The 2024 AI Leaders Summit was hosted by Alibaba Cloud in Shenzhen on June 6, 2024. Dr. Liu Yi, founder of Raisound and special expert of the National Key Research and Development Program, attended the summit. At the main forum, he discussed "How large models are reshaping intelligent terminal experiences" with many industry guests. He was also invited to give a keynote speech titled "Large models driving innovation in end applications" at a breakout session.

In his speech, Dr. Liu Yi noted that "Devices like computers, phones and watches have now become people's closest information portals. The integration of large models and terminals has become a new demand and direction in the era of large models."

The 2024 Alibaba Cloud AI Leaders Summit in Shenzhen sparked great enthusiasm for practical applications of large models. Representatives from multiple companies shared best practices in applying large models across various fields onsite. Many AI pioneers and enterprise developers gathered to explore how to seamlessly use generative AI services on the cloud, develop a new generation of applications, and modernize existing ones.

Large models are accelerating the intelligent development of end products

With the rapid development of technologies such as the Internet of Things, 5G, and artificial intelligence, terminal intelligence is developing in the direction of multi-technology integration, among which large AI models have brought new opportunities for the development of terminal intelligence. End-side large language models, as the built-in core intelligence capabilities of intelligent devices, can securely and instantly respond to user needs with personalized service capabilities. At the same time, large models in the cloud industry, thanks to their huge data processing capabilities and algorithm optimization, play an irreplaceable role in service integration across devices and scenarios. The fusion of edge and cloud greatly improves service efficiency and user experience, creating the next-generation natural user interface NUI for the future.

In his keynote speech, Dr. Liu Yi noted that Raisound integrated large models into intelligent wearable and input device solutions. By combining speech recognition, speech generation technologies with large models, users can leverage these smart devices "assistants" to accomplish cross-lingual barrier-free dialogues, automatic medical record entry, and other tasks. The voice interaction latency has been shortened to 100 milliseconds. The introduction of large models has effectively driven upgrades to core natural speech interaction technologies, bringing users an improved experience with faster-updating end products.

Cloud collaboration has enabled new intelligent experiences for terminal devices

The integration of large models and terminals has spawned a new application ecosystem. By combining the generalization capabilities of large models with the specific needs of terminal users, more personalized services can be provided to users.

During the roundtable discussion "Leaders Talk" session, the four heavyweight guests engaged in an in-depth exploration of how large models can shape new experiences for intelligent terminals. Dr. Liu Yi noted that "in cloud-centric scenarios, terminals will take on some AI workload from the cloud based on their own capabilities, where possible. A typical example is the upgrade of intelligent vehicle cabins - we moved the voice capabilities from the cloud to the edge, and also deployed appropriately sized large models to the edge. This allows most interactions to be handled locally on the terminal, with only unsupported requests going to the cloud, greatly improving interaction speeds and stability.

Dr. Liu Yi mentioned at the forum that based on Raisound's independently developed edge intelligence technologies and Alibaba's powerful cloud computing capabilities, the two have jointly built an edge-cloud collaborative platform for terminal intelligence innovation. This platform is designed to provide integrated hardware-software solutions, and they plan to develop a series of edge intelligent products. Currently, Raisound has launched a portfolio of smart terminal products, including smart mice, AI headphones, smart voice watches, and more, applied in smart government, healthcare, and other scenarios to deliver unprecedented intelligent experiences for users.

The full-link architecture of cloud-edge integration has facilitated the practical application of industry solutions

The collaborative development of edge large models and cloud industry large models, especially technologies like AI chips, has rapidly advanced edge human-machine interaction technologies. Together they outline a vision for the future where artificial intelligence becomes more widespread and the experience more human-centric. This also marks the arrival of a new era for human-machine interaction.

As a domestic leader in intelligent voice interaction, Raisound will continue using artificial intelligence voice technology as its core and deeply integrating multi-modal interaction at endpoints. It aims to provide integrated hardware and software artificial intelligence products and solutions for customers, creating new paradigms for human-machine interaction in the era of large models.

Dr. Liu Yi noted that Raisound has now built a "Cloud-Edge" full-link architecture to realize integrated intelligent perception and cognition. This enhances user interaction experience while protecting data privacy and security. Raisound will also continue empowering terminal applications through artificial intelligence to fuel renewal, focusing on high-quality development. This will help accelerate the formation of new productivity and contribute to industry development and practical application.