“AI Superpowers: China, Silicon Valley, and the New World Order,” by Kai-Fu Lee, was published in June. A book that served as a type of primer for themes that I’ve been gradually learning about from a fresh viewpoint since beginning work at Parashift in July 2019.
Kai-Fu Lee is the Chairman and CEO of Sinovation Ventures, as well as the President of Sinovation Ventures’ Artificial Intelligence Institute. Sinovation Ventures was launched in 2009 and is presently on its seventh fund. The team oversees a $2 billion portfolio and works with over 350 technology businesses across China. Prior to that, Kai-Fu Lee served as President of Google China and held several executive positions at Microsoft, SGI, and Apple.
Kai-Fu Lee addresses the driving reasons for the AI ecosystem, how China is aligning its infrastructure to accommodate AI, the direction he believes the AI revolutions will go, and how we must adjust to the new environment in his book.
He also offers a comparative appraisal of progress based on his perspective, which gives him a lot of information about AI research and business applications in China and the United States. He distinguishes between four AI waves:
Internet-based artificial intelligence
Internet AI has most likely reached us all at this point. This sector is mostly concerned with firms such as Google, Facebook, and Amazon, which have access to vast amounts of data and use it to improve the user experience. They mostly use recommendation algorithms to do so, which recognise behaviour patterns based on user behaviour on the services. The services curate customised content to optimise for X based on these insights. User engagement is the most common cause.
Jinri Toutiao, or ByteDance in English, is one of China’s top performers in this category. Founded in 2012, the company is known as “China’s BuzzFeed” because it, too, has created a hub for timely viral content. There are, however, considerable distinctions when compared to BuzzFeed. Unlike BuzzFeed, Toutiao does not have editors on staff. No. The editors at Toutiao are algorithms. They search the internet for material and analyse and analyse articles and videos from partner networks and commissioned contributions using technologies like Natural Language Processing (NLP) and Computer Vision. To put it another way, to comprehend what the content is about. The individual user footprint, which includes clicks, articles read, videos viewed, views, comments, and much more, is then used to tailor a news feed to the user’s specific interests. The algorithms may modify the titles of articles as part of the personalisation process to enhance click rates. The recommendation algorithm improves as more users navigate through the sites and actively “label” data points. As a result, content is becoming increasingly targeted.
China and the United States are tied on a national level. However, a seemingly little point favours China’s eventual pole position: China has more internet users than the United States and Europe combined. Furthermore, China already has a mobile-first infrastructure that allows for seamless internet payments, promoting the development of innovative and economically viable internet applications.
Business Artificial Intelligence
The second wave, known as Business AI, makes use of data that has already been labelled with company-specific labels. This data, which in some cases dates back decades, is extremely helpful for the creation of more accurate prediction models. For example, banks and insurance companies, as well as a number of medical institutions, typically have access to very large data sets containing information such as credit histories, claims and fraud cases, as well as archived diagnoses and health status developments that have been recorded and stored for years.
Machine forecasts are more delicately organised than ours, thus prediction models created on such data sets are particularly helpful. While we humans make predictions based on evident links (so-called hard features), AI-based forecasts also include subtle connections (weak features) that appear to us to be insignificant in the overall picture. The more data the computer can analyse, the more likely it is to uncover more important correlations and so improve the accuracy of the predictions. It’s fairly straightforward.
As a result, as long as there is a large enough data pool with organised, categorised data with relevant outcomes, these technologies can surpass even the best analysts in their analytical jobs. The employment of these technologies for disease diagnostics is an excellent illustration of their added value (interesting results were recently obtained in the case of breast cancer). In this case, AI is used to supplement the expertise of field professionals, improving diagnosis accuracy.
Companies like Palantir and IBM Watson first entered the Big Data field in 2004, offering consultancy services to businesses and governments based on their expertise. These companies were market leaders in their fields for a long time. However, in 2013, new players such as Element AI and the 4th Paradigm emerged as Deep Learning, a particular technique connected with Machine Learning, grew rapidly in both exploration and application.
While US corporations currently have a clear advantage in terms of implementing Business AI quickly and profitably, Kai-Fu Lee believes China will become a leader in the application of AI in public services and in some industries where legacy systems are still in use. China’s very immature financial and healthcare institutions, for example, provide tremendous incentives to challenge how consumer credit and medical care should be planned and delivered. This is where Business AI can help, by transforming weaknesses into strengths through a fundamental bottom-up restructuring of structures and processes.
