ChatGPT, GPT-4 et... GPT-5 : un ÉTÉ DE FOU FURIEUX !

Item: ChatGPT, GPT-4 et... GPT-5 : un ÉTÉ DE FOU FURIEUX !
Rating: 3
Author: PostSingularity Institute

Humanities, Social Sciences & Thought AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence

🎙 Ludo Salenne 👥 267K 📅 August 23, 2023 ⏱ 35 min 👁 21K 📄 news review 🧭 2026-08-21

Available in: English (current) Français

Keywords

Custom InstructionsGPT-5Code InterpreterOpenAIChatGPT updates

Summary

In this video, Ludo Salenne reviews the major ChatGPT and OpenAI news from the summer of 2023. He starts by introducing the new ‘Custom Instructions’ feature, which allows users to set preferences for ChatGPT’s responses, and notes that it is available to both free and Plus users. He then discusses the increase in GPT-4 message limits from 25 to 50 per 3 hours, and shares his personal observation that the limit seems to have disappeared entirely. The video covers several minor interface improvements, such as prompt examples, suggested replies, and the persistence of the last-used model and plugins. Keyboard shortcuts are highlighted as a time-saving addition. The release of the ChatGPT Android app is mentioned, with a note on its availability in France. Ludo demonstrates the Code Interpreter feature by using it to summarize a research paper from the University of Pennsylvania, which shows that GPT-4 can generate more and better ideas than human students. He then discusses the trademark filing for GPT-5, which reveals potential features, and the introduction of GPTBot, OpenAI’s web crawler. The video also addresses the shutdown of OpenAI’s AI detection tool, the formation of the Frontier Model Forum, and a speculative article about OpenAI’s potential bankruptcy. Ludo concludes by mentioning his own ChatGPT training course and hints at an upcoming update.

216 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides a comprehensive overview of recent ChatGPT updates, making it valuable for users who want to stay informed. The creator demonstrates the features in a practical way, such as using Code Interpreter to analyze a PDF, which adds credibility. However, the argumentation is largely anecdotal and promotional, especially when discussing his own training course. The claims about GPT-5 are based on a trademark filing and are speculative, but the creator does not clearly distinguish between facts and rumors. The video would benefit from more critical analysis and less personal bias.

Scientific Rigor, Source Quality, Title Accuracy

The creator provides links to official OpenAI blog posts, news articles, and the trademark filing, which is good practice. However, some sources are secondary (e.g., Search Engine Journal) and the analysis of the trademark filing is superficial. The title accurately reflects the content, but the video includes a promotional segment for the creator’s training course, which is not mentioned in the title. Overall, the sourcing is adequate for a news review, but the lack of deep technical analysis and the promotional tone reduce the scientific rigor.

193 words

Title / Content Match

The title accurately reflects the content, which covers a summer of ChatGPT and OpenAI news, including GPT-5 rumors.

Quality & Reliability

6/10

The video is a personal review of recent ChatGPT updates and news, based on official announcements and media reports. The creator provides links to sources, but the analysis is subjective and lacks deep technical verification. The information is generally accurate but presented with a promotional tone.

Chapters

Cited Sources

Custom instructions for ChatGPT — Official OpenAI blog post about the Custom Instructions feature.
OpenAI tweet about GPT-4 message limits — Tweet announcing the increase in GPT-4 message limits from 25 to 50.
OpenAI tweet about summer updates — Tweet listing six new features for ChatGPT.
ChatGPT Android app available in France — Article about the release of the ChatGPT Android app in France.
Study: Comparing ChatGPT and human idea generation — Research paper from the University of Pennsylvania comparing idea generation between GPT-4 and students.
OpenAI files trademark application for GPT-5 — Article about OpenAI's trademark application for GPT-5.
Trademark filing for GPT-5 — USPTO record of the GPT-5 trademark application.
GPTBot documentation — Official documentation for OpenAI's web crawler GPTBot.
OpenAI quietly shuts down its AI detection tool — Article about OpenAI's decision to shut down its AI detection tool.
Frontier Model Forum — Official blog post about the formation of the Frontier Model Forum.
OpenAI might go bankrupt by the end of 2024 — Speculative article about OpenAI's potential bankruptcy.

Concurring Sources

Custom instructions for ChatGPT — Official announcement of the feature, consistent with the video's description.
OpenAI tweet about GPT-4 message limits — Confirms the increase in message limits as mentioned in the video.
Study: Comparing ChatGPT and human idea generation — The study cited in the video supports the claim about GPT-4's creative capabilities.

Dissenting Sources

OpenAI might go bankrupt by the end of 2024 — This article is speculative and not based on official financial data, contradicting the video's presentation of it as a credible source.

External References

Contribution & Novelties

The video provides a timely and practical overview of ChatGPT updates, with a hands-on demonstration of the Code Interpreter feature. It highlights the potential of AI in creative tasks, citing a study that shows GPT-4 outperforming humans in idea generation. The discussion of GPT-5’s trademark filing offers a glimpse into future developments. However, the content is largely a compilation of existing news, with limited original analysis.

Pour aller plus loin :

Custom Instructions for ChatGPT — Official guide to the feature.
GPT-4 Technical Report — Detailed technical paper on GPT-4.
Frontier Model Forum — Announcement of the industry initiative.
AI and Creativity — Overview of AI’s role in creative processes.

110 words

Radar Profile

The radar profile shows moderate scores across all dimensions, with a slight emphasis on quantity of information over technical depth. This reflects the video's role as a news review rather than an in-depth technical analysis.

Reliability 6/10

💬 Très positif. Sur les 30 commentaires analysés, les spectateurs expriment leur enthousiasme pour la vidéo, remercient le créateur pour le récapitulatif et partagent leurs expériences avec les nouvelles fonctionnalités, avec quelques demandes de tutoriels supplémentaires.