谷歌實現(xiàn)人工智能新突破

For many of us, it’s easy to tell if it’s a computer that’s calling you on your phone. Its voice
makes it obvious before it has the chance to finish telling you that you can get a free loan or that you are invited for a customer survey. It’s also easy to hang up – it’s just a computer, after all.

我們不少人很容易就能辨別出是否是一臺電腦在給你打電話。在電話那頭告訴你獲得免息貸款或者邀請參加客戶調(diào)查之前，電腦的聲音就已經(jīng)相當明顯了。這也讓人們很容易就掛斷電話 —— 畢竟那只是一臺電腦而已。

But Google Duplex, an artificial intelligence (AI) voice assistant, has shown that computers are able to “conduct a natural conversation with a human over the phone and complete simple real-world tasks,” reported tech site Wired.

但據(jù)科技網(wǎng)站W(wǎng)ired報道，谷歌人工智能語音助手Duplex能夠與人類進行自然的電話溝通，并完成現(xiàn)實中的簡單任務(wù)。

On May 8 during its annual developer conference Google I/O, the company showed Duplex making phone calls and having full conversations. First, it called a hair salon to make an appointment, before calling a restaurant to make a reservation.

在今年5月8日舉行的谷歌年度開發(fā)者大會(Google I/O 2018)上，谷歌公司用Duplex打電話并進行了完整的對話。它先給美發(fā)店打了個預(yù)約電話，然后再給一家餐廳打電話訂位。

As CNET noted, not only has Google changed the typical computerized voice, but the company has taken “extra steps to disguise the system to sound more like a human”.

正如科技資訊網(wǎng)站CNET所指出的那樣，谷歌不光改變了典型的電腦語音，還“進一步讓該系統(tǒng)聽起來更像人類”。

Duplex comes with imperfections – “umms” and “uhhs” – and uses typical human phrasing such as “Oh, gotcha.”

Duplex 會使用一些如“嗯”、“啊”之類的停頓語氣詞，以及常見的人類使用的措辭，如“啊，明白了。”

As a result, “the receptionists on the other end of the call have no idea they are conversing back and forth with AI,” commented CNET.

所以，“電話另一頭的人并不知道他們在和人工智能交談，”CNET評論道。

Some see Duplex as quite a breakthrough – it will be able to help humans cut down on boring tasks. But for the rest, it “straddles a fine line between being convenient and deceptive,” as Wired wrote. People have a hard enough time knowing whether news is real or fake, or what photos are to be trusted. And now, we may not be able to trust our own ears.

一些人將Duplex視為一大突破 —— 它能幫助人們完成枯燥的工作。但正如Wired所寫的那樣，還有一些人認為，它“跨越了便利和欺騙的界線”。人們現(xiàn)在已經(jīng)很難分辨新聞和照片的真?zhèn)瘟?。而如今，我們或許也無法相信自己的耳朵。

Google also mentioned that soon, Duplex will be able to imitate the owner’s voice. While this is a great step forward in AI, it also raises many concerns. For example, students could use Duplex to call their school to pretend they’re sick by using their mother’s or father’s voice.

谷歌也指出，很快，Duplex就能模仿主人的聲音。盡管這是人工智能向前邁進的一大步，但它也引發(fā)了諸多擔憂。比如，學生們可以用Duplex模仿父母的聲音，給學校打電話謊稱病假。

Since people are shocked by the fact that a machine can imitate a human so well, Google has promised that in the future the voice will tell the listener that it’s not a real person.

人們對于機器能夠如此惟妙惟肖地模仿人類感到震驚，鑒于此，谷歌承諾將來的語音將告知對方所聽見的聲音并非真人。

According to Alexander Rudnicky, a researcher of human-computer speech interaction at Carnegie Mellon University, technology like this is something people are only going to become more aware of. “We just assimilate this interaction into our culture,” he told Wired.

美國卡內(nèi)基梅隆大學人機言語互動研究人員亞歷山大·魯?shù)履峥苏J為，人們對這樣的技術(shù)會越來越了解。“我們只是將這樣的互動融入我們的文化，”他在接受Wired采訪時說道。