TECH - Apple Quietly Unveils New Multimodal AI With A Staggering 30B Parameters: Could It Power Text And Vision Features On iPhones? | Benzinga
Apple Inc. (NASDAQ:AAPL) researchers have made significant progress in the field of artificial intelligence (AI). The iPhone maker published its latest research detailing new techniques for training large language models (LLMs) using text and images, potentially leading to more powerful and flexible AI systems.
What Happened: The new methods, outlined in a research paper titled “MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training,” were quietly released on arxiv.org. The paper details how combining various training data and model architectures can result in top-notch performance across various AI benchmarks.
The researchers found that a diverse dataset incorporating visual and linguistic information was crucial for the MM1 models to excel at tasks such as image captioning, visual question answering, and natural language inference.
Interestingly, researchers found that the image’s resolution also ...