Google’s newest app-based project, Smart Reply, is offering to read your mail, then draft short replies based on the content. For example, it could recognise an invitation to an event, cross-check the date with your phone calendar, then suggest responses along the lines of “Busy that night”, “Love to attend” and “Check and get back to you”.
The programme is able to “learn” from the phrasing in both mail and replies, and thus emulate how you would usually respond to such requests. For executives with crammed inboxes, this does sound immensely convenient. While electronic skimming of e-mail is not new – how else do you think digital ads are tailored to your searches? – the large-scale “learning” across billions of accounts and responses is ground-breaking.
It’s a huge first step in creating a machine-based understanding of how the human language, and mind, works. A similar, self-educating system has been developed and launched for Google’s search engines as well.
Eventually, this might lead to robots that can behave in human fashion. Think call centres staffed by only computer programs, or a robot Jeeves that comes with personality. The trade-off is that some third party has access to your personal correspondence and who knows where that information will end up? As always with the issue of privacy in the digital age, the best preparation is awareness of the risks.