Lesson 4 of 7 · 8 min
Photos, videos and artificial intelligence
You can’t spot a fake by eye, and a real photo can be used to lie. What matters is where the picture comes from.
Try it yourself: which photo misleads?
Two posts with a photo each time. You can’t see the photos themselves, and that is the point: what matters is what you know about them. Sometimes one post misleads, sometimes both, sometimes neither.
Pair 1 of 4
Illustrative example
A post with a photo of a flooded town square
This is what it looks like here right now!!! Share!
A post from the town council with a photo
Eastbrook, 14 June, 7.30 am: water in the square by the church. Photo: town council. Up-to-date information is on the council website.
The most common trick: a real photo, a false caption
Deceiving people mostly doesn’t require faking anything. It is enough to take a real photo and write that it is from somewhere else or from another time. So with a picture, don’t only ask “is it real?” but above all “where is it from?”
Image search in a minute
- On a phone: press and hold the picture and choose image search (in the Chrome browser it is the option with Google Lens; other browsers can do something similar).
- On a computer: right-click the picture and choose image search. You can also upload the picture to Google Lens or TinEye.
In the results, look for the earliest appearance. If a photo was published three years ago, it can’t be showing today.
Artificial intelligence: don’t look for six fingers
Not long ago the advice was to look for mistakes: odd hands, nonsense lettering. That is no longer enough. A review of 56 studies with more than 86,000 participants found that people recognise a deepfake in roughly 55% of cases. That is almost the same as tossing a coin.
More reliable than the eye is origin: who published it, where it first appeared and whether anyone else confirms it. Some platforms and programs add a label to pictures saying they were made with AI. It helps when it is there. When it is missing, that proves nothing.
The other side of the coin
When anything can be faked, anything genuine can also be waved away with the words “that’s a fake”. Legal scholars call this the liar’s dividend. Believing nothing is the same mistake as believing everything. In both cases the same thing helps: finding out where the recording comes from.
A voice on the phone
The voice of someone close to you can be imitated too. When someone on the phone presses you about money, hang up and call back on a number you know. The course Don’t Get Scammed covers this in detail.
Check your understanding
Quiz · 0/31. You receive a dramatic photo “from today’s storm round here”. What is the best first step?
2. According to research, how well do people recognise a deepfake just by looking?
3. Under a genuine recording on a TV station’s website someone writes “that’s a deepfake”. What do you do with that?
Found a mistake or inaccuracy in this lesson?Report a mistake
What next
Done? Next lesson: Numbers and “a study proves” (8 min)
The whole course is free, no sign-up needed.