Apple continues to work on artificial intelligence at full speed. Finally, it launched a new series of open-source artificial intelligence models called OpenELM (.
Aiming to offer users the power and functions of artificial intelligence without the need for cloud servers, Apple states that the most important feature of these models is that they can work on the device.
In the meantime, we recently shared with you Bloomberg’s Mark Gurman’s statements and details about AI LLM in iOS 18. Here’s all we know about the new artificial intelligence models.
What is OpenELM?
OpenELM consists of eight models. Four are pre-trained, while others can be customized with specific instructions and inputs. The models are relatively small (270 million to 3 billion parameters) and optimized to run on devices.
Apple uses a layer-based scaling strategy to improve the accuracy and efficiency of the models.
On the other hand, the use of OpenELM models offers many advantages:
- On-device AI processing: With OpenELM models, AI functions can be performed directly on the device without needing cloud servers. This significantly increases both privacy and data security.
- Lower costs: Reduced reliance on cloud servers results in cost savings.
- Faster response times: On-device data processing enables faster response times and a smoother user experience.
- More research and development: OpenELM models are open source, allowing researchers and developers to study and improve them. This accelerates the advancement of AI technology and the development of next-generation applications.
Apple made the following statement about the new language models:
The reproducibility and transparency of large language models are crucial for advancing open research, ensuring the trustworthiness of results, and enabling investigations into data and model biases, as well as potential risks. To this end, we release OpenELM, a state-of-the-art open language model. OpenELM uses a layer-wise scaling strategy to efficiently allocate parameters within each layer of the transformer model, leading to enhanced accuracy. For example, with a parameter budget of approximately one billion parameters, OpenELM exhibits a 2.36% improvement in accuracy compared to OLMo while requiring 2 times fewer pre-training tokens.
Diverging from prior practices that only provide model weights and inference code, and pre-train on private datasets, our release includes the complete framework for training and evaluating the language model on publicly available datasets, including training logs, multiple checkpoints, and pre-training configurations. We also release code to convert models to the MLX library for inference and fine-tuning on Apple devices. This comprehensive release aims to empower and strengthen the open research community, paving the way for future open research endeavors.
Also, if you want to learn more about OpenELM, you can check this detailed report.
Apple’s plans include making OpenELM models available in the iOS 18 operating system. iOS 18 is expected to include significant AI-focused enhancements and run a large language model (LLM) on the device.
OpenELM models are considered an important step for the future of artificial intelligence. With new models, AI is expected to become more private, secure, and accessible.
Apple’s move could be important in spreading artificial intelligence and integrating it into every aspect of daily life.
Featured image credit: Keming Tan / Unsplash
