ReALM artificial intelligence from Apple
According to the research article, the system called Apple ReALM is references to visual elements It uses large language models to transform complex references, including understanding (like “this” or “that”), into a pure language modeling problem. This enables ReALM to achieve significant performance gains compared to existing methods.
Still, researchers warn that relying on automatic parsing of screens has its limitations. Addressing more complex visual references, such as distinguishing between multiple images, will likely require the inclusion of computer vision and multimodal techniques.
Source
https://arxiv.org/pdf/2403.20329.pdf
https://appleinsider.com/articles/24/04/01/apple-ai-research-realm-is-smaller-faster-than-gpt-4-when-parsing-contextual-data
This news our mobile application Download using
You can read it whenever you want (even offline):
Source link: https://www.donanimhaber.com/realm-apple-gpt-4-ten-daha-kucuk-ve-daha-hizli-yapay-zeka-gelis–175950