{"id":25240,"date":"2018-05-12T21:45:08","date_gmt":"2018-05-12T21:45:08","guid":{"rendered":"http:\/\/www.kurzweilai.net\/?p=315642"},"modified":"2018-05-14T22:10:26","modified_gmt":"2018-05-14T22:10:26","slug":"two-major-advances-in-autonomous-technologies-that-rival-human-abilities","status":"publish","type":"post","link":"https:\/\/hoo.central12.com\/fugic\/2018\/05\/12\/two-major-advances-in-autonomous-technologies-that-rival-human-abilities\/","title":{"rendered":"Two major advances in autonomous technologies that rival human abilities"},"content":{"rendered":"
\"\"<\/a>

The voice of singer John Legend, shown here at Google I\/O, will be among six celebrity voices to come to Google Duplex voice technology and Google Assistant later this year, along with phones and home speakers. (credit: Google)<\/p><\/div>\n

Google Duplex<\/h4>\n

Google’s new artificial-intelligence Google Duplex<\/a> voice technology for natural conversations, introduced at the Google I\/O<\/a> event this past week, cleverly blurs the line between human and machine intelligence.<\/p>\n

Here are two impressive examples of Duplex’s natural conversations on phone calls (using different voices):<\/p>\n

Duplex scheduling a hair salon appointment:

Duplex calling a restaurant:

Google Duplex is designed* to make its voice on phone conversations sound natural** — \u201cthanks to advances in\u00a0understanding<\/em>,\u00a0interacting<\/em>,\u00a0timing<\/em>, and\u00a0speaking<\/em>,\u201d according to Google AI Blog<\/em><\/a>. For example, it uses natural-sounding \u201chmm\u201ds and \u201cuh\u201ds, and the appropriate latency (pause time) to match people\u2019s expectations. \u201cFor example, after people say something simple, e.g., \u2019hello?,\u2019 they expect an instant response.\u201d<\/p>\n

Google also said at its I\/O developer conference that six new voices are coming to Google Duplex, including singer-songwriter-actor John Legend\u2019s. Legend\u2019s voice, among others, will also come to Google Assistant later this year, and will be included in phones and home speakers.<\/p>\n

* \u201cAt the core of Duplex is a\u00a0recurrent neural network<\/a>\u00a0(RNN) … built using [Google's] TensorFlow Extended<\/a> (TFX).\u201d<\/em><\/p>\n

** To address \u201ccreepy\u201d concerns right out of a Westworld show, a Google spokesperson provided an email statement to CNET: \u201cWe are designing this feature with disclosure built-in, and we’ll make sure the system is appropriately identified.\u201d<\/em><\/p>\n