百度研究院：2020年10大人工智慧科技趨勢

近日，百度研究院釋出了一份關於2020年人工智慧科技趨勢預測的報告，報告從十個角度對2020年AI的主要發展趨勢做了闡述。

以下是十大預測趨勢的詳細解讀：

趨勢一：AI 技術已發展到可大規模產業化階段，2020年將出現多家AI工廠

AI技術以及各類商業解決方案已日臻成熟，並快速進入產業化階段。隨著全球科技巨頭對AI技術的持續投入，到2020年，全球範圍內將出現多家人工智慧模型與資料工廠，從而大規模推動人工智慧技術和相關的商業解決方案更新產業。例如客服行業的AI解決方案將可以大規模複製運用到金融、電商、教育等諸多行業中。

(英文原文，下同：The increasingly mature AI technology and all types of associated business solutions are rapidly entering the stage of "industrialization". With the continuous investment global technology giants pumped into AI technology, there will be many factories of AI models and data emerging in 2020, facilitating AI technology and associated commercial solutions on a large scale to update industries. For example, AI solutions in the customer service industry can be copied and applied to finance, e-commerce, education and other industries on a large scale. )

趨勢二：2020年將會是AI晶片落地的關鍵年

近幾年，AI晶片逐漸達到了可用的狀態，2020年將會是AI晶片大規模落地應用的關鍵一年。端側AI晶片將更加低成本、專業化、解決方案整合化。同時，神經網路處理單元(NPU)將成為下一代端側通用CPU晶片的基本模組，未來越來越多的端側CPU晶片將會以深度學習為核心進行全新的晶片規劃。除了晶片以外，AI還將重新定義計算機體系架構，支援人工智慧訓練和推理，成為異構設計架構的新思路。

(In recent years, AI chips have gradually reached a usable state, and 2020 will be a critical year for the large-scale implementation of AI chips. AI chips on the edge will be more low-cost, specialized and seamlessly integrated into downstream solutions. At the same time, the neural processing unit (NPU) will become the basic module of the next-generation edge-based general-purpose CPU chips. In the future, more and more device-based CPU chips will integrate deep learning framework as the core to their designs. In addition to chips, AI will redefine the computer architecture and support AI training and inference as a new idea of heterogeneous design architecture. )

趨勢三：深度學習深入滲透產業，並大規模應用

深度學習是人工智慧領域最重要，也是被產業界證明最有效的技術。以深度學習框架為核心的開源深度學習平臺大大降低了人工智慧技術的開發門檻，有效提高了人工智慧應用的質量和效率。2020年，深度學習將大規模應用於多個行業，實施創新，加快轉型升級。

(Deep learning is the most important and effective technology in the field of artificial intelligence. At the core of open-sourced deep learning platforms is the deep learning framework, which greatly lowers the development threshold of AI technology, and effectively improves the quality and efficiency of AI applications. In 2020, deep learning will be applied across many industries at scale to implement innovation and accelerate transformation and upgrading. )

趨勢四：AutoML將大大降低機器學習的門檻

AutoML的快速發展將大大降低機器學習的門檻，擴大AI應用的普及率。AutoML將能夠把傳統機器學習中的迭代過程綜合在一起，構建一個自動化的過程。研究人員只需輸入元知識(如卷積運算、問題描述等)，演算法就可以自動選擇合適的資料、最佳化模型結構和配置、自動地訓練模型，並將其部署到不同的裝置上。

(AutoML will be able to integrate the iterative process in traditional machine learning and build an automatic process. Researchers only need to input meta-knowledge (such as convolution operations, problem descriptions, etc.), the algorithm can automatically select the appropriate data, optimize the model structure and configuration, train the model, and deploy it on different devices. The rapid development of AutoML will greatly lower the threshold of machine learning and increase the popularity of AI applications. )

趨勢五: 多模態深度語義理解進一步成熟，並得到更廣泛應用

多模態深度語義理解以聲音、影像、文字等不同模態的資訊為輸入，融合感知和認知技術，實現對資訊的多維度深層次理解。隨著計算視覺、語音、自然語言理解和知識圖譜等技術的快速發展和大規模應用，多模態深度語義理解逐漸成熟，應用場景更加廣闊。結合AI晶片，將廣泛應用於智慧家居、金融、安防、教育、醫療等行業。

(Multimodal deep semantic understanding takes the information of different models such as voice, image, and text as input, and integrates perception and cognition technologies to achieve a multi-dimensional deep understanding of information. With the rapid development and large-scale application of computing vision, speech, natural language understanding, and knowledge graph, multimodal deep semantic understanding is gradually mature, which leads to a broader application scenario. Combined with AI chips, it will be widely used smart home, finance, security, education, healthcare, and other industries. )

趨勢六：自然語言處理技術將與知識深度融合，面向通用自然語言理解的計算平臺得到廣泛應用

隨著大規模語言模型預訓練技術的出現和發展，通用自然語言理解能力有了極大地提高。基於海量文字資料的語義表示預處理技術將與領域知識深度融合，不斷提高自動答疑、情感分析、閱讀理解、推理、資訊提取等自然語言處理任務的有效性。集合超大規模算力、豐富領域資料、預訓練模型和完善研發工具的通用自然語言理解計算平臺將逐漸成熟，並在網際網路、醫療、法律、金融等領域得到廣泛應用。

