Module 3: Image, audio and video models · 7 min
What image and video AI cannot do well
- Exact text inside pictures is often wrong or garbled.
- Counting objects and showing hands, legs or fine mechanical parts can fail.
- It has no understanding of physics or truth, only of how things usually look.
- It can copy the style of famous artists, which raises fairness and copyright questions.
- Realistic fake photos, voices and videos (deepfakes) can be used to cheat people.
Good practice: label AI-made images, never present them as real photos of real events, and never make fake content that harms a real person.
← Multimodal, audio and video AI | Prompting, RAG or fine-tuning? →