當AI越來越成熟，就會有人用嚟犯法！例如早幾年台灣有個Youtuber，利用AI「deepfake」技術，將名人嘅樣貌套落AV女優個身上，再轉售獲利，最後就梗係被人控告啦；而係香港，有人就利用我地特首嘅樣同聲音，推介人去投資某某計劃！
When AI matures, someone uses it to break the law. For example, youtube in Taiwan a few years ago, using AI 'deepfake' technology, put celebrity profiles on AV girls and then sell them for profit, and it's Hong Kong, where someone uses the same voice as our leader to invest in a project!
?
網上真假難分，特首講嘢你要聽？
事後反詐騙中心就梗係立即開Post，表示上流傳一段偽冒政府官員推介投資計劃的影片，以欺騙市民相信並作出投資；騙徒亦偽冒證券及期貨事務監察委員會（證監會）高層人員以即時文字訊息發出投資建議，相關部門已作出嚴正澄清影片及有關訊息全屬虛構。
After the incident, the Anti-Fraud Center immediately opened the Post, indicating that it had uploaded a video of a pseudo-public official offering an investment plan in order to deceive citizens into believing and making an investment; that the liars, as well as the senior members of the Securities Commission and the Futures Supervisory Board (CVC), had sent out proposals for an investment in instantaneous text, and that the relevant ministry had made a serious effort to clarify that the film and the related information were entirely non-existent.
政府官員又點會公開推介做私人投資呢？不過自從有「deepfake」技術之後，再加上「AI尹光」嘅聲音模仿生成技術流行起來，要自制某名人嘅聲音、動作、外貌真係話都無咁容易！
Government officials will publicly promote private investment. But since the "deepfake" technology, combined with the "AI Yoon Gwang" sound imitation technology, it's not so easy to make a celebrity's voice, action, appearance!
反詐騙協調中心：可從影片中人物眼神、嘴型、光線等睇下自唔自然
Anti-fraud Concord Centre: it's natural to look at people's eyes, mouths, light, etc. in the film
根據反詐騙協調中心所講，呢類AI片其實都有漏洞，只要留心以下事項：
According to the Anti-Swindle Coordination Centre, there are loopholes in this type of film, provided that the following matters are taken into account:
．市民可從影片中人物眼神、嘴型、光線等有否不自然，或把片段放大看看是否有模糊或剪接來分辨真偽；
. Citizens may use the image, mouth, light, etc. of a person in the film, or zoom in to see if there is ambiguity or clipping to distinguish authenticity;
．視訊通話中，可要求對方在鏡頭前做某些動作，例如頭部向上下左右轉動、以一隻手指慢慢橫越臉部，由於電腦需要針對突如其來的改變重新運算，可留意對方在屏幕的影象有否異樣，很可能出現破綻，如手指或五官出現模糊；
. In a message message, the other party may be required to do certain actions in front of the camera, such as moving the head up and down, slowly crossing the face with a finger, since the computer needs to recommence the sudden changes, to notice whether the image on the screen is different and is likely to be broken, such as a blurring of the finger or five officers;
．切勿隨意點擊超連結，登入任何可疑網站或下載附件；
. Do not randomly click overlinks, log into any suspicious website or download attachments;
．切勿將個人、信用卡資料、三位元數安全碼（CVC/CVV）及一次性密碼輸入不明來歷的應用程式或網站；
. Do not enter personal, credit card data, three-digit security codes (CVC/CVV) and one-time passwords into applications or websites with unknown origins;
．除了向有關機構查證外，市民也可在守網者網站「防騙視伏器」或「防騙視伏 App」手機應用程式，輸入可疑網址查核，以評估其詐騙及網絡安全風險；
. In addition to checking with relevant institutions, citizens can also check suspicious URLs to assess their fraud and cyber-security risks on Internet watchmen's websites, "Swindle Watchers" or "Stalking App" mobile phone applications;
．提醒身邊親友慎防受騙；
. Reminds relatives and friends to guard themselves against deception;
撰文:李皆穎
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