2016 was full of ups, downs, and surprises: from continued chaos of the Middle East to Britain’s surprising exit from EU, from the dramatic US Presidential Election to China’s further rise… the world is changing every second. These changes may not match our expectations, but we can see that behind the countless changes, technological innovation has been one of the most powerful driving forces that shaped 2016.
As Deep Learning stepped into the scene, the development of artificial intelligence is pivotal force that cannot be neglected. Some professionals in academia and industry even think artificial intelligence will become one of the most essential technological innovations in years to come. It is roughly estimated that artificial intelligence will substantially influence the following fields: translation, advertising, communication, transportation, imaging… Some farsighted scholars also think that artificial intelligence will change not only our lifestyles, but also the way we understand the human race, intelligence, and the essence of the universe.
In 2016, the world of artificial intelligence hit significant milestones as AI technology spreads: AI became a gaming wizard, a strong Go player, a speech transcription expert, a driver, a translator, an artist, and a weapon… Here, SYNCED sorts out 2016 for you, the eventful year of AI.
No.1 Human-Computer Chess Match: Lee Sedol vs. AlphaGo
Milestone Rate: ★★★★★
Attention Rate: ★★★★★
In March 2016, machines conquered the last land human have in Board games – Go: Developed by Google DeepMind, a computer program named AlphaGo beat an international Go expert in Seoul in a 4-1 series, rising the public’s attention and discussion toward artificial intelligence. After the match, The Korea Baduk Association awarded AlphaGo an “honorary 9 dan” – the highest Go grandmaster rank. After that, AlphaGo’s world ranking increased up to No.2. On July 27th in the same year, Ke Jie, the original world rank No.1 Go professional, with his cumulative point decreases, which later led that AlphaGo’s world rank go up to No.1 for a while.
The development team for AlphaGo is a British artificial intelligence company called DeepMind. After being acquired by Google in 2014, DeepMind grow fast to be one of the core companies in artificial intelligence. Led by the genius researcher Demis Hassabis, DeepMind started to expand and attracted talents from all over the world, along with achieving many important technology breakthrough in many research fields. These fields include excelling in Atari games, speech synthesis, lip reading, design routes for subway system, hybrid computing with neural networks, and more. In addition to research, DeepMind also tried to turn its research outcome into products. In July, Google mentioned that DeepMind’s Deep Learning skills can save Google servers a millions-dollars worth electricity bill. Furthermore, the collaboration between DeepMind and National Health Service(NHS) is seen as a great model in term of using artificial intelligence in medical applications.
After DeepMind conquered Board games, the company switched its focus to the more complicated, real-time strategy game. In November, a collaboration between DeepMind and the famous gaming company Blizzard was announced, opened up StarCraft II to AI and Machine Learning researchers around the world. This time, researchers are trying to challenge artificial intelligence from a human perspective and based on human control speed.
No.2 No More Killers On the Road, Self-Driving Cars are Up
Milestone Rate: ★★★★★
Attention Rate: ★★★★★
In addition to Google Self-Driving Cars, there were many other companies and organizations started to achieve research or application outcomes. These organizations include not only Internet or Technology companies like Baidu, Uber and IBM, but also traditional car manufacturer General Motors and Honda Motor, along with the new generation in car industry Tesla Motors. Moreover, there are many independent research projects related to self-driving cars, covering topics like self-driving cars and the sociology, ethics and law behind them, in prominent academic universities like MIT and Harvard.
At the beginning of 2016, the first driverless bus travelled public roads in Wageningen, Netherlands. This electrical vehicle is named WePod, and is designed by French vehicle manufacturer, EasyMile, and EU transportation plan Citymobil2. WePod can take 6 passengers at once, but it can only travel 200 meters far with limited speed up to 8 miles/hour.
In August of the same year, the first driverless taxi called nuTonomy officially had a road test in Singapore. People who were interested in participant can use smartphones to reserve. The startup developed nuTonomy was from MIT, and now focuses on developing driverless taxi. Even though companies like Google and Volvo had already did road tests with their driverless cars, nuTonomy became the first company that open their driverless cars to the public. By the end of the year, the driverless car project under Alphabet/Google had achieved significant success, and later transformed to an independent company called Waymo to further continue the project.
