practical demonstrations and links to resources. It highlights the rapid pace of AI development and the shift towards more efficient models like Mixtral. The creator’s explanation of Mixtral’s architecture is particularly valuable for understanding the future of LLMs.

Pour aller plus loin :

  • Mixtral of Experts — Official announcement and technical details.
  • SMERF — Research page for SMERF, a novel 3D scene representation.
  • Imagen 2 — Overview of Google’s image generation model.
  • Stable Zero123 — Stability AI’s 3D generation model.
  • NotebookLM — Google’s AI-powered notebook tool.

98 words

Radar Profile

The radar profile shows high scores in quantity of information and technical level, reflecting the video's comprehensive coverage and explanations. The quality and reliability scores are moderate, due to the inclusion of unverified leaks and personal anecdotes. Overall, the video is a valuable resource for staying informed on AI developments.

Reliability 7/10

💬 Très positif. Sur les 30 commentaires analysés, la majorité exprime une gratitude et une admiration pour la qualité et l'organisation des informations, avec des éloges récurrents pour le pull de Matt et l'enthousiasme pour les nouveautés présentées.