First of all, what is artificial intelligence? You know what it’s not. It’s not the HAL 9000 computer from 2001: A Space Odyssey, nor is it the calming tones of Jarvis, Iron Man’s own inbuilt virtual assistant. Movies tend to have an interesting take on how you’ll eventually interact with an Artificial Intelligence, but such charismatic entities (or creepy ones in the case of HAL) still belong firmly in the realms of science-fiction.
Clear the clutter of this fantastical material from your mind and focus instead on what’s real. Deep Thought, a chess computer, is real, although this machine is named after a famous fictional supercomputer. Built-in the mid-1980s by a team of scientists at Carnegie Mellon University, this chess engine competed at the Grandmaster level, but it was defeated by the then World Champion. Deep Blue, a 1990s descendent of Deep Thought corrected this setback by taking the metaphorical scalp of that same legendary champion.
Of course, the algorithms that form the framework of an AI persona have come a long way since these famous chess battles. Indeed, the main branches of this technology tree still follow this maths-based approach. Chaotic weather patterned is studied by these thinking machines. They analyze financial trends, model biological processes, and generally crunch huge data sets until a solution to an issue is generated. But science is an osmosis-based discipline. That means these higher-echelon systems have a way of filtering their way down until they’re integrated into consumer products. The proof is in the pudding, as they say, for Alexa and her friends are examples of this principle in action.
AI meet the smart speaker
Before getting into the nuts and bolts that form the innards of an AI-provided device, you’re still in need of a consumer-grade virtual assistant primer. The description of the massive supercomputers that rule business and science ecosystems are one thing, but your AI system isn’t built to solve the world’s problems. Instead, it’s a personalized product, an electronic assistant that uses your voice to carry out hundreds, perhaps thousands of commands in your home and beyond.
A speaker listens for its Wake Word. In an Amazon Echo, for instance, the Wake Word is “Alexa.” although this default voice activation cue can be changed. There’s going to be a list of many common spoken commands further on in our Virtual Assistant study, so let’s leave this topic be for the moment. Next, the AI systems under review here are entirely capable of tracking your movements and activities. Then, since we’re talking about intelligence and memory, the device will actually learn from those movements. Amazon Echo and Google Home are both equipped with special communications protocols, so they network with all established smart home products.
So big supercomputers have developed to the point that their features have been miniaturized. Consumer gadgets have greedily adopted the smarter circuits and cognitive code. The result is a processing center that understands your spoken words and reacts to these words in real-time. And you’re not limited to basic monosyllabic utterances either, not when the coding encourages structured sentences. Naturally, if you’re unique voice is to be recognized by a voice-activated device, it’s going to require some training.
Mobile devices work under this assumption, so you’ve probably encountered this idiosyncrasy in an Apple or Android device. Siri is an intelligent assistant living inside your iPhone or iPad. For Android devices, this duty is assigned to Google Now, although the Google Pixel handsets are currently enjoying a slightly friendlier artificial intelligence experience thanks to Google Assistant. Essentially, every manufacturer offers a slightly different AI interface, and those virtual personalities have different names. You’re undoubtedly familiar with Siri, the Google Now and Google Assistant twins, and then there’s Cortana. This latter digital persona was designed for Windows 10, although she also has a persistent presence within her App.
Smart speakers and voice training
Stick with this segment, there’s some tech-speak on the way. Primarily, the Amazon Echo gadget hosts Alexa. Voice training is accomplished by repeating a series of phrases, twenty-five in all. You repeat the phrase after she has stated it, and this repetitive phrasing acts as the training system. Again, this technique is used in Cortana, Siri, and most other voice-activated products. Incidentally, Google has a potential edge in the field of voice recognition due to the strings of text that flow across its famous search engine every single minute of the day.
Basically, no other company has access to more spoken or written data. One example of this platform universalism is Google Translate. If you need to know what a word or sentence translates to in any particular language, it’s likely Google Home will have the edge, and that’s simply because of the vast database slumbering under the Google Translate interface.
Remarkably, all of the speech recognition modules running inside the programs of each of these intelligent attendants are fully realised. Training routines will improve commands, of course, and this routine may have to be repeated. Various regional dialects have caused some humorous responses from even the most attentive assistants, but they’re always learning. Accents and noisy background environments are issues, but they’re slowly being subdued by an AI boosted learning curve. But is everything really stored inside the attractive Bluetooth-enabled speaker of an Amazon Echo? Likewise, where did the voice-activated Google Home speaker get its language skills?
A powerful processor lives at the heart of both popular devices, but they just don’t contain enough digital horsepower to process thousands of commands. No, in order to process the kind of sound waves that contain nuanced, diction-heavy content, you’re going to need the addition of cloud computing. Without that server-based resource, a different one in each virtual assistant, only a few words will be recognized. It may even respond to the odd Wake Word and carry out a few basic operations, but not much more is possible without an internet connection.