The fast development in self-driving cars also caused some issues. For example, Tesla reported that there were many accidents related to the self-driving feature called Autopilot in the past one year. One of the accidents even caused people die. Later, Tesla warned drivers not to totally rely on the flawed auto-driving feature. Beside these twists, self-driving technology indeed achieved a huge improvement in 2016. Based on the self-driving routes in China and U.S., we might be able to use driverless cars in less than ten years in our daily life, if the development goes well.
No.3 Is Artificial Intelligence Dangerous? Theories of Threat
Milestone Rate: ★★★★☆
Attention Rate: ★★★★★
After another year, warnings like “artificial intelligence are threats of human beings” from famous figures like Hawking and Max continued to evolve. Especially after AlphaGo beat famous Go professional Lee Seoul in March, these warnings started to become more persuasive. More people think that artificial intelligence might be a threat to humans.
There is another event caught people’s eyes this year. In March 2016, Microsoft launched an artificial intelligence chatbot called Tay on Twitter. The original design idea for Tay is to imitate how a 19-year-old girl talks, with its ability to learn from the conversations it had. However, in not even one day, Tay became a racist who supported Nazi.
Some inappropriate sayings from Tay
In addition to this event, there were similar events like Google’s image recognition program identified blacks as gorilla, Google search system recommended distinct professional or career advertisements to men and women, Dallas police used a bomb robot to kill sniper, news recommendations on social network might affect the U.S. Presidential Election, and more. One the one hand, these events and the issues they caused further spread out the words about AI becoming a threat. On the other hand, some researchers saw that artificial intelligence might cause inequality and threats, regardless from biased algorithms or the selected data. At the same time, these societal events and issues revealed to researchers and organizations the public’s misunderstanding toward artificial intelligence, and the needs to educate the public about the importance of artificial intelligence.
Hence, many organizations and researchers worked on various projects to improve the existing issues. For instance, at the end of September, Google DeepMind collaborated with Microsoft, Amazon, Facebook and IBM to found Partnership on AI, an organization aimed at exploring a better way to develop artificial intelligence and help the public to have correct understanding toward artificial intelligence. Additionally, professionals from academia, technology industry and media industry reported many articles about the same topic. *
Nonetheless, the future associating with artificial intelligence and human beings cannot be investigated by discussions or arguments; it can only be examined by practice. Law or societal issues related artificial intelligence have been and will be discussed by various governments and related organizations. All of these shows that the development in technology will substantially influence our life. As the Chairman of the Council of Economic Advisers in U.S., Jason Furman, said, “We need artificial intelligence for the future.”
No.4 Governments Started to Focus on Artificial Intelligence: Encourage Technology Development, Improve Regulation Methods
Milestone Rate: ★★★★☆
Attention Rate: ★★★★★
Now, is the time for artificial intelligence. In October 2016, the White House released two reports — “Preparing For The Future Of Artificial Intelligence” and “National Artificial Intelligence Research and Development Strategic Plan” — to describe the future directions and considerations for artificial intelligence, like the development plan in the future, and the opportunities and challenges artificial intelligence might bring to the government. VentureBeat, a technology media company headquartered at San Francisco, CA, summarized these two reports into seven key points: 1. Artificial intelligence should be used to benefit humans; 2. Governments should embrace artificial intelligence; 3. Need better regulations for driverless cars and drones; 4. Children need to learn and catch up with the technology development; 5. Use artificial intelligence to complement human workers, instead of replacing; 6. Eliminate the bias in data or avoid using biased data; 7. Consider safety and influence toward the rest of the world.
U.S. President Barack Obama talked about the future of artificial intelligence, self-driving cars, and human, with reporters from Wired.
In December, the White House released another report named “Artificial Intelligence, Automation, and the Economy”, focusing on the influence intelligence and automation technologies might cause the economy and the corresponding potential strategies to solve those problems.
