AI is the hype, and pretty much all tech companies are either experimenting on it, or use it to empower their products.
Apple, considered one of the largest tech companies of them all, also use AI extensively. But unlike most others, the company doesn't really use the term that much.
Besides being its marketing strategy, Apple also lacks using the word to focus more on its products' capabilities, and trying to avoid hype and the rivals tend to generate. After all, as a company known for its sophisticated simplicity, Apple is also known for emphasizing human-centered design that focuses on experience and making technology work seamlessly.
So here, highlighting "AI" may be able to shift the focus towards the technology itself, and not to the human at the center, or the experience itself.
But that doesn't mean the company is refraining itself from ever creating a generative AI.
Generative AI is a term to describe a class of AI algorithms that are capable of generating new content. These algorithms are trained on vast amounts of data and learn to generate new content by identifying patterns and structures within the data.
Popularized by the likes of OpenAI when it first launched ChatGPT, others followed suit, and scrambled to come up with their own solutions.
Generative AI is like having a magician in the digital realm, because it can create something out of thin air, capable of bring users' imagination to life in ways they never thought possible.
This time, Apple introduces its first-ever generative AI product.
Calling it the MGIE, or short for 'MLLM-Guided Image Editing),' allows users to describe the changes they want to make to a photo in everyday language.
The tool, developed in collaboration with the University of California, Santa Barbara, can perform a range of tasks, ranging from simple edits, like cropping and resizing to more complex ones, such as altering the shape or brightness of specific objects in an image.
In addition to those, it can also edit specific areas of a photo and can, for instance, modify the hair, eyes and clothes of a person in it, or remove elements in the background.
MGIE can even understand ambiguous text prompts, and interpret them into more detailed and clear instructions the photo editor itself can follow. For instance, if a user wants to edit a photo of a pepperoni pizza to "make it more healthy," the tool can interpret it as "add vegetable toppings" and edit the photo as such.
To do this, MGIE uses multimodal language models to interpret user prompts and then ‘imagines’ what the edit would look like.
Once the user inputs the desired changes, the model automatically applies them to the image.
In other words, the tool has the ability to edit photos based on the text the user types in.
As an open-sourced project, Apple has made the MGIE available for download on GitHub.
The company also released a web demo on Hugging Face Spaces.
Apple's MGIE is far from being the first tool that can do what it does.
Meta has one, and so does Microsoft, OpenAI and others.
But according to the researchers, " [...] human instructions are sometimes too brief for current methods to capture and follow," the project's paper (PDF) reads.
Apple's entry to the generative AI space is significant, not because it's unique, but because the company is not known for embracing the AI hype like pretty much everyone else.
Apple uses AI in some of its products, but the company is far from being the top players in the AI game.
Let alone using the word "generative AI," because Apple is known to be shy in even using the word "AI."
But this time, as the company reveals its open-sourced AI model for image editing, the company is showing the world what it's capable, especially when it comes to contribution to the space.
Read: Apple Will Continue To Invest In AI Because It 'Will Shape The Future'