一窺未來，智能服務大使的原型Seeing the future: Knowing more about A.I. Ambassador
兩個目標：團隊認為智能大使除了能夠提供實用的功能服務之外， 亦需要為使用者帶來前所未有的體驗。
Two Principals: Providing useful service, meanwhile, creating new user experience.
人工智能好處
Advantage of Artificial Intelligence
雖然初期人工智能公司 Asiabots 曾經將技術應用在市面上的robots上， 雖然在前期相當「吸睛」、令人有耳目一新的體驗， 但是整體的表現仍然未達理想，除了感覺好玩之外，硬繃繃的沒有什麼表情，某程度上和星球大戰中的3po仍有一段距離。 也許一些其他的工作機械人還能夠勝任，但如果真的需要做到客戶服務， 可能還有一段距離。
Though Asiabots once attempted to integrate their technologies onto the robots, the first impression is great but the overall quality is not satisfying. It is interesting, but kind of robotic and have no emotion. Comparing with 3PO in Star Wars, it is not good enough to handle customer service enquiries.
所以要同時做到提供服務之餘亦需擁有嶄新的體驗的前提下，團隊的眼光移向虛擬角色Virtual Avatars。虛擬角色在科技電影一再出現， 例如星球大戰莉娜公主的全息影像， 或者是日本動漫超時空要塞《可曾記得起愛》四維虛擬影像，都是充滿未來感而且令人深刻。其實並不需要一個硬繃繃的機械人，虛擬機械人不是已經能夠滿足需要嗎？
The team moves on to develop Virtual Avatars in order to provide decent service and new user experience. Virtual characters often appear in Sci-Fi movies, like the hologram of Princess Lia in Star Wars or Macross, are so futueristic. Maybe virtual character can already satisfy one’s need?
人工智能服務大使 A.I. Ambassador 轉向虛擬影像角色
A.I. Ambassador is changed to Virtual Avatar
虛擬影像角色確實能夠滿足大部分服務需要，同時亦能帶來耳目一新的體驗。但是前提是有幾個重點的需要攻克：
Virtual Avatar can satisfy most of the users’ needs and bring a whole new kind of service at the same time, but there are a few points need to be noted:
面部表情必須流暢自然，口型能夠配合對答， 而且能夠根據答案內容表現合適的面部情緒。
面部表情之外身體的動作， 亦必須有細節而且自然流暢，例如平時在沒有人溝通的時候，需要作出自然升降起伏。
聲線亦為關鍵之處， 雖然可以以真人錄音代替， 但日後要修改時便會非常麻煩， 因為對答的內容可能因應情況改變而有所增減。
NLP Natural Language Processing unit 自然語言處理技術是整個關鍵的核心， 能夠理解人類說話。
Natural facial expression, with different natural facial expression according to the speech content.
Body gesture, even if it is not talking, the avatar needs to be natural and human-liked with movement.
Voice. A.I. Voice is another key element, though recording could solve the problem, when editing is needed, re-recording will become a huge workload.
Natural Language Processing (NLP). Understanding human language is the main component of the whole system.
選擇哪一種虛擬角色風格亦是關鍵所在， 畢竟這個是第一眼所接觸之處，需要考慮到市場的接受能力和用家的喜好。 這類屬於Human Computer Interaction 的設計以及牽涉人形的設定, 不能不提及恐怖谷理論 Uncanny Valley。當機械人或者虛擬人物越來越擬似真人， 移情作用便會進一步提升；但當擬似真人到達某一個像真度， 使用者會突然發現對象越來越詭異。簡單來說，當像真度未能夠完全十足十貼近真人， 所謂兩頭唔到岸時，恐怖感油然而生；用家會對任何模擬得不夠真實的地方特別敏感。直到擬真度進一步提升至極高水平才能脫離恐怖谷效應。
The character style is another main point since first impression is important. You must need to know what is Uncanny Valley when we talk about Human Computer Interaction design. When robot looks alike human, transference level is raised. However, when robot really ‘looks’ like human-being, users will suddenly feel the robot looks weird. In conclusion, when it could not perfectly look like a human, users will be specially sensitive to those imperfect parts, until the quality is improved to extinctive level.
2019年年底三星集團投資的一家新創公司, 展示了名為artifical human “Neon”, 並且號稱這是一種新的生命形態, Neon 成像非常漂亮而且動態自然, 故成功脫離恐怖谷之列。 但展示時仍然為開發狀態，在有些片段當中看到說話的口型與聲音仍然未能夠完全貼合。可能因為這原因所以未能夠現場做live demo。 不過期待技術的更新，Neon 應該將會更貼近真人。
In late 2019, Samsung invested a start-up and showed ‘Neon’, an artificial human, saying Neon is a new nature of life. Neon is beautiful in outlook and natural in gesture, therefore is not affected by Uncanny Valley. However, as Neon is still in R&D process, the shape of the mouth is not consistent with the speech, this might be one of the reasons that live demo is not shown. Neon should be more human-liked in the future with better technology development.
Comments