Apple seems to be a latecomer to the large-scale language modeling (LLM) space, but the popular technology company released MM1 to compete with some of the early AI adopters, including: OpenAI And Google.
Apple MM1 features
Apple MM1 was recently introduced in a preprint research paper titled “MM1: Methods, analysis, and insights from multimodal LLM pre-training” by the Apple development team. New AI tools provide a system that leverages AI-generated data and labels to enhance the training of other LLMs. Apple researchers believe this is the way to get cutting-edge results.
More precisely, the study aimed to demonstrate that “a careful combination of different types of training data and model architectures can achieve state-of-the-art performance on a variety of AI benchmarks.”
Additionally, the research team believes that Apple MM1 not only speeds up performance, but also reduces the amount of follow-up prompts required to achieve desired results. Touted features of Apple MM1 include enhanced in-context learning, multiple-picture inference, and enabling few-shot thought chain prompts.
Some of these features make this LLM Chat GPT and google gemini. The researchers also acknowledged that this brings competition, saying that Apple MM1 produces very competitive performance on several benchmarks “while enabling multiple-image inference and few-shot prompts.”
Extensive multimodal pre-training allows Apple MM1 to conveniently perform in-context predictions. MM1 looks like a family of AI models with about 30 billion parameters. This is a tiny fraction of the trillion parameters found in OpenAI's GPT-4 and Claude 3 Opus.
Expanding the AI landscape
In particular, Apple MM1 is one of the many AI models released recently. A number of early applications have been released that compete equally with ChatGPT.
Newly launched applied AI company Cognition Labs makes its first debut AI software engineer He was named Devin a few days ago. Devin has been vetted by several top AI companies and is said to have the ability to perform autonomous coding and fine-tune AI models.
There is also Google Genie It was launched by DeepMind. New AI tools can create playable virtual worlds from a single image. It's clear that the AI landscape is slowly expanding, and we're sure to see more models emerge in the coming months.
