Mr. Zeng Xianrui, a consultant of Asian Business Investment, reminded everyone: **There are risks, and investment needs to be more careful to avoid unnecessary risks.
1. The necessity of AI: focus on creative production work
Microsoft conducted a survey on the necessity of AI applications and found that the pace of work has increased exponentially as data, information, and interactions become more frequent, resulting in a heavy burden on both businesses and employees. After the investigation, Microsoft came to three conclusions:
1) Digital debt is killing our creativity
According to Microsoft's survey, nearly two-thirds (64%) say they struggle to find the time and energy to get the job done. Nearly two-thirds of leaders (60%) have already felt the impact, citing a lack of innovation or breakthrough ideas in their teams as a problem. More and more emails, meetings, notifications, documents, and data are taking up more and more of employees' time and energy, so they can't focus on productive work.
This can be clearly observed from the proportion of time that Microsoft spends on users using various Microsoft 365 apps. Team meetings, team communication, and emails for communication occupy 57% of users' time, while Excel, Word, PowerPoint, and Onenote, which truly embody creativity, only occupy 43% of the time.
The heaviest email users (top 25%) spend $8 per week on email8 hours, the heaviest meeting users (top 25%) spend 7. per week in meetings5 hours. Employees generally agree that the number one disruptor of productivity is inefficient meetings. With the intervention of AI, it is expected to reshape the way of working, solve trivial tasks, and help employees focus on creative work.
Every employee needs to be able to use AI
Microsoft believes that AI will reshape the way work is done, and employees need to learn how to use AI to do their jobs. As of March 2023, the share of U.S. job postings mentioning GPT on LinkedIn has increased by 79% year-over-year. In Microsoft's survey, 82% of leaders said their employees will need new skills to prepare for the development of AI. Analytical judgment, flexibility, and emotional intelligence are the most important skills that leaders believe employees will have when they work with AI.
A general-purpose large model that provides the underlying functionality is expected to increase productivity. Applications such as Microsoft 365 (i.e., Office) and Tencent Meeting are basic software that can be widely used in almost all industries, and embedding AI models is expected to reshape employees' working patterns and improve productivity.
ai+office（wps）：Microsoft Office and Kingsoft WPS have embedded AI models into office product series, which can help users write drafts, revise articles, convert text to each other, and automatically generate PPT. Microsoft has embedded Copilot into Microsoft Team, Word, Outlook, PowerPoint, Excel and other office family buckets to help users deal with the needs of various scenarios.
AI+ Conferences:Automatically record meeting minutes and summarize them, mainly for **, courts and other departments, and some for corporate departments. For example, Tencent's meeting AI assistant can complete a variety of complex tasks such as information extraction, content analysis, and conference management and control through simple and natural instructions, improving the efficiency of meetings and information flow.
2. Necessity of device-side AI: privacy and cost
The AI era requires a hybrid computing architecture that combines cloud, edge, and local. In the traditional PC era, most users can only solve their needs by invoking cloud models through networking, but after the popularization of AI, relying solely on cloud AI will become expensive, complex, and insecure. For some general operations, the use of local large models for calculation is enough to meet the requirements, such as optimization, meeting background blur, data visualization, paragraph summarization, and so on. Hence the need to assign different tasks to the right subjects, and the future of AI will be mixed.
Privacy leakage is a serious problem, and AI localization can effectively protect data privacy. In the scenario of invoking cloud AI, it is inevitable to encounter privacy and data leakage problems, and an AI that fully grasps key information such as user identity, preferences, language style, and address book will bring multi-faceted threats to users, and the local large model that calls the local knowledge base can greatly reduce the privacy leakage risk caused by network communication.
On-device AI can effectively solve network latency problems and can continue to work in offline scenarios. Data transfer will always be limited by the physical distance of the server, and applications like AI chatbots must respond in near real-time for a positive user experience. Processing generative AI models on the device avoids potential latency caused by network or cloud server congestion, while improving reliability by being able to execute queries anytime, anywhere. In scenarios such as lack of network connection and poor network signal, the device-side AI model can still help users complete work tasks offline.
The device-side AI model can enhance the personalized experience. The generative AI model on the device side will be able to customize the model and response based on the user's unique voice patterns, expressions, reactions, usage patterns, environment, and even external data, so as to provide services to local end users in a more reasonable way.
Cloud-based AI models are facing rapidly increasing cost pressures. According to a Reuters report, generative AI search is estimated to increase the cost per query by 10 times or more compared to traditional search methods. At the same time, more and more AI applications and AI users are doubling the pressure on cloud AI access. According to TiriasResearch, the cost of AI infrastructure could exceed $76 billion by 2028, and there is currently no effective business model to pass this cost on to consumers. Therefore, it is necessary to share the cost pressure of cloud AI by developing end-to-end AI capabilities.
Disclaimer: The views are for reference only and do not constitute investment advice, and the operation is at your own risk. (Edited by Zeng Xianrui's team of Asia Business Investment Advisors, qualification number A0240622030007).