Apple AI edits images based on text
Google, Microsoft, and nearly every other tech company delved into the AI craze, but where’s Apple?
Most may not be aware of the iPhone firm’s artificial intelligence projects. That is why CEO Tim Cook confirmed his company is developing AI products and services, and MGIE is one of the latest. It is a program that edits images from text prompts.
For example, you can submit a photo and tell it to increase the brightness. In response, MGIE will boost the image saturation to make it easier to see. Apple admits that its competitors have similar features but claims MGIE can understand shorter text commands. As a result, it might become more user-friendly than DALL-E and Stable Diffusion.
How does the Apple AI work?
Apple’s picture-editing AI tool is MGIE, which stands for “MLLM-Guided Image Editing.” MLLM means multimodal large language models and the “multimodal” aspect means it can process text and images.
VentureBeat said Apple AI uses MLLMs in two ways. First, the program uses MLLMs to get expressive instructions from user input. These are clear and concise enough to guide the editing process.
For example, tell it to “make the sky more blue,” and MGIE will “increase the saturation of the sky region by 20%.”
Second, it uses MLLMs to generate visual imagination, which captures the essence of the edit and can be used to guide pixel-level adjustments. In other words, it turns your commands into manipulations the program should make.
Apple developed the AI with University of California researchers, but they admitted that it is not the first program to have such features. For example, you can do the same with Stable Diffusion and DALL-E.
The Apple AI tool sets itself apart because it can understand short user commands. “Human instructions are sometimes too brief for current methods to capture and follow,” its research paper stated.
READ: Mercedes-Benz cars allow ChatGPT voice commands
Often, you must provide numerous details to get exactly what you want from these image-generating programs. On the other hand, Apple’s MLLM seems to understand brief commands clearly.
The tool can also add objects based on user descriptions. For example, the researchers told it to make a picture of a pizza healthier, so MGIE added vegetables.
MGIE proves that Apple wants to compete in the AI market. However, it hasn’t confirmed offering future products and services with this technology. You may try MGIE on this Hugging Face webpage and read more about it on arXiv.
What are Apple’s other AI projects?
Apple is testing a ChatGPT-like AI chatbot https://t.co/HaeFWwNhee
— TechCrunch (@TechCrunch) July 19, 2023
Apple is building AI-powered mental health services with the Quartz app. It will “keep users motivated to exercise, improve eating habits, and help them sleep better.”
You may read about it in my other article. Also, Apple insider Mark Gurman said the iPad company is developing an AI chatbot to rival ChatGPT.
It has two parts: Ajax and AppleGPT. The former is “a platform for creating large language models that enable AI chatbots.” It is similar to OpenAI’s GPT-4 large language model.
READ: Filipino student creates myth-inspired AI costumes
The latter is the interface that lets people use Ajax. Gurman said AppleGPT resembles competitors too closely to over anything groundbreaking.
Apple may have limited the tool because it focuses on improving privacy when using generative AI. The work includes trying to address potential privacy concerns related to the technology,” he explained.
Mark Gurman also claimed nobody can use the generative AI tool’s output. “Any output from it can’t be used to develop features bound for customers.” You may read it in my previous report. Moreover, check the latest digital trends at Inquirer Tech.
