Our reporter Qu Zhongfang reports from Beijing.
A few days ago, Dell Technologies (NYSE: Dell) held the 2024 Technology ** and Outlook Greater China ** Communication Conference, Dell Technologies Global Chief Technology Officer John Roese told the "China Business News" and other ** reporters that 2023 is the first year of generative AI, and in 2024, generative AI will enter the business application stage from the stage of pure theory. "We're going to see a real shift coming, with generative AI systems coming into production. ”
Our reporter learned that Dell Technologies releases its technology trends and outlook for the new year at the end of each year, and John Roese joined Dell in 2012 to develop the company's forward-looking technology strategy. For technology in 2024**, in addition to the shift from theory to practice in generative AI, John Roese said there are three other key trends to watch in 2024: In the security space, zero trust infrastructure will become a "must-have" across industries;The "edge platform" approach will become more widespread;Quantum computing and generative AI will be inextricably linked.
When asked about the impact of technology** on Dell Technologies' strategic decisions in 2024, John Roese responded that these are both outlooks and in Dell Technologies' portfolio and construction of product technologies, as well as in its ability to deliver to customers. "A vision without execution is just a fantasy, and what Dell Technologies is doing is with execution and vision. ”
The first generation of AI projects may be mature
John Roese noted that in 2024, the conversation about generative AI will shift from theory to practice, from training infrastructure and costs to inference and operational costs, where business leaders will take on more responsibility. Increasingly, the focus will shift from broad experimentation to a top-down strategic focus on picking out a handful of generative AI projects that are truly transformative.
While generative AI has sparked tremendous creativity in how it can transform business and the world, there are very few real-world generative AI activities at scale. As we move into 2024, we will see the first batch of generative AI enterprise projects maturing. John Roese said.
John Roese mentioned his feelings and experiences of the fire of generative AI, he pointed out that at the beginning, seeing that any scene that can use human natural language can use new AI tools such as ChatGPT has stimulated unlimited imagination, and almost every enterprise, organization and every functional department in the world is thinking about how to use generative AI to change the traditional way of working such as sales, finance, human resources, and writing**. Dell Technologies, for example, quickly identified hundreds of projects that could use generative AI. As we approach 2024 and when it comes time to implement generative AI, we quickly realize that each project using generative AI is very large, and building and implementing a generative AI system that is actually in production requires a lot of manpower, material resources, and financial support. But each company's resources are limited, and only a few of them can be done, so in 2024, business leaders, CIOs, and others must make strategic decisions to prioritize projects. Once you choose the right one, it is naturally a good thing to have a positive impact on the business, but if you choose the wrong one, there will be a lot of waste of resources, and you may face new risks in market competition.
In view of the huge demand for computing power for large AI models and generative AI, John Roese believes that quantum computing will solve the problem that generative AI and most large-scale AI have a huge demand for computing resources, quantum computing will bring a huge leap in the capabilities of AI systems, hybrid quantum systems will become the computing basis of modern AI, and AI work will be dispersed in a series of different computing architectures, including quantum processing units.
ai pcLanding is still in its early stages
The reporter observed that since the beginning of this year, with the fire of AI large models and generative AI, AI has penetrated into thousands of industries, and in the PC (personal computer) industry, a new concept - AI PC has also become popular, including Intel, Qualcomm, etc. Chip manufacturers are accelerating the improvement of AI computing power, and PC manufacturers have also laid out.
For AI PC, John Roese pointed out that although this is a new word, in fact, the development of AI PC has a history, taking Dell Technologies as an example, the deployment of AI on the PC side has a long history, through intelligent tuning software to support AI performance tuning, network optimization, audio noise reduction, etc., so as to improve the productivity of the PC. AI PCs in the broad sense are personal computing platforms that are designed to run AI-based applications optimally — such as Microsoft Office Assistant Copilot, Adobe Image Management AI, or intelligent chatbots or some kind of security technology. From the perspective of development and evolution, new workloads require more advanced processing power, and AI PC should have the capabilities of three processors: general-purpose computing CPU, graphics processing GPU, and neural processing unit (NPU), and have some dedicated processing power to optimize the operation of local AI tasks. In John Roese's opinion, AI PCs are coming quickly, and they must come.
Taking Microsoft's AI assistant Copilot as an example, John Roese pointed out that although the computing power demand for Copilot is relatively low for a single user, the total computing power demand will increase to a fairly large scale as the user scale is superimposed. In this case, if the computing power is concentrated in the data center, it will inevitably be expensive and difficult to achieve, and if the distributed strategy is adopted to execute the AI load locally, then the world's billions of PCs will be the ideal computing platform, which can not only effectively protect the data security of users, but also improve the user experience and ensure that the application of AI is more efficient.
Regarding the landing timeline of AI PCs, John Roese believes that 2024 will enter the early stages of AI PC deployment, and by 2025 and 2026, most PCs will have additional AI processing power. Therefore, Dell Technologies not only needs to make dedicated software, but also to make PCs capable of running a variety of AI work requirements, which is not only high-performance, but also low-power, and safe. In the near future, the computing terminal used by users will still be a PC, but it will add additional capabilities to efficiently run AI work locally, and many AI instances will run directly on the PC side in the future, instead of running in the cloud, and AI PCs will become more and more mature and powerful.
Editor: Wu Qing Proofreader: Yan Jingning).