The Chinese government’s Ministry of Industry and Information Technology (MIIT) announced its 3-year action plan for Artificial Intelligence (AI) in 2017.
Over the past year since the announcement, experts have come to realize the plan is ambitious to the point of being scary. The proposals set in the MIIT 3-year plan are massive, far-reaching and comprehensive.
Index of this article
1. Goals of China’s AI ambitions
2. System layout of the 3-year plan
3. Major projects to power China’s AI dreams
a. Intelligent Aerial Vehicles
b. Intelligent Home Products
c. Medical Diagnosis Systems
d. Intelligent Translation Systems
e. Intelligent Video Image Identification
4. Where all this ties up
5. Really Intelligent AI
6. Slideshare presentation on China’s AI action plan
7. China’s 3-year AI plan: A quick video overview
What are the goals of China’s 3-year plan for Artificial Intelligence?
Here are the major goals the AI action plan of China:
- promoting industrialization
- integration of a new generation of artificial intelligence technology
- developing high-end intelligent products
- improving the level of intelligent manufacturing
- improving the public support system
- promoting the development of a new generation of artificial intelligence industry
- building a strong country and a network powerhouse to help transform and upgrade the real economy.
What does all this lead to?
Apparently, China is looking for major breakthroughs whereby it can accelerate industrialization and application deployment.
With this plan, China plans to dominate AI in particular and business and technology in general. But more importantly, it is looking for what it calls ‘Collaborative Innovation’. That means China categorically is looking forward to combine production, education, research, SMEs, industrial ecology and government policy to come out a winner.
The system layout of China’s AI ambitions
Here is a graphical representation of China’s AI blueprint:
What all things are shaping up with China’s AI ambitions?
As one of the end-goals of the 3-year plan, China is looking forward to some genuinely smart products down the line. Here are a few of them.
Intelligent Aerial Vehicles
These vehicles will have to have completely automated cruise control. They must be fully capable of operating in highly complex environments. That means they will have a superior level of intelligence that will smartly avoid obstacles.
And no, they’re not vague targets.
These goals have been more specifically defined. For instance, China is aiming for 360-degree omindirectional perceptional avoidance and a precision of 0.005 degree.
Intelligent Home Products
This is where Internet of Things (IoT) comes into picture. A set of coordinated appliances, all smoothly integrated, will lead to the improvement of the overall quality of life.
These products will have intelligent sensors and should be able to conserve water and power consumption without sacrificing efficiency. Further, they should be safe to use and priced in a way to become articles of mass consumption.
The specifics? Achieve 90% penetration rates for smart TVs by 2020.
Medical Diagnosis Systems
While China’s healthcare spending, as a percentage of GDP, has risen from 3.7% in 1994 to 5.6% in 2013, services are still concentrated mostly in cities. In other words, there’s quite of a bit of stuff that remains to be done.
One of the key areas of China’s AI plans is medical imaging and diagnosis systems. The priority is making things streamlined and standardized. Next, AI must power diagnosis of the most prevalent diseases.
A key milestone, surprisingly, is speeding up the commercialization of medical imaging diagnosis support systems. That’s because the Chinese government both understands and acknowledges that private sector medical services are unavoidable, even in the communist state.
Numbers? False negatives must fall below 1% and detection of common diseases by these intelligent services must remain well over 95%.
Intelligent translation system
The bigger objective is to sharpen machine learning and apply it to translation. Currently, translation services falter beyond a point, in terms of accuracy and practicability.
The ambitious target is to further the application of high-accuracy smart translation systems. It will produce translation solutions that are extremely reliable. These translations services will be accurate in multi-language scenarios.
Target? By 2020, the intelligent translation system will have an accuracy of over 85%.
Intelligent video image identification
China is making giant strides in face recognition and is using a multi-pronged approach for that. Using big data and AI, it is not only improving face recognition to deliver social value but is also trying hard to dissuade dissent, prevent miscreants to stage protests and even punish erring drivers.
As a part of the AI plan by MIIT, China is driving innovations in fields as diverse as biometrics, video understanding and human-video integration. In other words, it is looking toward a future where cross-media convergence is an absolute reality. The 3-year plan in AI targets to achieve even image search, video surveillance, video summarization and improved resolution (read accuracy) from considerable distances.
If you’re wondering what goal it will serve beyond surveillance, the answer is an extremely robust credit and finance systems that will prevent frauds and reward high credit-score individuals.
Here’s the number: China plans to exceed 97% detection accuracy in face recognition by 2020.
Where does all this tie up?
The larger picture is offering policy support and funds, nurture talent, improve business environment, develop cutting-edge technology, establish technological leadership and dominate global business.
China is pursuing a complex network of objectives and understanding of China’s 3-year plan for Artificial Intelligence will be incomplete without understanding the associated goals.
One of the associated goals is “Made in China 2025” (MiC25). The MiC25 is a long-term strategic gameplan that China came out with in 2015, thus building over a period of 10 years. The MiC25 intends to, among other things, improve domestic content of core material by upto 70% by 2025. Many people see parallels between MiC25 and Germany 4.0.
The second associated or intertwined goal is the success of China’s Social Credit System. The Credit System is sometimes considered Orwellian, owing to the huge government surveillance in China, it is bringing and the degree to which it may invade privacy of individuals. The system calculates scores individuals and corporates.
Really intelligent AI
Experts might be seeing China’s massive AI initiatives as a comprehensive threat to the technological dominance that Europe and the USA have traditionally enjoyed. In that narrative, China was always a country that mass produced cheap goods and India was where you outsourced services at dirt cheap rates.
Suddenly, China is trying to change its own image. It is ready to don the mantle of technology leader the world over. Its push, investment and intent speaks volumes about what it intends to do.
The more data you feed your AI, the stronger it becomes.
The more data you feed your AI, the stronger it becomes. And with the millions of devices everywhere, data is hardly a problem for China.
There’s one more thing about AI in cases like face recognition.
Face recognition, to be fully reliable, must not just be fed huge amount of data. It must also be fed diverse data.
So if the Chinese cameras grow smart only by reading Chinese faces (Mongolian race), they may falter when they try reading other races, say Caucasians.
And China isn’t unaware of this. Against its multi-billion dollar investments in the African continent, it is getting one important return. It could be using this opportunity to train its camera AI recognize other races – in this case of Negroid origins.
The New York Times reports that modern AI is good at identifying white males, but performs poorly in recognizing people with color skin, especially dark-skinned women. Error rate of 35% was observed with face recognition of black women.
The option left for the US and European countries is clear: a really smart AI, with algorithms so developed and complex they won’t have to depend on huge data, big data.
If the US or Europe can really pull that off, they could beat China at its own game.
But today, that seems like distant future.
Slideshare presentation on China’s AI action plan
Prefer watching a presentation?