每日英語跟讀 Ep.K239: What are deepfakes – and how can you spot them?
Have you seen Barack Obama call Donald Trump a “complete dipshit,” or Mark Zuckerberg brag about having “total control of billions of people’s stolen data”? Answer yes and you’ve seen a deepfake. The 21st century’s answer to photoshopping, deepfakes use a form of artificial intelligence (AI) called deep learning to create images of fake events, hence the name deepfake.
你看過歐巴馬稱川普為「徹頭徹尾的蠢貨」，或是祖克柏吹噓「完全掌控了數十億人被盜的資料」嗎？若你回答是，那麼你已見識過「deepfake」（深偽）。deepfake是二十一世紀對Photoshop影像處理的回應，使用一種稱為「深度學習」的人工智慧來製作影像以偽造事情，因此而得名。
The AI firm Deeptrace found 15,000 deepfake videos online in September 2019, a near doubling over nine months. A staggering 96 percent were pornographic, and 99 percent of those mapped faces from female celebrities onto porn stars.
人工智慧公司Deeptrace二○一九年九月在網路上發現了一萬五千個deepfake影片，在九個月內幾乎翻了一倍。其中高達百分之九十六的比例是色情內容，所顯示的臉孔，百分之九十九是女性，從名人到色情片明星都有。
Is it just about videos? No. Deepfake technology can create convincing but entirely fictional photos from scratch.
deepfake只限於影片嗎？不是的。此技術可做出令人信服，卻是完全虛構的照片。
Audio can be deepfaked too. In March 2019, the chief of a UK subsidiary of a German energy firm paid nearly ￡200,000 into a Hungarian bank account after being phoned by a fraudster who mimicked the German CEO’s voice. Similar scams have reportedly used recorded WhatsApp voice messages.
聲音也可以被深度偽造。二○一九年三月，一家德國能源公司的英國子公司負責人接到一通詐騙電話，騙子的聲音模仿德國的CEO，該負責人便將近二十萬英鎊匯入一個匈牙利的銀行帳戶。據報導，類似的詐騙手法也用於錄製WhatsApp語音訊息。
How do you spot a deepfake? Poor-quality deepfakes are easier to spot. The lip synching might be bad, or the skin tone patchy. However, spotting a deepfake gets harder as the technology improves. In 2018, US researchers discovered that deepfake faces do not blink normally. No surprise there: the majority of images show people with their eyes open, so the algorithms never really learned about blinking. At first it seemed like a silver bullet for the detection problem, but no sooner had the research been published than deepfakes appeared with blinking. Such is the nature of the game: as soon as a weakness is revealed, it is fixed.
要如何辨識deepfake呢？品質差的deepfake較容易被識破──嘴唇可能不太同步，或者膚色不均勻。但隨著技術的進步，辨識deepfake變得越來越難。二○一八年，美國有研究人員發現deepfake的人臉無法正常地眨眼。這不令人意外：大多數影像所顯示的人臉，眼睛都是睜開的，因此演算法從未真正學習到眨眼。起先眨眼似乎是識破深偽的訣竅。但這項研究發表後不久，會眨眼的deepfake就出現了。這就是遊戲的本質：弱點一被發現，就會被修復。Source article: https://www.taipeitimes.com/News/lang/archives/2021/10/25/2003766696
更多通勤學英語Podcast單元:
每日英語跟讀Podcast，就在http://www.15mins.today/daily-shadowing
精選詞彙 VOCAB Podcast，就在https://www.15mins.today/vocab
語音直播 15mins Live Podcast, 就在https://www.15mins.today/15mins-live-podcast
文法練習 In-TENSE Podcast，就在https://www.15mins.today/in-tense
歡迎到官網用email訂閱我們節目更新通知。
商業合作洽詢: [email protected]