亡者意識以數位方式存在的世界將來臨了嗎--也許沒有這麼快
Has the time come for the deceased's consciousness to exist digitally? -- Perhaps not so soon.

海擇短評 Haize Comment：

Netflix"黑鏡(Black Mirror)中"San Junipero"與"Dark museum"談到的亡者意識以數位方式存在的世界將來臨了嗎？也許沒有這麼快。

近期一些淘寶商舖推出了"復活親人"的AI服務，讓人不禁意外，是不是黑鏡影集中亡者意識以數位方式重生的時代已然來臨。仔細了解該產品/服務後可以發現，其實，這也就是用軟體程式生成將照片動態化並配合口型說話的影片，這類技術其實已問世一段時間，並不算特別尖端的新技術。大體來說，這技術也沒有超過此前Amazon在拉斯維加斯"re:Mars"的Demo展中，讓Alexa語音助理用已故親人的聲音說話與朗讀文章的功能，而這個功能後來被衛道者批評為"病態的演示(morbid demo)"與"毛骨悚然的手段(a means for creepiness)"。

嚴格來說，淘寶商戶也好，Amazon也好，它們推出的產品/服務與生成式AI距離遙遠，一定要說的話，它與深偽(Deepfake)技術相關，可以說是基於人工智慧中的deep learning(深度學習)技術所創造出fake(假)訊息的一種。衛道者的批評已許有些過度引申，與其說是它把親人的存在娛樂化或商品化，它更值得擔心的可能會是在詐欺、盜版與破解人臉識別等上的應用。

親人或亡者的意識，就真的不能以生成式AI的方式"訓練"成功嗎？既然大語言模型都能訓練出ChatGPT與Bard，只要資源足夠，當然也能生成出能與親人的記憶、風格與語調與還在世者對談的連天機器人。真正的問題可能還在於誰有權創造或復活這些"親人"複製品，相反的來說，逝者的生成式AI版本，可能會在活著的親屬不知情或不同意的情況下被創造，類似問題已經在各種與已故名人相關的創作出現，比如一部關於名廚Anthony Bourdain的生平紀錄片中，曾出現了導演使用其聲音的爭議，這讓逝者的親人感到不安。當然，也有畢卡索、吳道子、張旭的新作在世間流浪千百年後重新問世的可能性，這應該也不是博物館或藝術品投資市場想看到的。

回到海擇資本開頭所說的，亡者意識以數位方式存在的世界將來臨了嗎？也許沒有這麼快，但，這不是因為技術做不到。

-------

Has the world depicted in Netflix's "Black Mirror" episodes "San Junipero" and "Dark Museum," where the consciousness of the deceased exists digitally, arrived? Perhaps not so soon.

Recently, some Taobao shops launched AI services to "resurrect relatives," unexpectedly suggesting that the era of digital rebirth of consciousness, as seen in "Black Mirror," may have dawned. However, upon closer examination, these products/services merely animate photos and sync them with lip movements to speak in videos, a technology that has been around for a while and is not particularly cutting-edge. In essence, this technology does not surpass Amazon's demo at the "re:Mars" event in Las Vegas, where Alexa was made to speak and read articles using the voice of deceased loved ones. This feature was later criticized by ethicists as a "morbid demo" and "a means for creepiness."

Strictly speaking, whether it's Taobao merchants or Amazon, the products/services they offer are far from generative AI. If anything, they're closer to Deepfake technology, which is a type of fake information created using deep learning from artificial intelligence. Critics might be overinterpreting the situation. Rather than commodifying or turning the presence of loved ones into entertainment, a more worrisome aspect could be its potential applications in fraud, piracy, and circumventing facial recognition technologies.

Can the consciousness of a loved one or the deceased really not be "trained" through generative AI? Given that large language models have been trained to create ChatGPT and Bard, with sufficient resources, it's indeed possible to generate robots that can converse with the living, mimicking the memories, style, and tone of loved ones. The real issue might lie in who has the right to create or resurrect these "replicas" of loved ones. Conversely, generative AI versions of the deceased could be created without the knowledge or consent of living relatives, a concern already evident in various creations involving deceased celebrities. For instance, a documentary about the renowned chef Anthony Bourdain featured a controversy over the use of his voice, unsettling his relatives. Similarly, there's a potential for new works attributed to Picasso, Wu Daozi, or Zhang Xu to emerge thousands of years later, a scenario neither museums nor the art investment market would welcome.

Returning to the initial question posed by Haize Capital, has the world where the consciousness of the deceased exists digitally arrived? Perhaps not so soon, but not because the technology isn't capable.

標籤 Label： BABA  AI  AIGC  AMZN  Deepfake  Deeplearning