As a new form of ubiquitous technology, artificial intelligence is driving unprecedented changes in all walks of life. The past few years of intelligent practices and the long term R&D investment at Huawei lead to the launch of the concept of "Inclusive AI" in May last year. The goal is to allow everyone to get what they need at an affordable price point, efficiently make easy use of cutting-edge technologies, and to assure them that their profile remains fully secure throughout the entire lifecycle.
Huawei practices have shown the difficulties met when implementing AI and using it in production in various industries. This has led to some regression in the popularity of AI. However, I think this is just the prelude to large-scale adoption of AI in industry. Problems are inevitable and will arise in AI implementation, but in essence, we use technology to solve these problems. There are currently three types of problems, the bottleneck of hardware, the scarcity of talent, and the gap in knowledge from industry to AI.
Computing power is expensive and requires investment in chips and infrastructure hardware. Large-scale AI deployment requires inclusive, always available computing power.
Talent capable of researching AI data and algorithms is scarce. It is especially difficult to recruit and retain AI talent. Research and production should be treated as two wheels driving the development of the AI framework. The existing AI framework is not sufficient for enterprise production scenarios. Enterprise production usually involves multiple distributed latency-sensitive subsystems. This requires the AI framework to adapt to suit independent or (device-edge-cloud) collaborated training and reasoning. The framework should also allow AI to be easy to use for existing software engineers, just like the software they are using.
Experts are familiar with either AI programming or the industry know-how, but not both, creating a gap in AI deployment. Breakthroughs are required to define the application of AI technology. Industry experts should be able to automatically generate data-based models without having to write one line of code. This gives them the power to enable an industry and maximize its experience.
Another issue brought about by ubiquitous AI is data sovereignty. Humans have managed property in the physical world for thousands of years, but the understanding and management of "digital rights" is just beginning. Data, as a fundamental part of AI, must be respected in compliance with laws and regulations. In my opinion, the most important thing is that those providing services are aware of, withhold and respect clearly defined boundaries for user data.
HUAWEI CLOUD is committed to becoming the "garden bed" of the smart world, providing enterprises with stable, reliable, secure, trustworthy, and sustainable smart cloud services. Huawei's 30 years of technical accumulation and R&D capabilities are now provided as cloud services to bring AI services to every developer, enterprise, and industry.