In recent years, always feel time is too fast, then quickly become a thing of the past 2019, although did not say that travel can be a stay away, but the world is always so much to open your eyes to see what the next will be 2020 What are black & C will soon occupy the position of the times, it will take everyone together to take inventory. This article is based on the following word cloud can be used as reference 2020 keywords.
Expect basic research breakthroughs
In order to avoid technical blockade off some Western countries and supply uncertainties brought about by China's development, independent innovation, the concept of continuous enjoys popular support, especially in terms of the basic disciplines of independent innovation, we achieve catch-up, came from behind to leapfrog development of fundamental way. And we do also very exciting in terms of quantum computing, new materials science.
1. Quantum computing has entered a crucial period
2019, "Quantum hegemony" Competition between the technology giants to make quantum computing, this profound concept into the public eye. First, Google in October in the journal Nature issued a document that required just 200 seconds to complete the calculation of years to complete ultra-count, but such a conclusion has also been followed by IBM and other companies public opposition to the author in this regard. " 300 seconds to get the first supercomputer to calculate the amount of 10,000 years, quantum hegemony era has come, "the article has to be introduced.
Our quantum computing is fruitful, such as the recent Chinese Academy of Sciences Professor Pan Jianwei, the Netherlands and the German scientists, for the first time on a 20-photon quantum Bose sample input line 60 × 60 mode interference calculations are sharply on the four key indicators refresh international record, approaching the quantum computing research to achieve an important goal of "quantum hegemony."
Since the unit is not a simple 0,1 quantum state but a coherent superposition states, and thus fault-tolerant quantum computation and quantum practical demonstration of the advantages of the industry has been looking forward to tackling the milestone breakthrough, quantum computing practical turning point, I want to achieve any one of which It will be a very difficult task, quantum computing technology will enter a crucial period. 2020 is expected to undergo quantum computing will further increase investment, competition intensified, accelerating industrialization and ecology richer stage.
2. The semiconductor device of innovative new materials to promote
The current mainstream of classical transistor process is 7nm, but because of quantum entanglement effect, the industry 3nm or less for the chip manufacturing process generally pessimistic. Under Moore's Law slowing down and calculate power and storage requirements of the dual pressures of the outbreak, the scientific community is also looking for a classic semiconductor material - alternative to silicon.
Currently Bi2Se3, Sb2Te3, Bi2Te3 representative of topological insulator, etc. can be achieved without loss of electron and spin transport, so that the new high-performance logic and interconnect calculation unit becomes possible. New material and new resistive material such as to bring high-performance magnetic memory and SOT-MRAM resistive memory.
Taking into account the traditional silicon semiconductor material has the potential to be close to the limit, we would like to improve the performance of only find a breakthrough in terms of new materials, so in 2020 I believe that this area is likely to usher in a breakthrough.
3. The modular design allows the world not difficult chip
In 2019 China's chip industry, which spread proverb called "In addition to water and air, and the rest are bought from abroad", sounds like quite an exaggeration, but in fact is not an exaggeration, in addition to technology, high chip costs, long payback period has become a stumbling block hindering the development of its industry. How to reduce costs and improve operational efficiency of the chip is the primary problem in today's major companies under consideration. Wuzhen Internet Conference this year, flat head Ali's brother chip company pioneered open things up a chip (MCU), I have been in the "100 lines of code interpretation Ali AloT chip platform no sword 100" made a detailed interpretation.
As the traditional chip design model can not effectively respond to the needs of fast iterative chip, customized and fragmented. RISC-V to open the instruction set, represented by their respective open source SoC chip design, hardware description language and high-level abstraction of templated IP-based chip design methods, design methods and promote the rapid development of chip-chip agile open source ecology. Further, based core particles
(Chiplet) of modular design with advanced packaging different functional way "chip module" package together, you can skip taped to quickly customize an application in line with demand, further accelerated the chip's delivery. 2020, China's investment in chips will certainly increase even further, then the modular design of the chip mode, the chip industry can make our country a higher level, especially worth the wait.
4. AI-aware intelligence to cognitive development from intelligent direction
Before AI since AlphaGo debut, cause and effect can not be found in the "listening, speaking, see" intelligent perception of the field to meet or exceed the level of humanity can achieve, but they can only find a statistical link natural language processing, image recognition, speech synthesis, etc. relationship. As the fourth game of chess three years ago Shishi and AlphaGo in the 78th dig a winning hand, not impossible to crack, but the AI is to be defeated this hand, the reason is still unknown, which reflects the perceived inadequacies of intelligence.
