Google Gemini Pro est LÀ — Bluffant ! (mais en bien ou en mal ?)

Google Gemini Pro est LÀ — Bluffant ! (mais en bien ou en mal ?)

by Ludo Salenne

Applied Sciences & Engineering AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 Ludo Salenne 👥 267K 📅 February 2, 2024 ⏱ 24 min 👁 11K 📄 expert opinion 🧭 2026-08-21
Available in: English (current) Français

Keywords

Gemini ProGoogle BardChatGPTpromptingAI test

Summary

In this video, Ludo Salenne tests Google’s new Gemini Pro model, available in France on Google Bard. He begins by showing how to access it and then compares it to previous models and to ChatGPT. He notes that while Gemini Pro is a step up from the old LaMDA model, it still lags behind GPT-4 in several tasks. He highlights issues with hallucinations and the importance of prompt structure, demonstrating that the order of information in prompts significantly affects output quality. He proposes a specific prompting method: task, characteristics, then role, delivered in a directive manner. He tests this method on a sales page critique and finds it effective. He also attempts to adapt a complex prompt (Promptor) for Gemini Pro, showing that it struggles with multi-step instructions. He concludes that Gemini Pro is not yet on par with GPT-4 but can be useful with careful prompting. The video includes a promotional segment for his ChatGPT automation course.

158 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides practical, hands-on insights into using Gemini Pro, particularly the importance of prompt structure and the need for a directive approach. The creator demonstrates his points with real tests, showing both successes and failures. The argumentation is coherent: he identifies a problem (hallucinations, poor prompt handling), proposes a solution (task-characteristics-role order), and validates it with examples. However, the evaluation is subjective and based on a limited number of tests, and the comparison with ChatGPT is not systematic. The value lies in the practical tips for users, but the scientific rigor is limited.

Scientific Rigor, Source Quality, Title Accuracy

The video does not cite external scientific sources. The creator relies on his own testing and observations. He mentions that Gemini Pro has 137 billion parameters, but this is not verified. The title accurately reflects the content, and the video is well-structured with clear chapters. The promotional segment for his course is clearly identified. The lack of verifiable sources and the subjective nature of the tests reduce the overall rigor.

179 words

Title / Content Match

The title accurately reflects the content: a test and review of Google Gemini Pro, with a balanced look at its strengths and weaknesses.

Quality & Reliability

6/10

The video is a hands-on test and opinion piece by a content creator, not a scientific study. It provides practical observations and comparisons but lacks rigorous methodology and verifiable data. The creator acknowledges limitations and offers a specific prompting method, but the assessment is subjective and based on limited testing.

Chapters

Cited Sources

Concurring Sources

Dissenting Sources

Contribution & Novelties

The video offers a practical, user-oriented perspective on Gemini Pro, highlighting the importance of prompt structure and providing a specific method (task, characteristics, role) to improve results. It also demonstrates the model’s limitations in handling complex, multi-step prompts compared to ChatGPT. This is useful for practitioners but not groundbreaking for the AI research community.

Pour aller plus loin :

92 words

Radar Profile

The radar profile shows moderate scores across all dimensions, with a slight peak in information quantity and a dip in technical depth. This reflects a practical, user-focused review rather than a deep technical analysis.

Reliability 5/10