In addition to US government, governments from countries and areas — like Europe, China, Japan, and Singapore — also started to see artificial intelligence as an important tool to achieve strength for the country and maintain competitiveness.
The Premier of the People’s Republic of China, Keqiang Li, often mention that the Chinese government strongly supports the technology development in artificial intelligence and robotics.
Furthermore, many governments and research laboratories paid more attention to management and law regulation related to artificial intelligence: How to resolve the employment problem caused by automation, how to equally distribute the economic worth created by artificial intelligence, how to investigate and follow up with the problems caused by automated devices, and how to prevent artificial intelligence technologies from being abused and further causing societal problems. These are all important topics, and we have not found the complete solutions for them so far.
One of the reports the Council of Economic Advisers submitted to the President in July mentioned that estimatedly 83% of the jobs with wage rate per hour less than 20 U.S. dollars will be automated. In order to resolve or improve this problem, the Council suggested two basic strategic principles: allowing flexibility and experimentations rather than enforcement and limitations; encourage jobs instead of reducing jobs.
Overall, the technology innovation with intelligence and automation technologies are brand new not only all humans, but also all the governments. In the future, the government should be careful and perceptive at the current environment so that the society can embrace this technology innovation in a healthy way.
No.5 Continue to Surpass Human: Microsoft Speech Recognition Skill Reached Pro Level
Milestone Rate: ★★★★☆
Attention Rate: ★★★★☆
Artificial Intelligence has already surpassed human not only in playing Go, but also in other fields. In October, researchers from Microsoft published an article named “Achieving Human Parity in Conversational Speech Recognition.” This report mentioned that their speech recognition system has reached 5.9%, the same or even lower word-error-rate(wer) as a professional recorder. Just one month ago, Microsoft’s system speedily reached the historical 6.3% WER, further challenging all human recorders.
Almost around the same time, the CEO of Smartisan, Yonghao Luo, successfully demonstrated the iFlytek Voice Input Method in their New Product Release Conference.
The development of such speech input method might lead the public to widely use speech assistants one day. In AWS re:invent 2016, Amazon launched various kinds of development tools to help developers to develop their speech assistant Alexa more easily. The Vice President of Alexa department in Amazon, Rohit Prasad introduced that Alexa has already had more than 5000 skills; the Amazon Echo, which has Alexa installed, can already helped users to accomplish many daily life tasks, like scheduling and searching music, just by speech. In addition to Amazon, Google, Microsoft, Samsung and Apple all have their own teams to develop their own speech assistants. Probably after a few years, there will be machines always sitting there and waiting for your commands.
No.6 Machine Translation achieves subversive breakthroughs via integrating Neural Network
Milestone Rate: ★ ★ ★ ★ ☆
Attention Rate: ★ ★ ★ ★ ☆
Since the fall of “Tower of Babel”, left unfinished, language unification has been the universal dream of mankind. Now, artificial intelligence based on neural networks allows us to see light at the end of the tunnel.
In September 2016, Google published a paper on arXiv introducing its Neural Machine Translation System (GNMT), which “uses the most advanced training techniques available today to maximize the quality of machine translation”. Google’s blog also announced that it had installed the system to its Chinese-English machine translation application. Within two months, Google published another paper announcing a further breakthrough: the realization of multilingual translation by neural machines, at the same time actualizing zero-shot translation!
The Model Architecture of GNMT
During the World Internet Conference at Wuzhen in November 2016, Sougou’s CEO Xiaochuan Wang demonstrated the company’s real-time machine translation application. Despite of controversies, this proved that machine translation is becoming a more established trend. Maybe after a few years, we would be able to plug “artificial babel fish” into our ears, and simultaneously understand the huge variety of cultural dialects.
In addition to the translation of text-to-text sequences between language pairs, another noteworthy breakthrough is the “translation” of lips to text (video sequences to text sequences), which is an uncontested transcendence of previous progresses. In November, at Oxford University, Google DeepMind and the Canadian Institute of Advanced Studies (CIFAR) jointly published an important paper describing the use of machine learning technology called LipNet to actualize automatic lip-reading. This will further lip-reading technology to an unprecedented capability, achieving a 93.4% accuracy, easily beating experienced human lip readers.
