聯合國教科文組織2021年Netexplo創新論壇於2021年4月14日至15日在網上舉行。由技術領域全球知名大學組成的Netexplo大學網絡歷時一年，在全球范圍內遴選出了10項極具突破性的數字創新技術，這些創新對社會具有深遠而持久的影響。
The United Nations Educational, Scientific and Cultural Organization (UNESCO) launched a new forum in Netexplo in 2021, which took place online from 14 to 15 April 2021. One year after the University of Netexplo network, organized by leading universities worldwide in technical and technological fields, 10 cutting-edge digital innovation technologies have been selected around the world that have had a profound and lasting impact on society.
阿根廷交易平台Abakus
受疫情影響，阿根廷經濟遭受重創，也使該國農業部門的動蕩加劇。鑒於此，阿根廷初創企業Abakus與瑞士區塊鏈基礎設施提供商CoreLedger推出了一個數字易貨平台，旨在幫助農民克服金融波動帶來的影響。
As a result of the epidemic, Argentina’s economy has been devastated, and the country’s agricultural sector has become increasingly active. As a result, Argentina’s first-ever company, Abakus, and the Swiss sector’s chain of infrastructure provider, CoreLedger, have launched a digital platform to help farmers overcome the effects of financial shocks.
Abakus系統基於一種以實物資產(例如大豆)為索引的加密貨幣，這種貨幣可能比本國不斷波動的貨幣更穩定，而代幣可以兌換成貨物或比索。這兩家公司表示，代幣化農業資產將有助於農民對沖通貨膨脹，並在國內和國際上獲得流動性。
The Abakus system is based on an encrypted currency indexed to real goods (such as soybeans), which may be more stable than the country’s constant currency, which can be exchanged for goods or peso. The two companies say that monetized agricultural assets will help farmers to inflate their prices and gain mobility within the country and internationally.
英國人工智能AlphaFold 2
在國際象棋和圍棋領域取得巨大成功后，去年年底，人工智能再次大出風頭。在國際蛋白質結構預測賽中，AlphaFold 2摘得桂冠，並破解了一個困擾人類50年的難題：預測蛋白質如何折疊。
At the end of last year, after the great success of chess and chess fields, artificial intelligence resurfaced. In the International Protein Structure Forecasting Competition, AlphaFold 2 was crowned and solved a 50-year dilemma for humanity: predicting how the protein collapses.
在過去50年中，“蛋白質折疊問題”一直是生物學界的重大挑戰。此前，生物學家主要利用X射線晶體學或冷凍電鏡等實驗技術來破譯蛋白質的三維結構，但這類方法耗時長、成本高。而AlphaFold 2不僅預測准確且效率高，有望促進醫學領域不斷取得進步。
In the last 50 years, the “protein folding problem” has been a major challenge for the biology community. Previously, biologists used experimental techniques such as X-ray crystallography or cold mirrors to interpret the three-dimensional structure of proteins, but these methods are time-consuming and costly.
美國圖像生成系統Dall-E
美國人工智能非營利組織OpenAI於今年1月份推出Dall—E，這是一個可以根據書面文字生成圖像的人工智能系統，該名稱來源於著名畫家達利(Dalí)和機器人總動員(Wall-E)。
In January this year, the American A.I.A. non-profit organization OpenAI launched Dall-E, an artificial intelligence system that produces graphics based on written text, from the famous painters Darí and Wall-E.
該系統可以根據簡單的描述創建極其逼真和清晰的圖像，精通各種藝術風格，包括插畫和風景等。它還可以生成文字來制作建筑物上的標志，並分別制作同一場景的草圖和全彩圖像。
The system can create accurate and clear images based on simple descriptions, mastering various artistic styles, including illustrations and landscapes. It can also generate text to make icons on buildings and separate sketches and colour images of the same scene.
多哥慈善機構“直接給錢”
“直接給錢”是一家通過無條件直接援助來緩解貧困的非政府組織，它使用美國伯克利大學開發的一種算法，以找出最需要幫助的人。這種算法可以利用電信運營商提供的數據，分析用戶的手機消費情況，確定用戶的貧困程度。
Direct Money is a non-governmental organization that uses untied direct assistance to alleviate poverty, using an algorithm developed by the University of Berkeley in the United States to identify those most in need. This algorithm can use data from telecom operators to analyse user cell phone consumption and determine user poverty levels.
多哥共和國政府啟動了“直接給錢—多哥”項目，利用衛星圖像和手機數據尋找最需要扶助的公民。研究人員利用了多哥政府用於發放新冠救濟金的手機平台Novissi，每月5次向多哥最貧困地區的5.7萬名弱勢群眾提供援助。
The Togolese government has launched the “Direct Money-Togo” project, using satellite images and cell phone data to find the citizens most in need. Using Novisi, the Togolese government’s mobile phone platform for the release of a new crown, researchers provide assistance to 57,000 vulnerable people in Togo’s poorest districts five times a month.
德國文本AI模型GPT-Neo
German text AI model GPT-Neo
OpenAI的GPT—3被認為是目前最好的人工智能文本生成器，其擁有1750億個參數，現已被數萬開發者用於300多個不同的應用程序，每天輸出45億詞之多。
OpenAI's GPS-3 is considered to be currently the best AI text generator, with 175 billion parameters, and has now been used by tens of thousands of developers for more than 300 different applications, with 4.5 billion words per day.
