譯/陳韻涵
A Silicon Valley startup recently unveiled a drone that can set a course entirely on its own. A smartphone app allows the user to tell the drone to follow someone. Once the drone starts tracking, its subject will find it remarkably hard to shake.
The drone is meant to be a fun gadget. But it is not unreasonable to find this automated bloodhound a little unnerving.
矽谷一家新創公司最近推出一款可以自主設定路線的無人機。使用者可利用智慧手機應用程式命令無人機去跟蹤某人。一旦無人機開始跟蹤，被跟蹤對象將很難擺脫。
這款無人機設計初衷是作為有趣的小機具，但若說這隻自動獵犬讓人覺得有點不安，卻也並不為過。
A group of artificial intelligence researchers and policymakers last month released a report that described how rapidly evolving and increasingly affordable AI technologies could be used for malicious purposes.
The tracking drone helps explain their concerns. Made by a company called Skydio, the drone costs $2,499. It was made with technological building blocks available to anyone: ordinary cameras, open-source software and low-cost computer chips.
一群人工智慧的研究和決策人員上個月發布了一份報告，把進展神速且價格日益親民的人工智慧可能被用來作惡的情況，做了一番描述。
Skydio製造的這款售價2499美元（約合台幣7萬3688元）的無人機，即可用以說明他們的憂慮。這款無人機以任何人皆能取得的技術元件組成：一般的相機，開放原始碼軟體和低價電腦晶片。
In time, putting these pieces together will become increasingly easy and inexpensive.
"This stuff is getting more available in every sense," said one of Skydio's founders, Adam Bry. These same technologies are bringing a new level of autonomy to cars, warehouse robots, security cameras and a wide range of internet services.
隨著時日推移，拼湊這些元件也會變得越來越容易，也越便宜。
Skydio創辦人之一布萊說：「這種東西在各方面都變得更容易取得。」這些科技正把汽車、倉庫機器人、保全攝影機和多種網路服務帶向自動化的新層次。
But at times, new AI systems also exhibit strange and unexpected behavior because the way they learn from large amounts of data is not entirely understood. That makes them vulnerable to manipulation; today's computer-vision algorithms, for example, can be fooled into seeing things that are not there.
In such a scenario, miscreants could circumvent security cameras or compromise a driverless car.
然而有些時候，新的人工智慧系統也展現了奇怪和預料之外的行為，因為對於它們如何從大量數據學習，我們尚未完全了解。這使它們容昜受到擺布；以今日的電腦視覺演算法為例，它就會受到矇騙而看見實際上不存在的物體。
這種情況下，歹徒可能躲過保全攝影機偵測，或讓無人駕駛車出問題。
Researchers are also developing AI systems that can find and exploit security holes in all sorts of other systems, said Paul Scharre, an author of the report. These systems can be used for both defense and offense.
Automated techniques will make it easier to carry out attacks that now require extensive human labor, including "spear phishing," which involves gathering and exploiting personal data of victims. In coming years, the report said, machines will be more adept at collecting and deploying this data on their own.
AI systems are also increasingly adept at generating believable audio and video on their own. This will make it easier for bad actors to spread misinformation online, the report said.
報告作者之一沙爾雷說，研究人員同時也在研發可在各種其他系統中找到安全漏洞並加以利用的人工智慧系統，這些系統進可攻，退可守。
有了自動化技術，執行目前需動用大量人力的攻擊行動會變得更為容易，包括需要收集並利用對方個人數據資訊的「魚叉式網路釣魚」。報告指出，未來各種機器對於自主收集並運用數據資料會更為熟練。
人工智慧系統對於自動產生可信的音頻訊號與影像也日益熟練。報告指出，這將讓心懷不軌的使用者更容易在網路散播不實訊息。
說文解字看新聞
人工智慧（Artificial Intelligence,簡稱AI）是電腦科技的子領域，強調智慧機器能像人類般運作與回應，包括語音辨識、學習、計畫和解決問題；旨在創造智慧機器的人工智慧，已成科技產業重要的一環。
與人工智慧相關的研究相當技術性且專業化，人工智慧的重要議題包括設計電腦使其具有獲取知識、推理、解決問題、感知、學習、規畫和能夠操控及移動物體的能力。其中，知識工程（knowledge engineering）是人工智慧相關研究的核心，機器惟有在存取許多與世界相關的豐富資訊時，才能像人類動作和回應。
機器學習（machine learning）則是人工智慧的關鍵領域；機器學習簡單來說，就是從大數據中歸納出有用的規則與模式。機器學習主要分為監督式學習、非監督式學習、半監督學習和增強式學習。監督式學習需要人工輸入訓練範本（base），再讓機器預測結果；非監督式學習毋須人工輸入範本；半監督學習則僅提供少數資料的答案，讓機器判別誤差並尋找答案。增強式學習以人腦的思考模式，讓機器根據當前的狀況選擇執行動作並獲得反饋，甚至依據所執行的動作創造新的環境狀態。
