12 Alibaba Experts Share Top Tech Trend Forecast
From quantum computing to self-driving cars and smart speakers, 2017 was a very exciting year for technological advances and developments. The past 12 months saw smart technologies rapidly become an integral part of everyday life, with millions of households across the world buying into home-based smart tech, as well as innovations such as “City Brain”, the software that is helping solve the traffic problems in some of the world’s largest cities.
Alibaba Hangzhou Campus
But what can we expect in 2018? Will the technology revolution keep going? What new ingenious ways will AI technology further human development? What are the big techs we should look out for in the coming year? We asked 12 of Alibaba’s top tech experts and scientists to share their thoughts on new emerging technologies and what sort of changes we might see in the next 12 months.
Head Scientist at Alibaba Cloud Quantum Laboratory, Deputy Director at Zhijiang Laboratory
”Improving language comprehension and diversifying visual recognition abilities will be key challenges in 2018 for AI technology.
As for machine learning, optimized integration between hardware and algorithms will serve as a gateway which could lead to breakthroughs in the effectiveness of deep model-based reasoning.”
In 2018, we are on the edge of exciting developments in quantum computing, which will pose certain challenges to existing applications of conventional computation such as data security.
The first major achievement for quantum computing in 2018 is likely to be quantum supremacy. Many teams involved in superconductor research or even ion trap research are on the edge of announcing the arrival of quantum processors whose capabilities far out match those of conventional computer processors. 2018 could see the first topological qubit. And seeds of new ideas in fields such as superconductors and ion traps may be planted this year.
The entry requirements to get started in quantum software are not high, so the field will continue to flourish. We can expect to see many theses on quantum software published, although many of these will simply re-combine previous findings.
At the same time, conventional computing is ready to put up a fight. Theoretical breakthroughs could raise conventional emulation capabilities exponentially. In turn, this could reset the starting point in the race for dominance. Cryptography is going to be one of the major fronts where quantum and conventional computing will continue to battle it out. After years of experience with post-quantum cryptography, classical cryptography can resist quantum computer attacks without requiring external resources. Small, inexpensive quantum cryptography products may appear this year, in a gambit intended to prise open the close range quantum cryptography market.
VP and Head of the Tech Lab at Ant Financial
“A huge amount of IoT data means that, now more than ever before, the physical world is also represented in the virtual world. Combining data mining with AI technology can help us understand this world with even more accuracy. I see blockchain technology starting to find practical applications within commercial systems. We will also see the emergence of third generation blockchain structure architecture, which will enable data exchange and value exchange.”
If 2017 was the year of AI, then 2018 might be the year of Internet of Things (IoT). User experience has already benefitted greatly from the progress in IoT. The new important issue is going to be how to handle the increasing amounts of data flowing in from the trend in increasing amounts of installed sensors. The data-processing demand created by countless new sensors will speed the ascent of edge computing. Large amounts of distributed small, light, IoT devices will create a never-before-seen challenge to the security management of the entire system.
In 2018, advances in image and video processing, and voice recognition will help machines to understand and interact with humans in more natural ways. Biometric identification methods will continue to replace passwords. As the number of IoT sensors increases, huge amounts of IoT data will project the physical world into virtual networks, integrating data mining and AI technologies, and enabling us to understand ourselves and this world more accurately.
Development of concepts such as zero knowledge proof will improve security and privacy. This in turn solves the paradox between privacy and mutual trust. In addition, when multiple platforms coexist, and where platforms and chains are interconnected, cross-chain exchange of value and data will become a vital function of blockchain technology. The development of core technologies such as consensus mechanisms and decision networks will continue to improve the performance and scale of blockchain systems. Following Bitcoin and Etherium, third generation blockchain structures will emerge.
In conclusion, businesses will continue to watch the development of blockchain technology closely. In fields such as financial services and supply chain management, some blockchain applications will develop from being proofs of concept into practical business systems. More and more traditional industries will rethink their current business models and try to integrate them with blockchain technology.
“Interactions between humans and AI could encounter challenges related to politics, economics, morals, and safety. These challenges include the problem of bias or discrimination on the part of AI technologies. The ROI from AI technology in 2018 may be overestimated, as training of AI talent still needs to be further developed.”
I see two major trends for 2018. First, consumer-oriented AI technologies such as smart speakers and translation machines will continue to enjoy popularity. At the same time we will start seeing mature business solutions to rise to prominence. Business AI technology will move from the peripheral to the core.
