The 4 waves of AI and battle for chip supremacy
The 4 waves of AI and battle for chip supremacy
Two experts project how the future of artificial intelligence will play out
Quick summary:
AI is a massive technological leap fueled by significant advances in deep learning
Deep learning has proven to have broad applications across several industries
Most headlines today showcase new implementations rather than new discoveries
Author Kai-Fu Lee projects “the four waves of AI” that will shape our future
Tensions over Taiwan are, in part, a “chip war” to control critical AI-enabling tech
Deep learning, advances in AI & our world today
In a previous edition of Forestview, I wrote about the future use of AI in insurance as envisioned in the book AI 2041 by authors Kai-Fu Lee and Chen Qiufan. Kai-Fu Lee wrote an earlier book titled AI Superpowers that examines both the development of AI (including the massive breakthroughs of deep learning) and the competition between China and the United States for future AI supremacy. According to the author, the history of AI research focused on two different approaches: the “rule-based” approach, which attempted to codify human knowledge into rules, and the “neural networks” approach, which mimics the way the human brain operates. The explosion in computer power and large data sets, along with new technical insights, has proven the efficacy of the neural network approach commonly termed “machine learning” or “deep learning” which is a subset. The last edition of Forestview recounted my exposure to the power of deep learning in the form of examining detailed weather observations and property damages to understand the “pathology of claims” in homeowners insurance.
In deep learning, there’s no data like more data. - Kai-Fu Lee
Kai-Fu shares in AI Superpowers that people are excited about the potential of AI and deep learning because it can be applied to many kinds of everyday problems. The core power of deep learning is its ability to:
recognize a pattern
optimize for a specific outcome
make a decision
Many media headlines over the past several years have focused on breathtaking new capabilities unlocked by AI, but Kai-Fu asserts that these milestones represent incremental improvements and optimization, not fundamentally new discoveries. He states these achievements represent “the application of deep learning’s incredible powers of pattern recognition and prediction to different spheres”. However, in Kai-Fu’s view, these achievements “do not signify rapid progress toward ‘general AI’ or any other similar breakthrough on the level of deep learning”. Kai-Fu cites Andrew Ng’s comparison of Thomas Edision and harnessing the power of electricity to AI today: the mass application of a discovery that revolutionized dozens of industries.
Rather than an age of discovery, Kai-Fu asserts that AI today is characterized as both the age of implementation and the age of data. Currently, in the insurance industry and many others, companies that are taking advantage of AI have a mix of talented entrepreneurs, engineers, and product managers that are bringing AI to a variety of domains, helping to solve problems and save expenses. Successful algorithms depend on three factors: big data, computer power, and smart data engineers and scientists.
In the age of implementation, data is the core. - Kai-Fu Lee
A key competitive advantage in our new AI world is having the most data. Why? Because, as Kai-Fu shares, “once computing power and engineering talent reach a certain threshold, the quantity of data becomes decisive in determining the overall power and accuracy of an algorithm”. The more examples algorithms can learn from, the more accurately it can pick out patterns and make better predictions.
Looking at AI from a geopolitical lens, this has profound implications. If AI is better characterized as being in an age of discovery, then the relative freedoms and advanced institutions of the United States and Europe would be a big competitive advantage. If, however, it is more accurate to view the current AI landscape as an age of data, then China has a distinct advantage over the U.S. and Europe: not only does it have a larger population, it also has fewer privacy restrictions that limit the potential use of data. For example, Kai-Fu cites the use of facial recognition in China and other Asian countries as being more commonly accepted than in the West. As AI becomes more widespread in a vast array of use cases, inequality will rise as labor is replaced. This will harm many developing nations whose main competitive advantage is cheap labor. Kai-Fu concludes that, “China and the United States have already jumped out to an enormous lead over all other countries in artificial intelligence, setting the stage for a new bipolar world order.”
The four waves of AI that will shape our future
To understand the geopolitical implications and how AI will impact people and nations over time, Kai-Fu provides a “four waves” framework for thinking about future developments. The first wave is Internet AI and largely focused on the use of AI as recommendation engines that learn (shape?) personal preferences and then serve up content to satisfy us. (I previously looked at the power of AI-driven recommendations in this edition of Forestview.) These recommendation engines rely on digital data and there is no greater trove of this data than Big Tech firms such as Amazon and Google, Alibaba and WeChat. While there have been many startups that have built successful businesses on Internet AI, most of the benefits have been limited to the tech sector and limited to the digital world.
The second wave of AI according to Kai-Fu Lee is Business AI, which expands into the corporate world and takes advantage of the databases that firms have been keeping and labeling for decades. Business AI looks for hidden patterns and mines human decisions to optimize efficiency. In this realm, AI has often shown to do better than trained human experts, who rely on strong features or a handful of data points with a clear cause-and-effect relationship that are more obvious to make decisions. By contrast, AI can mine weak features or thousands of peripheral data points that may appear unrelated but add predictive power when used across millions of examples. Since AI can optimize algorithms using complex mathematical relationships. it can often do better than humans in industries with large amounts of structured data on meaningful business outcomes, like banking, finance, and insurance. Even in areas such as medicine, AI tools can do better at diagnosing conditions because of its ability to examine millions of images and research papers.
It should be clear that both Internet AI and Business AI are waves that are prominent in our everyday lives today. The next two waves identified by Kai-Fu Lee are more future-oriented. The third wave he describes as Perception AI and is characterized by matching the power of AI with our offline lived environment through the use of IoT digital sensors and smart devices. Starting with our mobile phones but going well beyond, our world is increasingly becoming digitized and the offline world is merging with the online world. No longer is AI limited to interactions through a keyboard or touch screen. Technologies such as voice and facial recognition are created a blended world where it makes less sense to talk about being online or offline as a binary choice. Kai-Fu points out that, unlike the first two waves, Perception AI is a hardware-heavy enterprise requiring a diverse array of sensors to sync up the physical and digital worlds. This requires going beyond traditional software programming where the toolset was relatively fixed. Perception AI requires a powerful and flexible manufacturing ecosystem - areas where China has more experience and a competitive edge according to Kai-Fu Lee.
The fourth and last wave of AI development that Kai-Fu outlines is Autonomous AI. This period is characterized by machines going beyond their ability to learn to train themselves and actively shape the world around them. While this might seem scary, raising concerns about the need to have a human “in the loop” and spark debates about intelligent augmentation vs. true artificial intelligence, Kai-Fu asserts that Autonomous AI will revolutionize much of our daily lives annd often for the better. The example of rising auto insurance claims and the need to go beyond telematics and automated driver assistance systems (ADAS) in today’s vehicles to fully autonomous vehicles which is estimated to reduce traffic deaths by over 90% is one such example.
By giving machines the power of sight, the sense of touch, and the ability to optimize from data, we can dramatically expand the number of tasks they can tackle. - Kai-Fu Lee
What do these future waves of AI development mean from a geopolitical standpoint? This is the main focus of the book AI Superpowers, which examines the rise to date of AI in both the U.S. and China as well as how changes in government policy and cultural attitudes could shape future developments. In Kai-Fu’s view, Perception AI and Automation AI are fundamentally different than Internet AI and Business AI. These newer waves seek to blend much more information about local environments than the earlier waves; as a result, Kai-Fu believes an ecosystem of startups in each country could form to assist with the ground-up data collection and marrying world-class AI expertise with local needs and uses. He states that “projecting global power outward from Silicon Valley via computer code may not be the long-term answer”.
Control over chips that power AI creates tensions
In examining the future development of AI in geopolitical terms as a competition between the United States and China for global influence and supremacy to shape the future world order, Kai-Fu Lee offers many thought-provoking observations and points to think about. His framing of the four waves of AI and the need for more hardware solutions to power both Perception AI and Autonomous AI shines a new light on the recent tension between the U.S. and China over the island of Taiwan. Long in limbo and shrouded in uncertainty, the tension over Taiwan is not only viewed from a lens of the projection of power between China and America; it has profound implications related to AI. A newly released book titled Chip War from professor Chris Miller of Tufts University looks at the historical development of the semiconductor industry and the current advantage that Taiwan has globally. More powerful AI requires more powerful processing power, and the chips that provide that are manufactured in only a few locations worldwide with Taiwan accounting for 41% of the world’s supply, far and away the global leader.
The prevalence of AI in our daily lives is not an overnight success story, although it can seem that way to most of us who have not traced its history and background over the last century, including many setbacks over the last 50 years. But today AI is here, it is real, and its impact is profound and grows bigger every day. While organizations race to take advantage and find ways to capitalize on the monumental developments in AI to date, as authors Kai-Fu Lee and Chris Miller highlight, there is much more to come in our future. The impacts that future developments and implementations of AI have will be fascinating to watch, and the tug of war to harness the power of AI and gain an upper hand will continue to be heated both in the corporate and geopolitical realm.
Can you think of examples in your life of the different waves of AI? What is your outlook for future developments? Are you excited or nervous? What shapes your exuberance or caution? Where do you see AI heading within your organization? Is AI a primary focus for your firm?