Challenges | Key Takeaways from Webinar 'AI in China: 2020 and Beyond'

Technology Author: Yingwei Fu, Ivan Platonov Editor: Luke Sheehan Jul 11, 2020 08:00 AM (GMT+8)

EqualOcean held an online event on June 30. Here is what 500 Startups' Stella Zheng and Zhuiyi Technology's Terry Ke think about the pandemic and AI.

WIM Salon X Online: AI in China: 2020 and Beyond

EqualOcean held a live webinar on the topic 'AI in China: 2020 and Beyond' with Terry Ke, Strategy Director of Zhuiyi Technology, and Stella Zheng, head of Mainland China of 500 Startups. We discussed the status quo of AI’s development in China and the challenges ahead of passionate entrepreneurs. AI has become a norm these days. Certainly worth a conversation. Below are the key insights from our webinar.

In this series of follow-up articles, we will walk through selected insights shared by our guest speakers. The topics include:

AI in China: COVID19 and AI

AI in China: Challenges

AI in China: Good Practice

AI in China: Future

EqualOcean: In the 2010s, AI was riding the wave of hype. Many companies in the world tried to deploy AI to optimize their businesses. AI is powerful but is also limited – the core technologies can be troublesome and difficult to apply in real-world scenarios. In the new decade, new challenges are ahead. What are the problems AI companies are currently facing? How do China-based AI companies differ from those of the rest of the world?

Terry: Most people in the field are more realistic about AI’s capability, which is a good thing for us – bypass the peak of inflated expectations and come into a more mature state. As I shared with my clients, AI is just a baby. Regardless of models, algorithms and data, we need to know that there is no one-size-fits-all solution for AI, at least at this moment.

When we come into a new niche or new vertical or scenario to apply AI, we need time to finetune the model to fit the specific situation or tasks best, even to the target audience. So, in my opinion, the primary challenge is the mental process of our mindset to treat AI and to give more patience to the still-new technology.

The second challenge is lacking quality data, which should be well-annotated and ready to use. Data is like the fuel for AI model. Without fuel, the engine cannot sustain the operation. This is also a major challenge that most AI companies face, not only in China but also in the world.

As to the conceptional change, we barely had the concept of deep learning or neural networks a decade ago. At the time, we had much less computational power to run complicated models and people did not know about AI. We have made great progress in the past decade. The young technology is getting advanced.

EqualOcean: Indeed, the second challenge is especially crucial – in China, there is even a data service company that has aggregated over 300,000 teams on board just for data labeling. Other countries have their own pioneers in the space, Southeast Asia also has firms engaging in AI data cleaning jobs, too. As an investor, Stella has been in touch with numerous AI companies, and her insights from an investor’s perspective will be valuable for us.

Stella: I’ve talked to many AI companies and fundraising is the first challenge for them.

AI technology seems to reduce the work done by human labors but, on the contrary, it is considerably labor-intensive. As the data labeling mentioned earlier, it asks for huge human labor costs. The cost of AI is more than people’s expectation and hence, AI companies need to raise massive funds to support R & D.

Despite fundraising, AI companies have numerous obstacles to apply AI and deliver real products for clients. As Terry said that AI is still a baby, the area is merely mature to have some actual case products. A lot of technical professionals used to tell me that they can hardly explain AI to clients as there is few publications or real data to show the result or outcome after utilizing AI technology.

The third will be integrating AI technology with existing systems, which is more complicated than add-ins or plug-ins. Enterprises and the end-users need to work with suppliers to thoroughly understand the function and make efforts to build an environment, especially for AI companies, to do testing. The cost of building a real environment for testing is particularly high for biotech companies and manufacturing enterprises. If the setting fails to meet the requirements, AI can do nothing. It still will take years to realize ‘AI World.’

Other than the challenges mentioned, AI has been making progress – advancing from machine learning to deep learning and neural networks, which are buzzwords in the VC industry now. With different industries’ efforts on developing AI technology, it should be promising in the coming years to have commercialized products.

Regarding the second question, data labeling is cheaper in China than other regions – and hence China’s AI companies can be in advantage in pricing with lower data costs. We have invested in Supahands, a data labeling company located in Malaysia. The intention is to break into the Chinese market, but the price put it in a less favored position.

The full record of WIM Salon X Online | AI in China: 2020 and Beyond