Perception AI, as the name implies, is about providing machines senses and thereby expanding the context spectrum. As a result, the worlds of digital and physical converge.
Algorithms learn to cluster pixels in photographs and movies into relevant clusterings and distinguish objects in a snapshot in a similar way as humans do. Audio data is similar. In this case, too, algorithms improve their grasp of individual words over time and gradually acquire the meaning of sentences and words in various settings.
The digitalization of our surroundings through sensors and other smart devices – the Internet of Things – is critical for progress in this arena (IoT). So, when you talk to Alexa on Amazon or drive a Tesla, keep this in mind. You, for example, are making a significant contribution to the advancement of such technology.
“Perception AI will bring the comfort and abundance of the internet world to our offline reality,” Kai-Fu Lee claims. Connectors are various sensor-based hardware components. Amazon Go is a technologically amazing example that can already be utilised normally in some regions today.
It should come as no surprise that China is the world leader in perception AI and may yet make significant improvements from here. The Chinese culture and its tolerant attitude toward privacy, as well as Shenzhen’s hardware manufacturing capabilities, provide a competitive advantage in the worldwide market.
The wave of autonomous systems is the most intriguing to me. It is, nevertheless, the one that is the most difficult to measure in terms of its development. The wave builds on earlier AI achievements and strives to create systems that can operate entirely on their own (i.e. without any human interaction). To accomplish this, these systems must be able to not only represent an environment but also react to changes in it and cope with potential deviations and abnormalities. For the bulk of you, autonomous driving will probably be the application that comes closest to your thoughts. Apart from that specific use, autonomous systems will progressively transform many other aspects of our daily life.
According to Kai-Fu Lee, China must be investing a significant amount of money in AI. They also use their political structures to accelerate the implementation of goal-oriented initiatives and obtain a competitive advantage. For instance, regulators and government officials in Zhejiang province have begun planning China’s first intelligent roadway. It contains a variety of sensors, solar panels implanted in the ground, and wireless connections between the automobile, the road, and the driver. The goal is to improve traffic efficiency by 30% and minimise accidents dramatically. Fun fact: In the future, driverless vehicles should be able to charge while travelling on these roads.
Near Beijing, another intriguing project is being created. Xi’an, China. There will be $580 billion invested in infrastructure over the next 20 years. The project’s purpose is to create the world’s first metropolis constructed specifically for autonomous vehicles of any kind. In this context, Baidu, a top AI company, is collaborating with the government to take the project forward as rapidly as feasible.
Nonetheless, the United States led the world in autonomous systems in 2018. Silicon Valley firms, in particular, enjoy significant research and development advantages. Google began testing self-driving cars in 2009, and many of the engineers engaged then went on to found their own companies. This movement did not begin in China until 2016, when businesses like Baidu, Momenta, JingChi, and Pony.ai are particularly powerful and catching up quickly.
Kai-Fu Lee bases the question of who will be the long-term leader in this area on the following: Will the most significant impediment in the broad adoption of such technology be technological or regulatory? If a policy is the deciding element, China will most likely have a significant advantage. Otherwise, the United States. China’s only chance of remaining technologically relevant would be if new and unexpected advancements in computer vision spread quickly over the world, reducing technological disparities. China’s regulatory advantages, on the other hand, could be a trump card.
Is it true that the winner gets everything?
So, an intriguing question now is how and to what extent the dominant forces will be able to claim the expected economic added value for themselves. Because a substantial sum of money is on the line. According to a PwC estimate, AI goods and services alone are anticipated to generate $15.7 trillion in new economic output by 2030 — yes, trillion, not billion.
Because AI has a natural tendency to create monopolies, I find this subject extremely intriguing. This means that in many of the 4 Waves’ industries, there is an economic dynamic in which only a significant winner emerges. Various internet businesses have previously informed us of this. Google leads the search, Facebook social networks, where new players may gradually emerge, and Amazon, which is steadily consolidating its e-commerce domination. For AI companies, the scenario will be no different. A little more radical. As a result, a number of American and Chinese enterprises who are leaders in their fields will be able to create a massive wealth-generating concentration.