但它是收費的，這阻礙了更多開發人員的採用，也不利於文本AI的快速發展。有鑒於此，德國Eleuther人工智能公司於今年3月下旬推出開源的文本AI模型GPT—Neo，以彌補這方面的缺憾。研究人員稱，GPT—Neo的推出將會催生出一大批嶄新應用，也會以更為低廉的成本釋放人們對人工智能未來的想象力。
In response, the German company, Eleuther artificial intelligence, launched an open-source version of the GPT-Neo model in late March this year to fill this gap. Researchers claim that the GPT-Neo launch will generate a large number of new applications and release people’s imagination of the future of artificial intelligence at a lower cost.
中國量子計算機“九章”
Chinese Quantum Calculator Chapter 9
“九章”是由中國科學技術大學潘建偉、陸朝陽等學者研制的76個光子的量子計算原型機。
“Chapter 9” was developed by Chinese science technologists, such as Pan Jianwei, and Luang yang, to produce 76 photon quantum calculator prototypes.
實驗顯示，“九章”對經典數學算法高斯玻色取樣的計算速度，比目前世界最快的超算“富岳”快100萬億倍，從而在全球第二個實現了“量子霸權”，推動全球量子計算前沿研究達到一個新高度，其超強算力在圖論、機器學習、量子化學等領域具有潛在應用價值。
Experiments have shown that the nine-chapter calculation of the classic mathematical algorithm Gospel is one billion times faster than the world's fastest ultra-rich, and that the second global realization of the “quantage hegemony” has led to a new altitude in the cutting edge of global quantum calculations, with super-calculations with potential applications in the fields of graphics, robotics, quantum chemistry, etc.
當求解5000萬個樣本的高斯玻色取樣問題時，“九章”需200秒，而“富岳”需6億年﹔當求解100億個樣本時，“九章”需10小時，“富岳”需1200億年。
It takes 200 seconds to solve the Gospel sample problem of 50 million samples, while it takes 600 million years to solve the problem; it takes 10 hours to solve the 10 billion samples, and it takes 12 billion years to solve the problem.
對於“九章”的突破，《科學》雜志審稿人評價稱“這是一個最先進的實驗和重大成就”。
For the breakthrough of the “nine chapters”, the science magazine reviewer commented that “it was a first experiment and a major achievement”.
美國人工智能“神經破譯”
American A.I.D.
“神經破譯”是由麻省理工學院和谷歌公司的研究人員攜手開發的一款人工智能軟件，基於語言進化原理，可以在不知道語言來源的情況下破譯古代語言，並揭示某種語言和其他語言之間的關聯。
“Dynamic Breaking” is an artificial intelligence software developed by researchers at the Massachusetts Institute of Technology and Google, based on the theory of language evolution, which can decipher an ancient language without knowing the origin of the language and reveal the connection between one language and another.
研究者們利用同一語族內不同語言之間的聯系，用該人工智能破譯了兩種失傳的語言：烏加裡特文和線性文字B，堪稱現代版的“羅塞塔石碑”！羅塞塔石碑是一塊用3種語言寫了同一個內容的石碑，幫助語言學家們讀懂古文字。
Using the same connection between different languages in the same language, the researchers decoded the two lost languages with the artificial intelligence: Ugarit and Linear B, which is the modern version of the Rosetta Stone, which is a stone written in three languages that helps linguists to read ancient scripts.
法國消費者應用程序Rift
Rift應用程序由投資平台Lita推出。投資和儲蓄對可持續發展的影響至少與支出一樣大。Rift讓用戶知道銀行如何使用他們的錢。該應用程序也可以指出最符合用戶優先級的金融產品。
The Rift application is launched by the investment platform Rita. Investment and savings have at least the same impact on sustainable development as spending. Rift lets users know how banks use their money.
印度區塊鏈應用程序Smashboard
Indian section chain applicationSmashboard
Smashboard網站稱自己是一個“另類的社交媒體網絡，為性侵受害者提供獨特的功能，目的是通過減少報案帶來的心理創傷，讓他們的生活更輕鬆”。
The Smashboard website describes itself as an “other type of social media network that provides victims of sexual abuse with a unique function aimed at making their lives easier by reducing psychological trauma caused by the underreporting of cases”.
借助區塊鏈技術，Smashboard允許用戶創建他們所遭受犯罪的私人和加密賬本，並將其安全地存儲在網上。例如，Smashboard為用戶提供了收集材料的選項，“這些材料可作為日記，也可作為加密個人空間中的時間戳証據”。
With block-link technology, Smashboard allows users to create private and encrypted accounts of the crimes they have suffered and to store them safely on the Internet. For example, Smashboard provides users with options for collecting materials that “can be used as journals or as a time stamp in encrypted personal spaces”.
更重要的是，該應用程序允許性侵受害者找到法律顧問或律師，並與他們私下互動。
More importantly, the procedure should allow victims of sexual assault to find legal counsel or lawyers and to interact with them in private.
澳大利亞人工智能“蠕虫大腦驅動器”
研究人員從蠕虫微小的大腦中獲得靈感，開發出一種能夠控制自動駕駛汽車的人工智能系統“蠕虫大腦驅動器”。與大多數網絡中使用數百萬個神經元相比，該系統隻需要幾個神經元。
Researchers have developed an artificial intelligence system, a worm brain drive, that controls auto-drive cars. Compared to the millions of neurons used in most networks, the system requires only a few neurons.
該系統隻用了控制電路中的19個神經元和7.5萬個參數，就成功控制了一輛汽車。研究小組相信他們的新方法可減少訓練時間，並使人工智能在相對簡單的系統中實現成為可能。（記者 劉 霞）
With only 19 neurons and 75,000 parameters in the control circuit, the system succeeded in controlling a car. The research team believed that their new methods would reduce training time and make it possible for artificial intelligence to be realized in a relatively simple system.
分享讓更多人看到 
注册有任何问题请添加 微信：MVIP619 拉你进入群
打开微信扫一扫
添加客服
进入交流群
发表评论