Did Google fake their Gemini Video?

1. Gemini is a large multimodal model.

🥈85 00:18

Gemini is a new model released by OpenAI that can process images, text, audio, and more.

• Gemini is no longer limited to language processing only.
• It can generate both images and text as outputs.

2. Gemini's marketing page raises concerns.

🥉78 01:46

The marketing page for Gemini is criticized for being misleading and lacking transparency.

• The page compares Gemini Ultra to GPT 4 using different prompting techniques.
• The comparison is presented without disclosing how the numbers were achieved.

3. The technical report for Gemini is disappointing.

🥈82 05:28

The technical report lacks useful information and is more like a marketing piece.

• The report fails to provide details about the model's architecture and training process.
• It also avoids disclosing the size and compute requirements of the models.

4. The controversy around the Gemini video.

🥈88 10:49

The video showcasing Gemini's capabilities is criticized for being staged and not a live interaction.

• The video uses specific prompts and shows individual frames, giving the impression of a live interaction.
• However, it is revealed that the prompts were pre-determined and the video was edited.