On the other side, there are four challenges the AI technology faces. The first challenge is model optimization and improving computer power. As edge intelligence develops, it will have to perform model optimization and improve computing power within a restricted environment. Only by combining model optimization with improved computing power can breakthroughs be achieved.
The second challenge is improving AI decision making. For AI responsible for complex decisions, enhanced learning can partly improve decision making. But as many business decisions are extremely complex, methods are required to allow machines to learn from how business experts make decisions, and iteratively improve machine decision making.
The third challenge is improving communication between humans and AI. This includes not just direct interaction between people and machines, but also how AI and people interact in fields such as politics, economics and security, and how AI-human interaction affects ethics and morality. One challenge is that of AI "discrimination" against certain groups. We need to be able to understand why machine intelligence makes decisions, and know where the responsibility for those decisions lies.
The fourth challenge is reaching ROI goals. Business costs may increase as training for AI talent still needs to be developed. This may mean the ROI for the AI industry is currently overestimated.
Head NLP Scientist, DAMO Academy Machine Intelligence Technology Laboratory
“At the start of 2018, machine reading-comprehension overtook that of humans for the first time. People will become accustomed to being outperformed by machine intelligence in a variety of specific fields, but it will still be a while before machines can match the depth and breadth of human thought.“
In 2018, there will be many improvements in machine translation and information extraction technologies which will greatly enhance the AI linguistic abilities. The extensive use of methods such as transfer learning will speed the progress of machine learning.
With regard to machine translation, linguistic knowledge or database knowledge will be better compiled in machine translation models. As for translation of languages with fewer reference resources, monolingual and comparable resources will be used to improve translation. Information extraction technology will move on from text files to multimedia (text, forms, images etc.) and information extraction within a vertical field.
In early 2018, our machine reading comprehension technology, Exact Match, slightly outperformed humans for the first time. Hitting this milestone has provided researchers with hope and encouragement. However, we are still at the start of a long journey towards the ultimate goal of machines that can understand and think for themselves.
Head Scientist of Speech Technology, DAMO Academy Machine Intelligence Technology Laboratory
“Machines will be able to perceive tone of speech, body language, facial expressions, and other types of human expression.”
At the beginning of 2018, the interactive methods between humans and machines will begin to move beyond interactive interfaces and begin to more closely resemble the way humans interact with each other.
This change will be supported by modal technologies that incorporate the audio, visual, touch and even taste senses. Machines will be able to perceive tone of speech, body language, facial expressions, and other types of human expression. This will make machines better able to interpret human intentions.
AliOS Chief Software Architect
“By 2018, the mobile Internet era will officially end. User-hours spent on smartphones may decline for the first time in recent years. As the use of smart AI hardware devices explode, user’s online time will become more fragmented between different platforms. “
In 2018, the age of AI powered hardware will begin. More products that include sensors – microphones, cameras, and screens – will enter the market. Many of these devices will be able to go online, interact and see, meaning a step closer to reaching general AI. These changes will mean many daily tasks will be able to be completed without a phone. The use of smart AI hardware devices will explode.
As a result, the time spent on single terminal devices will drop dramatically. Smartphone users’ active time on their devices will decline for the first time in recent years. In contrast, the number of smart devices, including smart phones, will increase, and user time spent online will become even more fragmented. This will be the end of mobile internet as the dominant technology.
“Self-driving cars will hit the fast lane in 2018, but driverless vehicles still face great challenges ahead.”
In the future, our relationship with our cars will fundamentally change. Humans will no longer be the drivers of vehicles, instead we will become precious cargo, transported to our destinations by self-driving smart cars. Industry trends suggest that more research resources will be invested into optimizing AI understanding and perception of varied traffic environments, including pedestrians and vehicles and their behavior. This information will contribute to advancing level-4 and level-3 self-driving vehicles technologies.
Level 3 autonomous driving is where the car is autonomous for certain driving functions such as acceleration and steering, but the driver is present and ready to assist if required. At level 3 or lower, self-driving vehicles still rely at least partly on human drivers. From an application perspective, level 4 fully autonomous driving (operating a car with no driver) on random streets is still a great challenge. Self-driving vehicles will become more widespread, but they not be fully autonomous just yet.
Nevertheless, in 2018 we will start seeing more self-driving vehicles on the roads. China’s self-driving cars could easily overtake America’s in the near future thanks to strong policy support.
