J'AI ACCÈS À CHATGPT VISION ! (et c'est vraiment dingue)

J'AI ACCÈS À CHATGPT VISION ! (et c'est vraiment dingue)

by Ludo Salenne

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Ludo Salenne 👥 267K 📅 October 12, 2023 ⏱ 25 min 👁 39K 📄 tutorial 🧭 2026-08-21
Available in: English (current) Français

Keywords

ChatGPT Visionimage analysismultimodalGPT-4AI tutorial

Summary

In this video, Ludo Salenne demonstrates the newly available ChatGPT Vision feature, which allows users to share images with ChatGPT and engage in conversations about them. He starts with a fun test: locating Charlie in a ‘Where’s Charlie?’ puzzle, which ChatGPT successfully solves. He then shows practical applications, such as generating a recipe from a photo of a fridge, explaining a brain diagram for a biology student, and solving a visual math puzzle. The video also covers limitations, including the model’s refusal to identify public figures and occasional bugs. Ludo demonstrates how to use ChatGPT Vision to summarize a PowerPoint slide into an email and a blog post, and even to generate HTML/CSS code from a hand-drawn website mockup. He concludes by combining Vision with DALL-E 3 to create a ‘reverse prompt’ workflow, generating new images from an existing one. Throughout, he emphasizes the beta nature of the feature and encourages viewers to experiment.

154 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides a clear, hands-on demonstration of ChatGPT Vision’s capabilities, showing both its strengths and weaknesses. The creator’s argumentation is based on direct experience, which adds credibility. He highlights practical use cases that viewers can relate to, such as summarizing meeting slides or generating code from sketches. However, the video is more of a showcase than a critical analysis; it does not delve into the underlying technology or potential biases. The promotional segment for his training course is clearly separated and does not undermine the overall value of the demonstration.

Scientific Rigor, Source Quality, Title Accuracy

The video is a tutorial based on the creator’s own testing, so the primary source is his direct experience. He mentions an article about ‘Where’s Charlie?’ and provides a link in the description, which is a legitimate external source. The title accurately reflects the content. The creator is transparent about the beta nature of the feature and acknowledges potential errors, which adds to the reliability. However, the video lacks citations to official OpenAI documentation or research papers, limiting its scientific rigor.

187 words

Title / Content Match

The title accurately reflects the content: the creator demonstrates access to ChatGPT Vision and showcases its capabilities.

Quality & Reliability

6/10

The video is a practical demonstration of ChatGPT Vision, showing real examples and limitations. The creator is transparent about beta instability and errors. However, the content is largely anecdotal and promotional, with no in-depth technical analysis or external verification of claims.

Chapters

Cited Sources

Concurring Sources

Dissenting Sources

  • Potential biases in multimodal AI — The video does not address potential biases or ethical concerns in image analysis, which are discussed in academic literature.

Contribution & Novelties

This video offers a practical, early look at ChatGPT Vision, showcasing its potential in everyday tasks such as meal planning, studying, and content creation. It also introduces the concept of ‘reverse prompting’ with DALL-E 3, which is a creative workflow. The creator’s hands-on approach provides immediate insights into the feature’s strengths and limitations.

Pour aller plus loin :

  • Multimodal learning — Relevant for understanding how AI models process multiple types of data.
  • GPT-4 — The underlying model that powers ChatGPT Vision.
  • DALL-E — The image generation model used in the video for reverse prompting.

94 words

Radar Profile

The radar profile shows a balanced performance across all dimensions, with slightly higher scores in information quantity and technical level, reflecting the video's practical demonstrations. The lower reliability score indicates the need for external verification of the claims made.

Reliability 6/10

💬 Très positif. Sur les 30 commentaires analysés, les utilisateurs expriment un fort enthousiasme pour la fonctionnalité et la qualité des démonstrations, avec quelques demandes de précisions techniques et suggestions d'usages supplémentaires.