No. 7 The Start of AI hardware warfare: giants vs. start-up
Milestone Rate: ★ ★ ★ ☆☆
Attention Rate: ★ ★ ★ ★ ☆
As deep learning algorithms become more complex, the data sets used grow in size as well, and these developments demand upscaled hardware. In 2016, building platforms for artificial intelligence has become a major new direction in the development of computational hardware. In addition to chip giants like Intel and NVIDIA making high-profile moves in artificial intelligence, start-up with core technologies are also trying to make pivotal changes (although high potential start-ups are acquired at faster rates). In addition, even companies like Google are trying to make its own move in this area.
Traditional chip manufacturers like NVIDIA uses the combination of GPU with deep learning algorithms to further its own development, helping the company’s stock price to soar and notably the biggest winner in artificial intelligence computing.
Intel, with a bigger market share, naturally did not wait for new markets to be claimed by new comers and sought to catch up through acquiring start-ups. In 2016, Intel acquired a number of artificial intelligence start-ups, including Movidius (computer vision), Nirvana (in-depth learning chip) and so on. By November, intel announced its roadmap for artificial intelligence with Nirvana and FPGA vendor Altera Plus, which was acquired in 2015, and introduced its corporate strategy and planned produce ecosystem in artificial intelligence chips.
AMD is a noteworthy competitor in this field in 2016, as the companies announced its first VEGA GPU architecture-based machine learning chip. At the same time, DSP vendors CEVA, FPGA vendors Xilinx and processor technology vendors like Imagination are also mapping their presence in machine learning.
Gregory Wong, CEO of NVIDIA, giving a speech at GTC Europe 2016
At the same time, Internet giants find new opportunities in the same field. In May 2016, Google released a new custom-designed chip tensor processing unit (TPU / Tensor Processing unit). This chip is specifically tailored made for Google’s open source machine learning framework TensorFlow. Microsoft has also indicated support for FPGA through Project Catapult. In addition, IBM’s progress in neurological morphology has attracted a great deal of attention, and may even herald a new direction for the development of artificial intelligence.
For start-up companies, aside from the newly acquired Nervana, there are also companies like Wave Computing, Kneron, and China’s Cambrian, Shenjian technology developing their chip platforms dedicated to deep learning. In 2016, these start-up companies had performed very well, such as the Cambrian 1 A processor launched by the Institute of Computer Science from the Chinese Academy of Sciences, allegedly the world’s first commercial processor dedicated to deep learning.
No. 8 Stanford Releases One Hundred Year Study on Artificial Intelligence
Milestone Rate: ★ ★ ★ ★☆
Attention Rate: ★ ★ ★ ☆ ☆
In Fall 2014, Stanford University launched a research project on artificial intelligence spanning the entire century. Two years later in early September 2016, the research efforts of giants like Google and Microsoft, well-known universities like Harvard University and MIT, as well as the famous Allen institute of Artificial Intelligence organized by Stanford University jointly published the research results of many experts and scholars in a report called “Artificial Intelligence and Life in 2030: One Hundred Year Study on Artificial Intelligence”. This report includes the present development of Artificial Intelligence audits future impacts on employment, environment, transportation, public safety, healthcare, community involvement, and politics.
The report attracted a great deal of attention, and to a certain extent guides the development of artificial intelligence. The report says…
In addition to the aforementioned report, there are many other research institutions, science and technology media that released their own forecast report. In 2016, Synced published its own series of monthly report named “AI100: the world’s most noteworthy 100 AI companies”. Here, we tracked down one hundred NLP, computer vision, chips and hardware companies the field of artificial intelligence. The report will continue to be updated on a monthly basis, please click to read more.
No. 9 The Artificial Intelligence/ Robot Infiltration of Our Lives
Milestone Rate: ★ ★ ★ ☆☆
Attention Rate: ★ ★ ★ ☆ ☆
Although the current development of artificial intelligence and robots are far different from the sensible, self-aware protagonists of science fiction, they have already (sometimes silently) infiltrated our lives.
