
Human-centred AI
Artificial intelligence (AI) is becoming more capable every day. But how can we make the best use of it? And what should we avoid using AI for? Researchers at the University of Amsterdam are exploring both the possibilities and limits of artificial intelligence.
Can you spot a deepfake video? Here’s what to look out for:
1. Micro-expressions
Watch closely for micro-expressions. Sometimes, facial creases or wrinkles don’t move naturally when someone smiles or frowns.
2. Lighting
The lighting often doesn’t match reality. Pay particular attention to shadows, they’re frequently off.
3. Accessories
Check the small details, like jewellery. You might notice, for example, someone wearing an earring on one side but not the other.
4. Unnatural visuals
Look for inconsistencies like mismatched skin tones along the edge of the face, rarely blinking eyes, unnatural-looking hair, or a head that seems too large for the body.
5. Voice and synchronisation
It may seem real, but a deepfake voice has its imperfections. Emotion and intonation do not come through clearly. The synchronisation between mouth movements and voice is also often inaccurate.
What can humans do that AI cannot?

Why GPT can't think like us
GPT may appear smart, but it doesn’t think like humans. The language model struggles to understand small changes in problems - an area in which humans excel, according to research.
An eight-year-old outsmarts AI
AI may have many capabilities, but an eight-year-old can outthink it. Studies indicate that young children are better at solving logical puzzles than language models such as GPT.
Don't ask AI for election advice
AI may sound smart, but don’t ask it who to vote for. Research from the University of Amsterdam shows that chatbots often give biased and misleading election advice.
Interested? Dive deeper with a degree at the UvA.
- Bachelor's

Artificial Intelligence
- Master's

Artificial Intelligence
- Master's

Cultural Data & AI (Media Studies)

AI Foundation Models