而认知智能将从认知心理学、脑科学及人类社会历史中汲取灵感，并结合扩领域知识图谱、因果推理、持续学习等技术，建立稳定获取和表达知识的有效机制，让知识能够被机器理解和运用，让AI能够回答为什么的问题，实现从感知智能到认知智能的关键突破。
笔者在参加2019年BDTC大数据年会时就观察到有关因果推理等认知智能领域的话题得到了业内极大的关注，因此我们也有理由相信2020年人工智能在通过外部知识、逻辑推理或者领域迁移的认知智能将可能迎来突破。
技术发展新趋势
除了基础科研方面的突破，物联网、AI、云计算等新兴技术发展新动向也同样值得关注，具体如下：
1．物联网+工业的迎来爆发
随着5G的落地；IoT设备数量的爆发、边缘计算概念的提出的使得工业生产的各个要素得以深度整合，提高制造效率、改善产品质量、降低产品成本和资源消耗，将传统工业提升到智能工业的新阶段。从当前技术发展和应用前景来看，物联网在工业领域的供应链管理、生产过程工艺优化、产品设备监控管理、环保监测及能源管理、工业安全生产管理都发挥了巨大作，实现工控系统、通信系统和信息化系统的智能化融合。
通过物联网+的方案制造企业将实现设备自动化、搬送自动化和排产自动化，进而实现柔性制造，同时工厂上下游制造产线能实时调整和协同。这将大幅提升工厂的生产效率及企业的盈利能力。对产值数十万亿乃至数百万亿的工业产业而言，提高5%-10%的效率，就会产生数万亿人民币的价值。物联网+工业必将在2020年迎来爆发。
2．AI+物联网使机器间大规模协作成为可能
传统单体智能无法满足协调大规模的智能设备共同完成实时感知、决策工作。但随着物联网协同感知技术、5G高速通信技术的发展，使多智能体之间的协同合作成为可能。多智能体协同将使物联网进一步智能化，强化智能系统的价值。比如大规模智能交通灯调度将实现动态实时调整，仓储机器人协作完成货物分拣的高效协作，无人驾驶车可以感知全局路况，群体无人机协同将高效打通最后一公里配送，甚至将智慧城市更超大规模的智能终端合作成为可能。
3计算存储一体化突破AI算力瓶颈
目前深度学习需要采用规模非常庞大的网络，存储很多参数，完成大量的计算，同时在这些计算过程中会生成大量的数据。为了完成这些计算，芯片设计中会增加很多运算单元，如几千到几万个运算单元，随着运算单元数目的增加，每个运算单元能够使用的存储器的带宽和大小都在减小，存储器成为了计算的瓶颈。
由于深度学习并不属于通用计算的范畴，冯诺伊曼架构的存储和计算分离，已经不适合数据驱动的人工智能应用需求。频繁的数据搬运导致的算力瓶颈以及功耗瓶颈已经成为对更先进算法探索的限制因素。类似于脑神经结构的存内计算架构将数据存储单元和计算单元融合为一体，能显著减少数据搬运，提高计算并行度和能效。计算存储一体化在硬件架构方面的革新，将突破AI算力瓶颈。
4．云服务下沉成为新十年的基础设施
笔者在前文《神龙飞天，国士王坚》中曾经介绍过在阿里神龙服务器和飞天操作系统的加持下，阿里甚至整个云服务中的虚拟化层所带来的损耗正在被不断的降低，云服务凭借其标准化、弹性化的优势，使得用户只需专注应用开发，无需关注基础设施及基础服务，而且通过云原生的资源交付方式，可以提高计算效率、易用性，降低用户的计算和运维成本。云服务正在变得像电力、自来水一样的无处不在，成为数字经济时代基础设施。
来自于技术的关怀
技术本身不全是冷冰冰的，很多新兴的技术也会带有一定的人文关怀，比如：
1.保护数据隐私的AI技术将加速落地：
数据流通所产生的合规成本越来越高。使用AI技术保护数据隐私正在成为新的技术热点，其能够在保证各方数据安全和隐私的同时，联合使用方实现特定计算，解决数据孤岛以及数据共享可信程度低的问题，实现数据的价值，同时避免数据泄露的风险。
2.规模化生产级区块链应用将走入大众
2020年除了央行基于区块链的数字货币很可能会进行试点以外，随着BaaS(BlockchainasaService)将进一步降低企业应用区块链技术的门槛，专为区块链设计的端、云、链各类固化核心算法的硬件芯片等也随着国家的大力扶持而应运而生，实现物理世界资产与链上资产的锚定，进一步拓展价值互联网的边界、实现万链互联。
之前区块链的发展过于偏向于虚拟化了，而通过正本清源目前区块链产业的发展也越来越贴近实体经济，我们看到目前腾讯与阿里等头部企业在区块链方面的专利储备是非常雄厚的，2020年肯定会涌现大批创新区块链应用场景以及跨行业、跨生态的多维协作，日活千万以上的规模化生产级区块链应用将会走入大众。
