Some Thoughts about AI Industry & Opportunities

AI has become a buzzword across different industries, but it is still a vague concept. Frankly, I don’t know whether I should call AI an industry as its applications may be immersive in various industries. Recently I have been talking to friends working in AI verticals and understanding more opportunities in the AI areas.

AI Industry Structure

A very important segment in AI industry is the Hardware, including AI chips, sensors, servers, etc.

AI hardware manufacturers have benefited a lot from the 1st round of the development of the AI industry, such as well-known Nvdia, AMD and also various booming Chinese AI chip companies. Frankly, there is a lot of pressure in how actual applications and solutions land with AI chips and such compute solutions, because ecosystem is the key to land AI solutions.

Another important segment is the Cloud Service and other fundamental services such as AI capabilities on a cloud-based platform.

Cloud service companies are another beneficiaries from the development of the AI industry. They may even benefit much more than hardware companies, for the support for a wider range of industries and applications. Almost all global tech giants are investing a lot in building cloud computing platforms and data centers, with powerful AI capabilities. Global leading companies - Amazon's AWS, Microsoft and Google; Chinese ones such as Alibaba Cloud, Tencent, Baidu, Huawei and also the new player King Cloud listed on Nasdaq. There expected to be a lot of growth potential in China because of China's IT market size.

Big Companies vs Start-ups: Cooperation? Competition?

Along the infrastructure, there are many additional capabilities of artificial intelligence ontop, such as computer vision, intelligent speech, semantic recognition, machine learning, etc.

Computer vision, for example, is now a market with many unicorns who focus on differentiated segments, scenarios, and applications. But in the long run, this market seems to these unicorns, difficult to identify: on one hand, the government will unify the market standard, and therefore in the future it won’t be like the current situation where there are different systems or protocols in different regions. On the other hand, in the future it is very likely that most of the service solutions will be directly provided by platform players such as Ali Cloud, Tencent or Huawei, rather than the SaaS unicorns today, because platform companies have capabilities and resources as solution providers, and meanwhile some of them are even able to test these capabilities in their actual end products.

Besides image recognition and gesture recognition, medical image recognition in healthcare industry is another popular market, such as X-ray or B-ultrasound. These applications may be provided by 3rd party companies with medical domains, and also very likely by cloud platform companies as part of the cloud services.

AI services such as speech/semantic recognition and translation also will become the fundamental cloud services in the future. I wrote an article about translation device and the ecosystem. With these services provided, more scenarios with specific requirements on domains can be solved with lessening barriers.

When Actually Landing the AI Applications…

Someone summarized that it requires ‘ABCDE’ to land IoT& AI industry applications:

A – Algorithm; B – Big data; C – Compute; D - Domain; E – Ecosystem.

Domain is essential to landing a solution because traditional industries are not easy to disrupt and it requires specific domains as core barriers, such as industry data, experience and expertise, industry protocols.

One thing I have learned in Microsoft by working with different partners across different ecosystems, is that leading technology doesn’t mean core barrier, and thus sometimes something beyond simply innovation, such as industry domain that helps end-to-end integration would be a source of significant unrealized business value for many AI application companies. Especially in industries like healthcare, retail, autonomous driving.

When we review the industry structure of artificial intelligence, for the 1st level Hardware and the 2nd level Cloud Service, most likely there are not many opportunities because the threshold for start-ups is too higher, or because the competition barriers are low. In comparison, there are more opportunities in the 3rd level of the industry application layer in entrepreneurship or business.

Where are the opportunities?

In the mobile and cloud ecosystem, the technical barriers are becoming lower for the start-ups, for the past over 30 years’ the accumulated resources in open source software on cloud platforms with more open APIs. Therefore, most startups can spend more time on delivering what the customers want and build their own intellectual property and core barriers, thanks to the low-code development platforms.

Comparatively, the technical barriers of AI are higher than those of cloud and mobile solutions. For example, the startups focusing on NLP, natural language processing (dictation, speech recognition, machine translation, intent understanding, etc.)have to build an engineering team to develop the product for at least a few years, which is financially and timing demanding for most startups. Alternatively, startups can also build their solutions based on 3rd party platforms such as Microsoft, Google, iFlytek, etc. But the challenge under this circumstance is the startup’s core barrier being closely connected to the 3rd party platform capability because of the dependence upon the platform. It requires the entire industry and ecosystem, especially research institutions and large corporations, to work together on lower the threshold by providing more open APIs and dataset.

Many think that technology innovation is the core differentiation to make the AI startups competitive. But there is also another essential key differentiation in AI applications, the domain understanding of industry. I can always hear from my partners in AI battlefield that it is hard to get customer obsession. Well, the main reason is that most startups are tech-driven, and they want to disrupt an industry or a scenario with their AI application and capability. But it fails most likely in the end because of the lack of understandings of the industry and thus huge difficulties of how to disrupt the industry with solutions which may not work in actual operations. Sometimes when I screen potential partner candidates, I will subconsciously prioritize those with industry sales or BD cofounding team. The bundle with tech capability and domain expertise enables such startups to build their core barrier in AI commercial battlefield.

For companies in the traditional industry, where are the opportunities of digital transformation? And where to start?

Digital transformation undoubtedly a big trend, but it is a multi-dimensional, complex engineering problem.

Usually many business decision makers would start to experiment in leveraging digital tools in production, internal productivity process, which enables the capability to use data to engage in the decision making process.

The core of digital innovation is to acquire domain knowledge from information, use knowledge to quickly integrate resources, rapidly develop products, and quickly fit into the market. A good digital transformation - its essence is to build the digital capabilities, which can assist making decisions in the system, product, market strategies.

It is relatively easy for big corporations to attain digitalization and intelligence; comparatively it takes small and medium-sized enterprises a lot more to get into digital transformation. The premium priority for SMBs is to find the biggest pain point they have, whether it is sales and marketing, financial management, or process optimization.