Google’s Creepy New Technology that Sounds Exactly like a Human

At the recently held Google’s I/O conference, the organization made an astonishing demonstration of the Google Assistant making a phone call to book a haircut. It did an unbelievable job. Not only did it use all the right words but it managed to pause at just the right spots and modulate its voice in precisely the correct manner. The incredible part was that the Google Assistant has picked up the human speeches to such a fantastic level that the person at the other end of the call didn’t even speculate that it was not a human but a robot. No wonder, such a significant achievement has also opened up a Pandora’s box of questions for Google.

Google Duplex talking business artificial intelligence

What is Google Duplex?

Some of the questions that were raised soon after the demonstration was: Is Google under any obligation to inform the other person that he/she is talking to a robot? Does such a technologically eventually play a role in breaking our trust in what we see and hear these days?
Though Google did not describe the functioning of Google Duplex in great detail on stage, it did issue a blog post soon after to fill up the general public on the operation of this new development. Certain drawbacks are attached to Duplex to regulate its power. Firstly, it is utterly incapable of engaging in open-end dialogues. It can only function in close-ended domains with complete control.

In short, Google Assistant will manage to work reasonably well. However, its capabilities will be only in formulaic situations. However, in situations where the AI is stuck in an out-of-context talk, it is equipped with a lot of tricks in its sleeve to escape such cases and bring the conversation back on track.

Google duplex architecture
These tricks are much like the other computer programs that have been designed earlier to talk to humans. They include rephrasing and repeating questions. This practice was also showcased during the demonstration recently. At first look, you might feel that fractions of the conversation reveal real intelligence however on a closer look you ought to realize that they are nothing but snippets of preprogrammed gambits. The Google’s blog post which was released soon after the conference did reveal some fascinating facts about the same. To achieve this feat, Duplex uses tricks like interruptions, syncs and requesting elaborations.

Also, you must note here that Google has not gone ahead for a full-fledged launch of Google Duplex. Till now, it has stressed on calling it an “experiment”. This statement means that there are chances of its limited availability in the market shortly or this software might not be released at all. For now, Duplex’s conversations have been constructed around only three themes which include scheduling an appointment with a hairdresser, making reservations at restaurants and asking businesses for their holiday hours. For now, Google also plans to restrict its availability to a limited number of anonymous users.

The last resort of Google Duplex to steer clear of situations that have gone out of hand is for a human to take over. Duplex is equipped with a “self-monitoring feature”. Much like Facebooks personal assistant. M. which according to the assurances given by the organization will be restricted to customer service, Duplex hands over the call in such situations to a living human operator. It is noteworthy that M was a disappointment and was shut down in this January.

Till now, the article has discussed in a fair amount of details what Google Duplex can do. Let us now come back to the ethical issues with which we initially began the article. The first question was that is Google under any obligation to inform the person on the other end of the call that he is talking to a robot? In response to this concern, Google has comforted the followers by stating that it believes in this responsibility and all measures shall be taken to ensure the same.

One question that is raised now is that how will the AI inform the person at the other end? See, if it goes on to say “Hi! I am a robot!” the other person is very likely to cut the call and ignore it altogether. However, other measures can be adopted like voice modulations. Furthermore, Google hopes that with the time certain social norms will evolve which will make it easier for people to differentiate between a robot and an AI.

Another ethical aspect that revolves around is the issue is the tremendous scope of mischief AI offers if it can freely pose as humans. Scam calls and automated hoaxes can bother us to no end. Imagine getting a panicked call that there is a shooting in your vicinity. The caller answers enough questions to ensure you that the person is indeed a human. However, soon after divulging the information confesses that it was a wrong number. Will you be able to sit in peace after this?

There is a broad consensus on the opinion that such a crucial development cannot be left on companies to regulate. There is a growing need for stringent laws to control the entire mechanism. In the absence of such laws, some other company may also take unfair advantage of this development.

Then, there are other more subtle impacts of the technology that can’t even be regulated. The small conversations that take place in the public sphere between two humans are what builds trust. If we eliminate the need for it altogether and instead give them a curtain to hide behind and avoid other humans in the society, what culture are we promoting?

Though Google has assured that it would place limitations on the number of calls that Google Assistant can place, won’t it anyway make us ruder to the people around as we can’t even differentiate between a human and a robot called? Won’t we be tempted to shut them out if the conversation doesn’t go as per our wishes?

Before the robots start doing the talking for us, we need to be more informed as citizens because Google is indeed not the first organization to come up with such technology neither will it be the last.

You might also like to read: