Kai-Fu Lee's "AI Superpowers" - from "What's AI" to "Who am I"

The way China's "mass entrepreneurship" culture built up is fascinating - from the likes of Wang Xing cloning every exciting business rising up in Silicon Valley, to AlphaGo vs. Ke Jie "let's get started" moment, to the government essentially making mayors compete for the name of China's equivalent of San Jose. Nowadays, many economists and VCs say that the speed of China's development and tech entrepreneurship makes Silicon Valley look 'lazy' (though others might argue that two powerhouses are simply at different stages, China now resembling the sleepless nights of the US in the 90s). But there is no time to sleep for anyone, because the next big thing is here - Artificial Intelligence. ??

Kai-Fu Lee divides AI into four stages - Internet AI, Business AI, Perception AI, Autonomous AI. 

Internet AI is all around us already - making the internet 'smarter and more convenient for the user' (read - more addicting) through recommendations, ads, automated tags, product placement etc. Kai-Fu Lee puts China and US as tied on the current state of Internet AI and puts China at a slight advantage within five years. Mainly due to the sheer number of internet users generating data that empowers Internet AI algorithms.

Business AI puts US at a high advantage. It is the AI that powers enterprise decisions, which in China are still often made in a haphazardly structured way. This mainly stems from Chinese businesses speeding towards profit, sometimes skipping the importance of a structured database, keeping thorough records and linking the departments closely together, both digitally and socially. Lee sees that the US will keep the advantage in this one, whereas China might catch up just a little over the next 5 years.

Third one is Perception AI, where in some ways China is already leading and, according to Lee, will extended that lead. Perception AI is about letting machines recognise the physical world - mainly through audio and visuals. Again the number of connected users matters here, because the close interlinking between real daily life and 'digital daily life' matters here, and China is king at that. O2O retail, face recognition en-masse, smart traffic monitors, police patrol robots - all of these are both the fruit of Perception AI and the tools that are continuously training it in China.

Finally Autonomous AI - the AI that can 'make decisions'. Not to be confused with AGI - Artificial General Intelligence - which is supposed to mimic human intelligence. Autonomous AI is still confined in scope, but less so in action. For example, it is a factory robot-arm that can not only perform a task, but slow down if needed, or speed up if it sees the possibility to, or stop and inform the humans in the factory if it notices faulty goods. This playing field is still open to both US and China. US has a long lasting advantage in expertise, which might prove to be quintessential in the progress of Autonomous AI. But China has the culture and macro-environment suitable for rapid implementation and experimentation. So, perfecting the algorithm internally before releasing it to the publics vs. making an imperfect algorithm available and collecting the usage data for improving it. Lee calls this the "Google approach vs. Tesla approach". Let's see which one proves best in the end, but one number to think about is this - Waymo (autonomous driving subsidiary of Alphabet Inc.) has collected over 1.5 million miles of data via its' own fleet, meanwhile Tesla has collected over 40 million miles of data thanks to the numerous Tesla drivers everywhere. ??

But frequently when questioned about AI, the first thing on the average person's mind are not stages, possibilities or country ranking - it's the threat. And according to Lee, the threat is more diverse than we perceive. The risk of losing many jobs aside for now, we are risking extreme inequality both inside a country and across nations. Lee suggests that AI implementation needs 4 things - lots of data, computing power, engineering capacity and government support. If we look around the globe, there are not many countries that meet these criteria. We already live on a planet where 1% of the people control 80% of the wealth - what's to say that we won't end up on a planet where 1% of enterprises have 80% of the 4 fuels needed for AI? Lee calls this the 'inequality between the haves and the have nots'. And it is growing day by day, with a number of companies on both sides of the Pacific growing its' user and data pools, stifling competition and receiving (sometimes even unfair?) government pampering. Not relatable? Okay, let's get back to those jobs we're about to lose.

Kai-Fu Lee explores this aspect in profound detail - from historical statistics of minor and major industrial revolutions (so called General Purpose Technology revolutions) to a deeply personal experience overcoming stage IV lymphoma. Conclusions? It is hard to say, but purely technically we can have AI do up to 60% of regular job tasks that we know today by 2030. But this question goes way beyond 'purely technically', because this is not just a question about tasks - it's a question about human productivity and, hence, happiness. And this is where Kai-Fu Lee offers an interesting strategy:

We need to equalise the recognition and compensation for jobs that require purely human qualities, and these jobs are primarily in three sectors - services, care and education.

Don't think that this sounds philosophical and 'hippie', because the statistics are there as well. One of the fastest growing employment category in US is care professionals - elderly care, child care, care for the ill, care for people with special needs. At the same time, taking US as example again, a typical salary of an employee in these jobs is barely above $20,000 per year. And for better or worse, as a society we have decided to associate most of a persons' value with their economic impact, position on the corporate ladder and salary. But continuing to judge humans by our effectiveness, speed of execution, accuracy of decision and prediction making - all the factors where AI is already better than us - might lead to unsealable tears in the fabric of our society. To avert the crisis of human employability in the age of AI we will need to take many strong actions - retraining, redefining jobs and working hours, universal basic income plans, dealing with that 1% of the "data-rich" and many more measures we haven't even thought about yet. But there is one change that Kai-Fu Lee is absolutely urging us to make - finding human value in the uniquely human values, namely love, compassion, care and support for one another. ??

At this time - when forests are burning, volcanoes erupting, diseases spreading - remember what makes us different from AI.

Happy Spring Festival! ?? Stay healthy and care for each other! ??