← Back to course
1/6
Week 10 Β· Real or Synthetic?

Monday

A cat on the Moon
// Slow down before you share
⏱ about 20 min

Monday: A Cat on the Moon

A new picture card is pinned to the mission control board.

It shows a fluffy orange cat sitting on a gray crater, with Earth glowing behind it.

"A cat on the Moon!" Rocket shouts. "We have to tell everyone on the station!"

Raven leans in and frowns at the card. "What do you notice? There is no caption. Nobody signed it. Where did it come from?"

Nova hovers beside the board. "It looks real," she says. "But looking real and being real are two different things."

Rocket lowers the card slowly. "So how do we find out which one it is?"

"That," Raven says, "is this week's mission."

Synthetic content and deepfakes

In Week 8 you learned that generative AI makes new, synthetic content.

That content can be images, videos, audio or text that imitate the data the model learned from.

Some synthetic content is made to look or sound real.

A deepfake is synthetic audio or video, or a photorealistic image.

Photorealistic means it looks like a real photograph.

The cat card is invented for our story. No real cat has sat on the Moon.

Kind of contentExample (all invented and harmless)
Photorealistic imageA photo-style picture of a cat on the Moon
Synthetic videoA clip of a dog skating along the rings of Saturn
Synthetic audioA recording of a parrot reciting the planets in order
Synthetic textA news-style paragraph about a comet shaped like a rubber duck
NAME THE KIND
  • Read the question.
  • Tap your answer.
A video clip shows a dog skating on the rings of Saturn. What kind of synthetic content is it?
A recording has a parrot reciting all eight planets. What kind is it?
A picture looks exactly like a real photograph. What is the word for that?
StatementTrue or false?
A deepfake can be audio, video or a photorealistic image.?
If a picture looks real, it must be real.?
Generative AI makes content that imitates the data it learned from.?
Synthetic content can only be text.?
WHY THIS EXERCISEGenerative AI can make images, video, audio and text, and looking real is not proof.
New content made by generative AI is called ____ content.
Synthetic audio or video made to seem real is a deep____.
How many kinds of synthetic content are in the table? Type a number.
WHY THIS EXERCISEThe table lists images, video, audio and text: 4 kinds.
FROM DATA TO DEEPFAKE
  • ?People gather lots of training data.
  • ?A generative model finds patterns in that data.
  • ?The model makes new content that imitates those patterns.
  • ?Someone shares the new content, maybe without a label.
WHY THIS EXERCISEGenerative models learn patterns from training data, then make synthetic content that imitates them.
Remember
Nova is a story character. In real life, an image generator is a tool people build.
It does not know or care whether its picture is true. It can make false things look real.

Good start, explorer. Tomorrow you learn two words for false content and why small changes matter.