Will images replace words? Until now, pictures of all kinds have been slow and expensive to make. Now, artificial intelligence can make amazing visuals in the flash of an eye.
The AI that can draw anything in any style
Will images replace words? Until now, pictures of all kinds have been slow and expensive to make. Now, artificial intelligence can make amazing visuals in the flash of an eye.
An astronaut rides a horse through space. A panda on a motorcycle pumps the gas. A koala slam dunks a basketball. You're not dreaming. This is reality. Or is it? Welcome to the future of photography.
DALL-E 2 is an impressive new AIA computer programme that has been designed to think. , grabbing attention with its eye-catching creations. Users give it a word prompt like: "a sea otter in the style of Girl with a Pearl Earring". It then generates a remarkably accurate and high-quality image.
Its designers at OpenAI say this is a milestone in technology. "Vision and language are both key parts of human intelligence," says researcher Prafulla Dhariwal. DALL-E 2 connects them, seeing "the world the way humans do". This is a "very important step" toward human-like Artificial General IntelligenceThe hypothetical ability of AI to learn or understand any task that a human can. .
We are a visual species. Half of the human brain is involved in image-processing, taking a few milliseconds to identify what we see. And we now make more pictures than ever before. The world has around 45 billion cameras - at least five for each of us - and more than 3 billion photos are shared online every day.
Illustrator Christoph Niemann says we are all "fluent in the language of pictures" without even knowing it. But as technology makes it easier to create and share photos, are images replacing the language of words?
Humans were making pictures long before writing. A cave painting of a wild pig in Indonesia is at least 45,500 years old. Egyptian hieroglyphsThe ancient Egyptian writing system developed from pictograms to abstract shapes. It is the ancestral writing of most modern scripts, including English, Russian and Arabic. appeared around 5,000 years ago. But literacy did not become common until after the invention of the printing pressThe first movable type printing press was built in Germany by Johannes Gutenberg. The first major printed work was the Gutenberg Bible. in 1439.
Psychologist Piotr Winkielman says "emotive images" have a greater impact on us than words. In his experiments, an image flashing on a screen for only ten milliseconds can alter mood and behaviour. "But people were not swayed by emotional words," however long they stared at them.
However, some say this is why words are worth more than pictures. They involve intellectual work and are less ambiguous and emotive. The printing revolution led to the Enlightenment and giant leaps in science, education and ideas.
Online, "the bulk of information still comes from text and that's unlikely ever to change," says technology expert Paolo Gaudiano. Images catch our attention, inspire or infuriate us. But we need words to clarify, inform and understand.
The expression "the camera never lies" has been around since 1859, shortly after the invention of photographyThe oldest surviving photograph is called View from the Window at Le Gras and was created by French inventor Nicephore Niepce in 1826. . But it's never been true. People have always tampered with pictures, from the Cottingley FairiesFive hoax photos of fairies taken in the 1910s by two cousins living near Bradford, England. The author of Sherlock Holmes, Arthur Conan Doyle believed they were genuine. to AI-generated deepfakesThis technique uses artificial intelligence to change the identity of a person in an image or a video - or to make it appear that a person is saying or doing something that they never said or did. .
DALL-E 2's developers are aware of these downsides. Users will not be able to use it to create violent images or photos of real people. And researcher Mark Chen says it's "exceedingly unlikely" to produce offensive pictures accidentally.
For now, only a few "trusted" users can use DALL-E 2. But the company plans to make it public by the summer. Like the 15th-Century printing press, a powerful technology is about to be unleashed.
Will images replace words?
Yes: This is happening. We are all fluent in the visual language of symbols, pictures and emojis. The Internet is making this language universal, bringing humanity together in a way that words cannot.
No: A picture is not worth a thousand words. Images cannot express the complex ideas in a novel or a news article. And a picture almost always needs a caption to explain what we are seeing.
Or... Written words are images. They are graphical symbols that represent sounds and concepts. Our minds convert these back into images and this is what artificial intelligence is learning to do.
Keywords
AI - A computer programme that has been designed to think.
Girl with a Pearl Earring - An oil painting by the Dutch artist Johannes Vermeer. It is one of the first modern psychological portraits, making the viewer ask who she is and what is she thinking.
Artificial General Intelligence - The hypothetical ability of AI to learn or understand any task that a human can.
Hieroglyphs - The ancient Egyptian writing system developed from pictograms to abstract shapes. It is the ancestral writing of most modern scripts, including English, Russian and Arabic.
Printing press - The first movable type printing press was built in Germany by Johannes Gutenberg. The first major printed work was the Gutenberg Bible.
Photography - The oldest surviving photograph is called View from the Window at Le Gras and was created by French inventor Nicephore Niepce in 1826.
Cottingley Fairies - Five hoax photos of fairies taken in the 1910s by two cousins living near Bradford, England. The author of Sherlock Holmes, Arthur Conan Doyle believed they were genuine.
Deepfakes - This technique uses artificial intelligence to change the identity of a person in an image or a video - or to make it appear that a person is saying or doing something that they never said or did.
The AI that can draw anything in any style
Glossary
AI - A computer programme that has been designed to think.
Girl with a Pearl Earring - An oil painting by the Dutch artist Johannes Vermeer. It is one of the first modern psychological portraits, making the viewer ask who she is and what is she thinking.
Artificial General Intelligence - The hypothetical ability of AI to learn or understand any task that a human can.
Hieroglyphs - The ancient Egyptian writing system developed from pictograms to abstract shapes. It is the ancestral writing of most modern scripts, including English, Russian and Arabic.
Printing press - The first movable type printing press was built in Germany by Johannes Gutenberg. The first major printed work was the Gutenberg Bible.
Photography - The oldest surviving photograph is called View from the Window at Le Gras and was created by French inventor Nicéphore Niépce in 1826.
Cottingley Fairies - Five hoax photos of fairies taken in the 1910s by two cousins living near Bradford, England. The author of Sherlock Holmes, Arthur Conan Doyle believed they were genuine.
Deepfakes - This technique uses artificial intelligence to change the identity of a person in an image or a video — or to make it appear that a person is saying or doing something that they never said or did.