Google I/O 2018 started on 8th May, and on the opening keynote, CEO Sundar Pichai introduced us to Google’s latest and arguably one of the most ambitious projects called Google Duplex. It is a technology to help google assistant become more ‘human’ in its interaction with real people, thus stepping closer to beating the Turing Test.
What is Turing Test?
The Turing test, developed by Alan Turing in 1950, is a test of a machine’s ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. [Source: Wikipedia]. The test helps determine whether we are interacting with a human or a machine. You might have seen an example of this scenario in the 1982 Sci-Fi movie Blade Runner, where robots are distinguished from humans by asking a series of unrelated questions.
How the new Google assistant performed?
Sunder demonstrated a couple of demo phone calls made by the new Google assistant implementing Google Duplex and it seemed the people at the other end could not distinguish it as a robot. Here is the video of the assistant making the calls:
Still the technology is not ready to handle a rigorous “Turing Test” as it is trained to handle only a specific line of conversation. The remarkable thing about this conversation is how Google assistant mimics the “umm”s and “aa”s just as a human is expected to say between sentences. It was also able to understand the Chinese lady with her accent and “not so fluent” English.
The three main technologies that goes into making the system are Natural Language Processing, Deep Learning(AI) and Text-to-Speech, among others. Learning more about these technologies will certainly help you in your career, so I highly recommend that you go through these. If you are a student, these will make fine subjects for your degree projects.
As of now, google is very particular about its use cases. According to Sundar, the technology is not yet refined and need a lot of development before it can replace a real human secretary to make our calls for us. For now, one of the potential use will be to collect data over phone calls about the working timings of small business on weekends or other holidays and update their Google Places snippets, so that customers need not enquire themselves by calling those businesses.
Threats and concerns:
The inability of humans to recognize the Google assistant as machines in these calls created a lot of concerns among a huge number of people. One of the major questions is “is it ethical of Google to not inform the call recipient that its a machine they are talking to?”. Moreover how safe are we from fake calls from machines that mimics some person we know and trust, to get confidential information from us? What if all our calls from a google device gets saved, in the guise of using them for training the assistant/ helping them better serve us or other such excuse. It is a major threat to our privacy. I’ve discussed more on online privacy in my last blog.
Businesses are already using chat bots that mimic sales personalities in their websites and social media pages. A new technology called Deep Fake uses deep learning in image processing to alter images and videos, an illustration of it shows former US President Barak Obama making a fake speech [See here]. Combined with technologies like Google duplex, fake and misleading videos can be created of influential people.
So at the end of the day, it all comes down to how we, as humans, use technology. We should be optimistic but at the same time very cautious.
If you liked this article please comment and show your support and interest so that I’ll be motivated to continue this effort. Like our facebook page if you haven’t already.