(With the emergence and development of pre-training large-scale language model, the technology of general natural language understanding has been greatly improved. Semantic representation pre-training technology based on massive text data will be deeply integrated with domain knowledge to continuously improve the effectiveness of natural language processing tasks such as automatic question answering, emotional analysis, reading comprehension, reasoning, information extraction, etc. The general natural language understanding the computing platform, which integrates large-scale computing power, rich domain data, pre-training model and improved R&D tools, will be gradually improved and widely used in the internet, healthcare, legal, financial and other fields. )

趨勢七：物聯網將在邊界、維度和場景三個領域形成突破

隨著5G和邊緣計算的發展，算力將不再侷限於雲端計算中心，向萬物蔓延，會產生一個泛分散式計算平臺。同時，對時間和空間這兩個物理世界最重要維度的洞察，將成為新一代物聯網平臺的基本能力。這也將推動物聯網與能源、電力、工業、物流、醫療、智慧城市等更多場景發生融合，創造出更大的價值。

(With the development of 5G and edge computing, computing power will not be limited to cloud computing centers, expanding to everything and building a distributed computing platform. At the same time, the insight into time and space, the two most important dimensions of the physical world, will become the basic capabilities of the new-generation IoT platforms. This will promote the integration of IoT with more scenarios such as energy, power, industry, logistics, medical treatment, and intelligent city, and create greater value.)

趨勢八:智慧交通將加速在園區、城市等多樣化場景中落地

自動駕駛的發展正在趨於理性，未來幾年市場對智慧駕駛的發展也會更加有信心。2020年，自動駕駛汽車將被應用於物流快遞、公共交通、封閉道路等不同場景。同時，V2X(vehicle to everything)技術啟動規模化部署和應用，這使得車輛和道路形成一個廣泛的聯絡，進一步推動智慧車路協同技術的實現，智慧交通加速在園區、城市、高速等多樣化場景中落地。

(The development of autonomous vehicles is becoming more rational, and the market will be more confident in the development of intelligent driving in the next few years. In 2020, more autonomous vehicles will be applied to different scenarios such as logistics, public transport, geofenced areas and so on. At the same time, V2X (vehicle to everything) technology is ready for large-scale deployment and application, which makes vehicles and roads form a wide range of connections, further promoting the realization of Intelligent Vehicle Infrastructure Cooperative Systems (IVICS), and accelerating the implementation of intelligent transportation in parks, cities, expressways and other scenarios. )

趨勢九：區塊鏈技術將以更加務實的姿態融入更多場景

隨著區塊鏈技術與人工智慧、大資料、物聯網以及邊緣計算的深度融合，資料與資產的線上線下對映問題將逐一解決。圍繞區塊鏈構建的資料確權、資料使用，資料流通和交換等解決方案，將在各行各業發揮巨大的作用。例如，在電商領域，可保證商品全流程資料的真實性;在供應鏈領域，可保證全流程資料的公開和透明，以及企業之間的安全交換;在政務領域，可以實現政府資料的打通、電子證書的實現等。

(With the in-depth integration of blockchain technology with AI, big data, IoT and edge computing, the problems concerning the online and offline mapping of data and assets will be solved one by one. Solutions such as data authorization, data use, data circulation and exchange built around blockchain will play a huge role among people from all walks of life. For example, in e-commerce, blockchain can ensure the authenticity of the whole process data of goods; in supply chain, it can ensure the openness and transparency of the whole process data, as well as the safe exchange between enterprises; in government affairs, it can achieve the opening of government data, the realization of electronic certificates and so on. )

趨勢十：量子計算將迎來新一輪爆發，為AI與雲端計算注入新活力

隨著量子霸權的成功展示，量子計算將在2020年迎來新一輪爆發。量子硬體方面，可程式設計的中等規模有噪量子裝置的效能會得到進一步提升，並具備糾錯能力。具有一定實用價值的量子演算法將能夠在其上執行，量子人工智慧的應用將得到極大的發展。

量子軟體方面，高質量的量子計算平臺和軟體將會出現，並與AI和雲端計算技術深度融合。此外，隨著量子計算生態產業鏈的初步形成，量子計算必將在更多應用領域受到更多的關注。越來越多的行業巨頭陸續投入研發資源進行戰略佈局，這將給未來的人工智慧和雲端計算領域帶來新的面貌。

(With the successful demonstration of quantum hegemony, quantum computing will usher in a new round of explosive growth in 2020. In terms of quantum hardware, the performance of programmable medium-sized noisy quantum devices will be further improved and have the ability of error correction. Quantum algorithms with certain practical value will be able to run on them, and the application of quantum artificial intelligence will be greatly developed. In terms of quantum software, high-quality quantum computing platforms and software will emerge and be deeply integrated with AI and cloud computing technologies. In addition, with the emergence of the quantum computing industry chain, quantum computing will surely garner more attention in more application fields. More and more industry giants have invested in R&D resources for strategic layout, which has the opportunity to bring a new face to the future AI and cloud computing fields. )