The new AI surpasses ChatGPT and changes the future of voice assistants.
Apple is breaking into the future with a new development in the field of artificial intelligence, which can significantly improve the functionality of the voice assistant Siri.
In its research work, the company introduced the ReALM system, which is able to analyze and interpret information from the user's screen and active tasks, making Siri a smarter and more useful assistant.
ReALM considers 3 types of objects:
The ReALM system shows significant improvements over existing systems, especially in the area of recognition of objects displayed on the screen. In tests comparing the ReALM to GPT 3.5 and 4.0, the smallest ReALM model showed results comparable to GPT-4, while the larger models significantly outperformed the latter.
Examples of screenshots that the ReALM model recognized
These results highlight the potential of ReALM as an efficient link resolution system that can run on a device without losing performance. Technology can be the key to the next stages of development of the Apple platform, starting with iOS 18 and the WWDC 2024 conference, which is scheduled for June 10.
Thus, Apple promises not only to improve the interaction of Siri with images in messages and notifications, but also to significantly strengthen the intelligent capabilities of the assistant through the use of the latest developments in the field of artificial intelligence. Such innovations promise to make Siri more intuitive and convenient to use, increasing the practical value of the assistant for users.
Apple is breaking into the future with a new development in the field of artificial intelligence, which can significantly improve the functionality of the voice assistant Siri.
In its research work, the company introduced the ReALM system, which is able to analyze and interpret information from the user's screen and active tasks, making Siri a smarter and more useful assistant.
ReALM considers 3 types of objects:
- Objects displayed on the user's screen;
- Objects associated with the conversation, including data from previous user requests or suggestions from the virtual assistant;
- Background objects that are not directly part of the user's interaction with the device, such as music playing in the background.
The ReALM system shows significant improvements over existing systems, especially in the area of recognition of objects displayed on the screen. In tests comparing the ReALM to GPT 3.5 and 4.0, the smallest ReALM model showed results comparable to GPT-4, while the larger models significantly outperformed the latter.
Examples of screenshots that the ReALM model recognized
These results highlight the potential of ReALM as an efficient link resolution system that can run on a device without losing performance. Technology can be the key to the next stages of development of the Apple platform, starting with iOS 18 and the WWDC 2024 conference, which is scheduled for June 10.
Thus, Apple promises not only to improve the interaction of Siri with images in messages and notifications, but also to significantly strengthen the intelligent capabilities of the assistant through the use of the latest developments in the field of artificial intelligence. Such innovations promise to make Siri more intuitive and convenient to use, increasing the practical value of the assistant for users.
