Can AI predict what you look like, knowing only what dog you own?
We asked AI image generator Midjourney what it thought people would look like based purely on the dog they own. This is what it showed us for 20 popular breeds.
Artificial intelligence is revolutionising all aspects of our lives, from early disease diagnoses to fraud detection – to predicting what people look like based on the dog they own.
We gave this last very important task to AI text-to-image generator Midjourney and these were the results.
AI image generators – such as Midjourney, Dall-E, Leonardo and Stable Diffusion – each function in their own way, but essentially allow a user to give a descriptive “prompt” in plain language then produce an image to meet that brief.
They tend to lean into stereotypes as the AI models behind them are trained using large volumes of images that have been scraped from the internet then labelled and categorised by humans.
As a result, the models reflect human biases and stereotypes, and the AI outputs follow suit.
For these images, Midjourney was asked to depict a breed of dog with a typical person who would own that dog.
The results were varied.
How did it do?
HOW TO GET STARTED ON MIDJOURNEY
1. Log in or sign up to Discord
Discord accounts are free, just visit discord.com
You can access the Midjourney Bot anywhere you use Discord.
2. Subscribe to a Midjourney plan
Go to Midjourney.com/account and sign in with your verified Discord account.
Plans range from $US8 to $US120 a month, plus GST.
3. Add the Midjourney server on Discord
Press the “+” at the bottom of the server list on the left-hand sidebar.
Press “Join a Server” and paste or type http://discord.gg/midjourney
4. Go to any General or Newbie channel on the Midjourney server
These are visible in the left sidebar.
5. Use the /imagine command
Within the General or Newbie channel, type the “/imagine” command then a short text description of the image you want to create. Then send your message.
This description is known as a prompt.
6. Accept the terms of service
The Midjourney Bot will generate a pop-up asking you to accept the terms of service.
You must agree to these before any image will be generated.
7. Processes the job
The Midjourney Bot takes about a minute to generate four options.
Midjourney uses powerful Graphics Processing Units (GPUs) to interpret and process each prompt.
When you purchase a subscription to Midjourney, you are purchasing time on these GPUs. Each image generation uses some of your subscription’s GPU time.
Use the /info command – by typing “/info” into the chat window – to check your remaining time.
8. Upscale or create variations
After the initial images have been generated, two rows of buttons appear: U1 U2 U3 U4 and V1 V2 V3 V4.
U buttons upscale an image, generating a larger version of the selected image and adding more details.
V buttons create slight variations of the selected image.
You can also rerun the job using the re-roll/refresh button.
This will rerun the original prompt, producing a completely new set of images.
9. Save your image
Click on the image to open it to full size, then right-click and choose “save image”.
