Qualcomm, the leader in high-end smartphone processors, is gearing up to unveil its third-generation Snapdragon 8 chip. This announcement is particularly noteworthy given that its current top-of-the-line Snapdragon 8 Gen 2 currently holds the title of the most powerful smartphone chip ever, according to Antutu Benchmark, which has been selected only by the Dimensity 9200+. Taiwanese manufacturer MediaTek.
Details of the new chip, which is expected to be unveiled on Tuesday, were leaked online on Monday. The new chip promises not only improved performance but also a significant leap in on-device AI capabilities, according to MS Poweruser.
In addition to the expected incremental performance improvements—such as 25% faster GPU and 30% faster CPU—the new system-on-a-chip (SoC) has a powerful AI engine that can support high accuracy in local AI models, and local large language models (LLMs) with more than 10 billion parameters. can run. The chip can run the text-to-image model Stable Diffusion with a ControlNet neural network.
Mobile hardware manufacturers are increasingly turning to AI. Google, for example, has improved its Tensor chips, which bypass the need for cloud processing to equip the latest Pixel phones to handle many of the AI features that they implement locally. Similarly, Apple ups its game with the A17 Pro SoC and boosts its AI capabilities.
The advent of AI-enabled chips in consumer devices heralds an era of change in technology. Designed to run complex AI models natively, these advanced processors hint at the future of everything from personalized chatbot assistants and highly realistic gaming experiences to real-time health monitoring and friendly user interfaces.
The promise of faster processing, combined with better personalization and greater privacy, positions these chips as the cornerstone of a future where our devices are not only smart, but tailored to our needs and preferences.
Kent Kersey is the CEO of Invoke AI, a company at the forefront of AI image generation and leveraging the power of distributed computing. He told Decrypt that running AI on a device rather than in the cloud has many advantages.
“Cloud-based computing is a good option for large enterprises, but for individuals who want to create content, on-device AI offers a more affordable, personalized and personal option,” he said. He also emphasized the potential of Apple's presentation by deploying a large integrated memory, for example, he pointed to a future with a powerful local LML assistant instead of a cloud-based Siri.
But while this new chip is powerful, Kersey says Invoke has no plans to bring the app to Android or iOS devices.
“Our commercial solution is typically used for more complex workflow processes,” he said. “While we don't have immediate plans for a native OSS Android app, we are closely watching the development of innovative use patterns and modern technology to support mobile deployments.”
Right now, the go-to for local image generation is Object Drawing on iOS. There is no comparable option for Android users.
Kersey was optimistic about the disruption Qualcomm's new chip will have in the broader technology ecosystem.
“Providing this chip for device AI will improve the capabilities that can be deployed on mobile devices,” he explained, adding that he expects to see app developers incorporate on-demand AI capabilities into their services as hardware becomes more capable.
However, every innovation has its own problems. Kersey pointed out the potential problems with distributing model weight down to the final device, especially with proprietary models like GPT-3 or GPT-4 that are bulky and not open source.
“This management of intellectual property may not work for the ownership model above all,” he said. However, Kersey is quick to point out that open source models like Stable Diffusion or Mistral7B can be game-changers.
As reported by Decrypt, Mistral7B is a lightweight model, trained with 7 billion measurements, that is making waves in the AI community. Decrypt tested its responses against the most powerful competitors like LlaMA and Stable Beluga 2 and beat their results. For context, while the Mistral 7B operates on 7 billion measurements, the GPT-4 boasts an impressive 1.7 trillion measurements.
The Qualcomm Snapdragon 8 Gen 3 is more than a microchip; It is an AI-centric mobile future carrier. With industry leaders like Qualcomm, Google and Apple leading the way, it's only a matter of time before your assistant has all your secrets and is useful for answering questions and performing tasks.