Let’s leave aside search engines, recommendation systems, translation applications using intelligent and learning algorithms and talk about the more straight forward things: in the past year, Amazon’s Alexa has become more intelligent—— it has developed nearly 5,000 skills; Google announced in March its new speaker-type voice assistant product Google Home to out compete Amazon’s Echo; in addition, Microsoft’s Xiaona and Xiaobing twin sister are aiding people with real life scenarios in China.
Under software development, we also made tangible progress on the development of intelligent hardwares (namely, robots): autonomous cars are carrying passengers, unmanned aerial vehicles are carrying freights, humanoid robots in Japan are taking up receptionist roles inside enterprises, nursing robots are communicating with elders and autistic children in Europe, not to mention the hard working vacuum robots in family houses… In United States, robots have even joined the team of law enforcement. In the confrontation with a sniper who killed a police, the Dallas police sent in a robot carrying a mob which killed the suspect. This is the first time the US police uses robot. Of course, there has been unmanned aerial military aircraft in battlefields for sometime now.
In addition to daily life applications, scientists are also using more artificial intelligence/ robot in their research. In July 2016, NASA sent a mission update to the Curiosity Rover on Mars, enabling it to auto-select targeted rocks for research. Artificial Intelligence is becoming an aid to scientists looking for exoplanets and extraterrestrial life. As for the analysis of particle data, atmospheric systems, genomics, economic activity, social conditions and other data models, artificial intelligence monopolizes them very naturally. Also in early December 2016, Science publishes the first issue of Science Robotics, directly contesting the field’s academic prosperity.
Artificial Intelligence has not lagged behind in Art. There are now intelligent systems that can compose classical music. Applications like Prisma are prevalent on the internet. AI can also spot Shakespeare collaborators for literary connoisseur and even edit pornographic videos. Of course, don’t forget that some of them even wrote movie scripts! The following video is a sci-fi short film based on the script written by artificial intelligence Jetson:
No. 10 Industry and Academic Crossover: Combined efforts of Universities and Enterprises
Milestone Rate: ★ ★ ☆ ☆☆
Attention Rate: ★ ★ ★ ☆ ☆
The accumulated value of deep learning throughout years is beginning to show. Industries focus on artificial intelligence as their key attention areas. In order to strategically take advantage of this trillion dollar class market, talent and tech capitals are where we invest.
In 2016, we see academic researchers and scientists entering industries. In August, facial recognition company Seeta is founded Shiguang Shan, a senior computer vision expert from the Chinese Academy of Sciences. In October, Russian Salakhutdinov, a machine learning professor from Carnegie Mellon joins Apple, while Canadian professor Joshua Bengio from University of Montréal joins deep learning incubator Element AI. In November, Stanford professor Li Feifei joins Google and Carnegie Mellon professor Xing Bo founded the machine learning platform called Petuum. There are also researchers who say they will maintain their academic titles when entering industries.
While academic talents flock into industry, industry is also producing high-quality academic research utilizing of its own resources. Companies like Google, Microsoft, Facebook, Tencent, Baidu, and so on have their own research institutions dedicated to artificial intelligence. These institutions does not only help companies to enhance their products and applications, but also publish valid research results. In December Russ Salahutdinov, the aforementioned professor now at Apple, announced at NIPS 2016 that Apple will continue to publish its AI results to include a broader academic community. Soon after, Apple released its first paper on artificial intelligence, “Learning from Simulated and Unsupervised Images throughout Adversarial Training”.
The era of seclusion is over. In order to lead in the age of AI, the method is only one: participation.
In addition to the ten focus areas mentioned above, there are also many more aspects we can focus upon, including: realist speech synthesis (WaveNet), the steady advancement of image recognition… in addition to video prediction, reading comprehension, and generation of confrontation networks and other areas that have been progressing rapidly in 2016.
In other forefronts, we also see the development of quantum computing, optical computing, bio-computing, BCI, robotics, and some new cryptography applications.
Now, most people will not be able to live with AI technologies. In 2016, a new era commenced. What’s to come? We are waiting for more.