In the suburbs of New York night falls. For the Scott family’s son, who just turned five, it’s time for bed. Lying down and tucked in, he sings the same phrase as every day, a “Good night, dad; good night mom”. But before closing his eyes he also says goodbye to the other member of the family. “Good night, Alexa,” he says, heading for the air.
She is not your sister, nor is she your pet. Alexa is a virtual assistant that lives inside the Echo, an Amazon-designed speaker installed in the kitchen. Echo has no visible controls beyond a few simple buttons, but thanks to Alexa, an artificial intelligence similar to Apple’s Siri, it is able to answer questions asked in natural language and have simple conversations.
If you ask him the distance between the Sun and the Earth, he knows it. If it is asked to play a song, it does so. If you connect with smart home devices such as smart bulbs or electronic locks, Alexa can be asked to close the door or turn on the lights with just a verbal command.
It is, in a way, like the HAL 9000 computer from 2001: A Space Odyssey , but fortunately without the homicidal tendencies.
Alexa is unable to hear the boy’s goodbye from her room, but for the youngest in the house it is not a problem. For him, even if he doesn’t respond, Alexa is yet another person who lives under the same roof . One with a personality somewhat different from that of his parents, yes, since sometimes he does not seem to understand him completely, yes, but no less real for that.
The voice of his master
For that generation that is now beginning to speak or taking its first steps in the world, growing up surrounded by digital voices willing to answer any question or obey any order will be common, as natural as today is to slide a finger across the screen to pass the page of a document or enjoy a photo carousel.
“It is a simple concept but with an exciting potential, capable of profoundly changing our relationship with technology; even deeper than the leap to the graphical interface was in its day ”, says Satya Nadella, CEO of Microsoft, who believes that in the very near future voice will be the main way of using a PC or a mobile phone. We will continue to watch the screen to consume information, but the search and selection of the content will be done in natural language.
The company has bet heavily in this field. Cortana, the virtual assistant in Windows 10, is already used in 25% of searches carried out from that operating system. If the searches carried out through other platforms, such as Siri, Alexa and Google, are added, the total searches per month that are carried out using only the voice already exceed 50,000 million , according to figures from Timothy Tuttle, founder of MindMeld, a platform which offers companies the ability to create intelligent conversation interfaces.
The figure is surprising, there is no doubt. Especially if we take into account how imprecise current virtual assistants can be . There are tasks that seem to solve exquisitely. Program alarms and timers? No problem. Take the test: if you use an iPhone or your Android phone has an assistant, ask them to wake you up at a certain time. It is rare that it fails.
Instead, they fail miserably in other fields. They are often unable to understand instructions correctly or have to use exact voice commands that end the illusion of speaking in natural language.
Virtual assistants actually know how to listen to us very well . Thanks to machine or machine learning techniques, the effectiveness of voice interpreters has grown in recent years to exceed 98% in languages such as English and Spanish. Today it is possible to make a fairly successful transcription, in real time, of a conference or a live broadcast. Platforms such as Skype or YouTube take advantage of these techniques to offer automatically generated subtitles .
The difficult part, the key that makes us think of these assistants as still immature technology , is that many times they are not able to understand the context in which questions are asked or assigned a task. It is a complex problem; something even humans often err on, but which becomes especially frustrating when dealing with a machine.
They don’t even know when we are actually talking to them or addressing someone else and mentioning their name in the conversation. In the latest edition of the Super Bowl, for example, a Google ad triggered a wave of activations on Google Home devices across North America because the protagonist of the ad spoke the activation command – “Ok, Google” – out loud. The same has happened in the past with Alexa ads or in homes where one of the family members has the same name.
Chatting with the machine
A first step in achieving the illusion of speaking to an intelligence is to offer chained responses . Being able to understand that the conversation is not over. Siri and Assistant, Google’s new virtual assistant, already offer this service, albeit in a limited way. You might ask, for example, “which movie has won the Oscar for best director this year?” and then, after getting the answer, add: “Who was the lead actor?” Attendees will recognize that the second question is related to the first.
“Almost all the historical development of artificial intelligence has focused on learning to recognize the voice. It’s why machine [or machine] learning has been such a revered tool. She is very good in those situations. But recognizing the voice does not give us an idea of people’s motivation , and motivation is the key to understanding what you really want to achieve when verbalizing an order or a question ”, says Rob High, head of the Watson project at IBM, an artificial intelligence platform that rose to fame in 2011 after becoming the first computer to win over human contestants in the American television contest Jeopardy!
High believes that Watson is the first step towards the age of cognitive computing, machines that will be able to understand us and may even anticipate our needs . “They will be able to put the music we like according to the mood they perceive, for example,” he predicts. For High, however, talking about artificial intelligence is necessarily talking about non-human intelligence. “I don’t think it makes sense to digitally replicate the human brain’s way of thinking. We already have 8 billion human brains on this planet (sic). We have to focus on developing cognitive systems that complement us, ”he says.
There is still a long way to go to reach the level of interaction that films like Her show us, where the protagonist falls in love with an artificial intelligence , but that does not mean that in their current state these types of assistants are not useful. Its market, in fact, will grow during the next decade at a rate of 34% per year, reaching 12,000 million dollars in turnover, according to Global Markets Insights. In just a decade, 50% of interactions with machines could be done through voice . Some of that growth will come from new smartphones with better assistants or new devices for the home.
Google, which introduced its conversational artificial intelligence Assistant only a year ago , has begun to extend its presence to Android terminals equipped with the latest version of the operating system and Amazon has reached agreements with several electronics and telephony manufacturers to bring Alexa to their products.
In both cases, they are artificial intelligence platforms that have set their sights, above all, at home. These virtual assistants could end up becoming the butler that home automation technology needs to convince the general public.
Today it is possible to automate almost every element of the home, from security to lighting, through cleaning or temperature control, but the more smart devices are placed at home, the more cumbersome it becomes to control them . It takes several applications and always having a phone or tablet at hand as a control center.
Virtual assistants can be the lubricant that eliminates friction between the user and the smart home elements, since they allow to control them with the voice in a more natural way.
The other part of that growth will also come when considering its application outside the personal realm. If the voice becomes the new method of interaction with computing and technology – not the only one, but the preferred one – these assistants will also be common in our work .
Instead of spending hours analyzing an Excel sheet, we will ask for a summary of the most significant figures. Instead of navigating through complex menus, we will ask the computer to draw or design a structure by defining its physical characteristics. None of these tasks will be performed exclusively with voice – the image will continue to be an effective means of presenting information – but the combination of tactile or graphical interfaces with expert systems capable of understanding natural language will simplify many of the tasks we spend hours on today. .
Concerned about health
Medical information and the first diagnosis are also promising fields for these types of assistants. A study by the University of California at San Francisco and Stanford University estimates that 60% of adults who have a smartphone have used it at some point to search for health information , and with the arrival of virtual assistants the number of searches or inquiries by voice has skyrocketed.
Today’s assistants, however, leave much to be desired when it comes to answering these types of questions . They trivialize the consultations or are not able to offer numbers for medical attention or psychological support services in the face of phrases such as “I am thinking of committing suicide” or “I have been a victim of rape. According to Eleni Linos, one of the authors of the study, “these are cases that should be considered”.
Before making the purchase, it could be as simple as requesting what we want from the computer , saying out loud to send us those pants or that jacket that we just saw on a web or ordering a pizza. The virtual assistant himself will take care of letting us know when the order arrives at the door and, who knows, it may not be long, even picking it up.
Superior photo: Maurizio Pesce via Fickr
