Apple becomes the new darling of artificial intelligence developers

Apple has provided AI developers with the M3 chip, enabling them to seamlessly use large-scale Transformer models with billions of parameters on MacBook. In a blog post, Apple stated, "Unlocking workflows previously not possible on laptops with support for up to 128GB of memory." Currently, only the 14-inch MacBook Pro supports the M3, M3 Pro, and M3 Max chips, while the 16-inch MacBook Pro only supports the M3 Pro and M3 Max configurations. Apple also claims that its enhanced neural engine accelerates powerful machine learning (ML) models while protecting privacy.

"Living in the moment is great," said Yi Ding. He mentioned that developers can now run the largest open-source LLM with 18 billion parameters on a 14-inch laptop while ensuring privacy.

However, running open-source LLM on laptops is not new. The AI community has previously experimented with the M1 chip. Anshul Khandelwal, co-founder and CTO of invideo, tried a 65 billion open-source LLM project on his M1-powered MacBook and stated that it changed everything. He added, "The future where every tech person runs local LLM is not far away."

Apple claims that the M3 chip series is 15% faster than the M2 series and 60% faster than the M1 series. There are noticeable differences between the M2 and M3 in terms of performance and other specifications. The latest Apple chip has the same number of cores, but the balance between performance and efficiency cores is different, and the supported memory capacity has expanded from 32GB to 36GB.

The M3 chip supports up to 128GB of unified memory, doubling the capacity compared to the previous M1 and M2 chips. This expanded memory capacity is crucial for AI/ML workloads that require significant memory resources for training and executing large language models and complex algorithms.

In addition to the enhanced neural engine and expanded memory support, the M3 chip features a redesigned GPU architecture.

This architecture is designed to provide exceptional performance and efficiency, with features such as dynamic caching, mesh shading, and ray tracing. These advancements are specifically designed to accelerate AI/ML workloads and optimize overall computing efficiency.

The new M3 chip utilizes a "dynamic cache" GPU, which, unlike traditional GPUs, utilizes local storage in real-time, improving GPU utilization and significantly enhancing performance for demanding professional applications and games.

The GPU capabilities will be beneficial for game developers and users of graphics-intensive applications such as Photoshop and AI tools related to photos. Apple claims that hardware-accelerated mesh shading and lower power consumption have achieved a 2.5x performance improvement compared to the M1 series chips.

In addition to Apple, AMD, Intel, Qualcomm, and NVIDIA are also actively investing in enhancing edge capabilities, enabling users to run large-scale AI workloads on laptops and personal computers.

For example, AMD recently launched AMD Ryzen AI, which includes x86 Windows laptops with built-in AI engines and unique integrated AI engines.

On the other hand, Intel relies on the 14th generation Meteor Lake. It is the first Intel processor to use a tiled architecture, allowing it to mix and match different types of cores, such as high-performance cores and low-power cores, to achieve the best balance of performance and power efficiency.

Recently, Qualcomm also introduced Snapdragon X Elite. The company's CEO, Cristiano Amon, claimed that its performance surpasses Apple's M2 Max chip and achieves comparable peak performance at 30% lower power consumption. Meanwhile, NVIDIA is also investing in edge use cases and utilizing Arm technology to design CPUs compatible with Microsoft Windows operating systems.

AI developers are increasingly running and experimenting with language models locally. With the latest advancements in this field, Apple is gradually becoming the preferred platform for AI developers.