GPT-4 : LES NOUVELLES INFOS (25 000 mots ? Images ? Accès complet ?)

GPT-4 : LES NOUVELLES INFOS (25 000 mots ? Images ? Accès complet ?)

by Ludo Salenne

Humanities, Social Sciences & Thought AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Ludo Salenne 👥 267K 📅 March 16, 2023 ⏱ 24 min 👁 4K 📄 news review 🧭 2026-08-21
Available in: English (current) Français

Keywords

GPT-4ChatGPT PlusAPImultimodalbar exam

Summary

In this video, Ludo Salenne provides an update on GPT-4, two days after its release. He explains that to access GPT-4 via ChatGPT, users must subscribe to ChatGPT Plus ($20/month), but notes that the model is slower and has usage limits (25 messages per 4 hours). He clarifies that the 25,000-word processing capability is not yet available in the ChatGPT interface; it requires access to the OpenAI API via the Playground, for which users must join a waitlist. Similarly, the multimodal feature (image understanding) is not yet functional in ChatGPT; the model only interprets text in URLs, not the images themselves. He demonstrates this with a chess position example. He then highlights impressive community projects, such as a Chrome extension and a Snake-Pong game built with GPT-4, and proposes to create an iPhone app with GPT-4, asking viewers for feedback. He reports that GPT-4 beats 90% of lawyers on the bar exam, citing a Forbes article, and mentions a company using GPT-4 to automate legal filings. Finally, he compares GPT-3.5 and GPT-4 with a creative prompt, showing GPT-4’s superior instruction-following. The video includes a promotional segment for his training course on automating ChatGPT.

193 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video offers practical, up-to-date information on how to access and use GPT-4, addressing common misconceptions about its features. The creator’s demonstrations, such as the chess image test, effectively illustrate the current limitations. However, the argumentation is largely based on personal experience and anecdotal evidence, lacking systematic testing or references to official documentation. The promotional segment for his training course is clearly separated but still influences the presentation.

Scientific Rigor, Source Quality, Title Accuracy

The creator cites several sources, including the OpenAI waitlist page, the Playground, a Forbes article on GPT-4’s bar exam performance, and a demo video. These are relevant and credible. However, the video does not provide a critical analysis of these sources, and some claims (e.g., the 25,000-word limit) are not fully verified. The title accurately reflects the content, which is a news review rather than an in-depth technical analysis.

152 words

Title / Content Match

The title accurately reflects the content, which covers the latest news about GPT-4, including word limits, image capabilities, and access methods.

Quality & Reliability

6/10

The video provides a timely overview of GPT-4's capabilities and limitations, but relies heavily on anecdotal evidence and personal experiments. The creator clearly distinguishes between confirmed facts and rumors, but the lack of deep technical analysis and the promotional segment for his training course reduce the overall reliability.

Chapters

Cited Sources

  • OpenAI GPT-4 API waitlist — Mentioned as the way to get API access to GPT-4.
  • OpenAI Playground — Shown as the interface to use GPT-4 API for longer prompts.
  • GPT-4 Beats 90% Of Lawyers Trying To Pass The Bar — Cited as evidence of GPT-4's performance on the bar exam.
  • OpenAI GPT-4 Demo — Referenced as the official demo showing GPT-4 capabilities.

Concurring Sources

Dissenting Sources

  • OpenAI GPT-4 announcement — OpenAI's official announcement states that GPT-4 is multimodal, but the video claims this feature is not yet available in ChatGPT, which is a temporary limitation.

External References

Contribution & Novelties

The video provides a timely, practical guide to GPT-4’s current limitations and workarounds, particularly the distinction between ChatGPT and API access. It also highlights community projects and proposes a novel use case (building an app with GPT-4).

Pour aller plus loin :

73 words

Radar Profile

The radar profile shows moderate scores across all dimensions, with a slight peak in quantity of information and a dip in technical level. This reflects a video that is informative but not deeply technical, suitable for a general audience interested in AI news.

Reliability 6/10

💬 Sur les 0 commentaires analysés, aucune tendance n'a pu être dégagée.