“The digitization of the world through smartphones is an early stage of digitization and virtualization.”
Over the past ten years, people's reliance on mobile phones has become a clear social phenomenon. Smartphones blurred the lines between the offline and online worlds.
In 2018 this trend will continue as smart voice assistants will be popularized and integrated into people's lives, along with smart speakers and IoT devices. Big improvements can be made in AI that can learn to communicate with humans using more natural human language by learning through interaction. The digitization of the world through smartphones is an early stage of digitization and virtualization.
Smart assistants will work as people's "virtual substitutes" in the virtual world, and assist in the completion of repetitive work, allowing humans to focus on more creative tasks. At the same time, natural language processing technology will iteratively learn and improve through mass interaction with users. This will help AI use more natural language and become more accurate in receiving and understanding requests.
This step forward will help to liberate human creativity and productivity. People in the future will grow more used to dialogue with physical devices. It is expected that the frequency of voice interaction with devices will overtake the frequency of touch control interaction with devices within five years. The next ten years are the era of artificial intelligence.
Leading Data Scientist at Ant Financial
“The financial sector will be strongly affected by the advancement of AI."
In 2018, people will pay more attention to the practical applications of AI. People’s priorities will move away from application of AI to chess problems and image identification to solving issues the world is currently facing.
This will mean challenges for many AI companies. But there will be a handful of companies that make it through these difficulties. These successful startups, along with traditional companies that succeed in remodeling their business, may become industry leaders.
The battle to dominate the AI chip industry will become even more hotly contested. For cloud and edge computing, chips that support faster and more energy efficient deep learning and machine learning will be developed. These will surpass all GPU and CPU frameworks. Manufacturers of all scales will have opportunities in this field, but only a handful of winners will emerge to be the key players.
After acquiring huge amounts of data through deep learning, AI technology will expand gradually from deep learning to strengthening its quality of learning. This may mean exploring fields such as small data learning, graphic calculations, explainability and model compression. More machine learning technology will be applied and developed in industry.
In particular, the financial sector will be strongly affected by the advancement of AI. AI will impact everything from risk management to loan and financial management. Data and algorithms will gradually reshape the finance sector.
“AI will enter all kinds of industries, discover all kinds of industry-specific issues, and bring a change to productivity.”
Computer vision is likely to continue being the hottest direction for artificial intelligence research. The advancement of computer vision, and its effect on various industries, may increase competition between humans and AI in the workplace. In the short term, workers employed in areas that are easy to replace with AI may face unemployment or a career change. In the long-term, though, new industries will create new employment opportunities.
Computer vision will continue to be under the spotlight in the medical industry as healthcare work begins to move further in this direction. This will mean the combination of industry experience with practical technologies, which will surely provide a solid foundation upon which to build.
AI will enter all kinds of industries, discover all kinds of industry-specific issues, and bring a change to productivity. Computer vision technology will become gradually more recognized in the security and transport sectors, as well as industrial, agricultural, and environmental sectors. But there is no sign that humans will be permanently replaced by AI, on the contrary AI will improve human living standards.
“The challenge will be to optimize hardware and algorithms to improve inferences from data models.”
In 2018, improving text comprehension by organically combining Big Data methodology and knowledge with linguistic knowledge will become more important. Increasing the variety of images which machines can recognize is an important research direction for machine vision. As for speech recognition, the ability to recognize, and automatically adjust to different accents, dialects, and background interference is important. Looking at machine learning, past research mainly focused on hardware optimization and algorithm/model optimization. But in the future the challenge will be to integrate the optimization of hardware and algorithms in order to improve inferences from data models.
“We will see robots enter the home and change people's lifestyles after years of research and development and software and hardware preparation.”
Improvements in video comprehension and editing technology will enable the entire video industry to take great strides forward. This will mean more accurate personalized search recommendations and standardization and improvement of video generation and in-video purchasing.
Facial scanning and recognition technology will become commonplace in 2018，appears in numerous scenarios, and become an accepted part of daily life. Furthermore, smart visual technology will be in numerous creative ways to bring about a qualitative change to the shopping experience across new retail. Additionally, all major car manufacturers will release a driverless car prototype. Autonomous vehicles will move from the research stage towards the application stage. Finally, we will see robots enter the home and change people's lifestyles after years of research and development and software and hardware preparation.
. . .
First hand, detailed, and in-depth information about Alibaba’s latest technology → Search “Alibaba Tech” on Facebook