智能機器人將終結人際互動
Whatever happened to the human touch? If you believe the latest fad from the tech world, it won’t be long before we give up talking to real people and start talking instead to machines — or “bots”. Or it might be machines pretending to be people, or people pretending to be machines. The permutations of human and artificial intelligence are about to become interesting.
人際互動到底發生了什麼？如果你將科技界的最新時尚當真，過不了多久，我們將不再與真人對話，而開始與機器或者“機器人”交談。或者，我們談話的對象可能是假扮成人的機器人，或是假扮成機器人的人類。人類與人工智能之間的組合變化將變得越來越有趣。
But if it works as advertised, we will move one step further beyond a world of direct human interaction towards a future when screens — and new ways of engaging with computer intelligence — govern much of our lives. For sheer convenience, it will be hard to resist. Less obvious is what might be lost in the process.
但如果這樣做的效果像宣傳的那麼好，我們將更進一步遠離人際直接互動的世界，邁向一個由屏幕——以及與計算機智能交流的新方式——操控我們大部分生活的未來。如果純粹爲了便利，這種誘惑將難以抗拒。但較不易被察覺的是這個過程中我們可能失去什麼。
Facebook has just given this trend its biggest lift with the release this week of a way for companies to plug their bots into Messenger, its chat system. Want to know the weather? Ask a question and an intelligent agent will answer. Interested in buying something? Bots on hand will show you the latest offers and guide you through to the checkout.
Facebook剛剛爲這一趨勢帶來了史上最大幅度的提振，該公司上週發佈了一種讓企業將自己的機器人接入其聊天系統Messenger的方法。如果想了解天氣，只需提問，智能機器人就會作答。如果你對購物感興趣，身旁的機器人將向你展示最新報價，並指導你直到完成付款。
If people take to the idea, entire call centres could be rendered obsolete. Why hold on waiting for a human when you can get an immediate response from a bot on your smartphone? The call-centre reps are operating from scripts anyway and these are interactions that could be programmed into a piece of software.
如果人類接受這一理念，所有電話服務中心將被淘汰。如果你可以立刻在自己的智能手機上得到機器人的迴應，還有什麼理由繼續等待人工應答呢？反正電話服務中心人員是按腳本應答，這些互動完全可以被編入一款軟件。
Whether the bots catch on will depend on not stretching the technology beyond its limits. Microsoft’s wayward chatbot Tay, the algorithm pretending to be an adolescent girl that started spouting racist comments on Twitter, is Exhibit A for what can go wrong. Most artificial intelligence is not very intelligent and when the main technique for training the new systems involves something called machine learning, it turns out that machines can be taught bad behaviour as well as good.
要想讓機器人受歡迎，必須不讓這一技術走過界。微軟(Microsoft)任性的聊天機器人Tay就展現了機器人可能出現怎樣的錯誤，這個扮成一名青春期少女的算法程序已開始在Twitter上散佈種族主義言論。大多數人工智能並不是很聰明，當培訓新系統的主要技術涉及所謂的機器學習時，事實證明，機器人既可能學好，也可能學壞。
Not that the chatbots on Facebook are about to have a mind of their own and turn rogue. Most take a very limited approach to AI, using machine learning to “understand” the questions being put to them and returning largely pre-scripted answers. They try to identify a need, then draw you down a chain of conversation to a result.
並不是說Facebook上的聊天機器人很快會有自己的思想，而且變得下流。大多數機器人的人工智能模式都很有限，它們利用機器學習來“理解”接收到的問題，並主要反饋預設的答案。它們嘗試先確定需求，然後在和你交談的過程中給出答案。
Deeper levels of intelligence will also be on tap. But it is as likely to involve a human brain as one made from silicon. M, another Facebook chat service, relies on real people sitting in the background, returning answers, though to the user, it is not at all clear whether information is being generated by a person or a computer. As the machines become smarter they will take over more of the responses, eventually pushing human workers out of the loop altogether.
更深層次的人工智能也將隨時派上用場。但它既可能涉及硅腦，也可能涉及人腦。Facebook推出的另一款聊天服務M，依靠真人坐在幕後反饋答案，不過對用戶而言，他們根本不知道自己得到的信息是由人還是由計算機生成。隨着機器人變得越來越聰明，它們將承擔更多的應答服務，最終將人工完全擠出反饋流程。
The prospect of another layer of human interaction being lost to technology is bound to bring a pang of regret, not to mention a wave of concern about the impact on jobs. But then, who would want to give up ATMs and go back to queueing up for a bank teller? Most people won’t mind at all when the days of call-centre queueing have receded into history.
人際互動在又一個層面上敗於技術的前景必然會讓人類感到痛惜，更不用說對影響就業的大量擔憂。但誰願意放棄自動取款機而回去排隊等待銀行櫃員服務呢？當排隊等待電話服務中心應答漸漸成爲歷史時，大多數人都不會介意。