Tell me more about far-field voice recognition
Amazon Echo and any other Amazon-branded products run a tight ship. The microphones and hi-fidelity speaker play music while listening for a voiced request, perhaps the score in tonight’s big game or a news update. Far-field voice recognition even accommodates this request from the other side of an echoing kitchen as the music plays, but more on those special microphones later. The human voice, meanwhile, generates very subtle waves of sound. These information-dense wavelets are so complex that only a cloud-equipped data processing center can properly decode them.
Alexa Voice Service (AVS) represents one such processing architecture. This is a domain reserved for Amazon Web Services and its Cloud Services subsidiary. On the Google Home front, those duties are similarly distributed, except this time it’s the Google Cloud Platform that addresses the voice-to-text translation service.
What we’ve got so far in the virtual assistant products market is a series of devices, not many so far, that incorporates a hands-free build. On the face of things, you’re looking at a smart wireless speaker, but it’s equipped with some interesting extras. Arguably, those extras are taking over, so the music playing feature will fall back into second place. Primarily, then, it’s the voice-activated personality that impresses most buyers. A question, uttered in plain English, with other languages to follow, is granted a snappy response. Localized actions are triggered by the device’s “Skills.” It plays music from an entertainment center, dims the lights, orders a pizza, or uses its deep learning capabilities to manage a smart home. Most of this code and data algorithms are installed in the compact housing of these remarkable gadgets, except for the voice processing muscle, a feature that’s stored in the cloud.
And all of these built-in features are far from static. Again, because of the wireless internet connection, everything is growing and improving. Deep neural networks are responsible for those language learning skills, by the way, while smart home thermostats and security sensors provide the same level of system adaptability for the thinking machine’s Smart Home capabilities.
The AI or Artificial Intelligence primer described in this post is intended more as a historical reference than anything else. Yes, there are algorithms that approximate the way you think. They make decisions, they learn, and they use voice-activated assistants as conduits to your home. The goal of this challenging work, all that server-contained hardware and software, is to build a truly frictionless machine interface.
No, that’s not exactly right. The real goal is to make this frictionless experience feel natural so that you forget you’re talking to bits of floating code. The older AI constructs may be chess winners, but their descendants are all about natural language skills. That’s not to say you can’t play a game with a smart home assistant. No, your new companion can play games, tell jokes, read poems, and even liven up a dull party.
And what about AI based smart homes
The upside of this kind of seamless interactivity is having things done on a whim. Smart home control is regulated on a whim, as are your entertainment needs. That party you’re having to celebrate your new family member (Your SHA, not a baby) is pulling in the right dance mix. The games interrupt the music, with the virtual home assistant playing “Tic Tac Toe” or “Rock Paper Scissors Lizard Spock.” One of your AI candidates even plays “Simon Says.” Try the game with Alexa. She’ll repeat everything you say when you play this game. Just don’t try to trick her into swearing; she’ll only substitute the swear words with bleeps.
Finally, card games are a natural fit for artificial intelligence. However, the card games and many of the other games mentioned here will require some imagination on your part. It’s not as if your talking box or cylinder can just pull out a board game, after all. Anyway, these are home-located activities, so what could go wrong?
Stuff to be careful about
There is, however, a downside. Put simply, an AI is an equal opportunity servant. The voice-activated persona is extremely smart but not quite smart enough to differentiate different sound wave characteristics, you know, the ones that tell you you’re talking to a different person. Because of this failing, an artificial home assistant will obey anyone’s requests. That simple fact could end up causing some not so simple problems.
There’s a number of Internet pages dedicated to privacy and security. Essentially, these pages are required because Alexa, Google Assistant, Siri, and Cortana don’t discriminate. They hear words and phrases, carry out actions based on those words, and that’s all there is to say about this subject. Imagine setting up all your bank accounts and financial data, then your party distracts you, just for a few minutes, of course. That’s more than enough time for an unscrupulous party guest to swing in and access the sensitive data.
Fortunately, there are ways to guard against such disastrous scenarios. They include 4-digit authentication codes and other measures. Don’t skip the Privacy and Security section, especially if you often invite strangers in your home.
Virtual assistants dress as lone smart speakers. They seem attractive and audio-capable, but, other than that, each one looks basically isolated. That’s not even close to the truth, but it is an appearance that’s encouraged because these devices should be able to fade into the background.
A voice-activated brain lives both inside and outside each synthetic intelligence. Remarkably complex algorithms establish the product’s larger-than-life information-serving credentials, along with plenty of servings of streamed music. From there, the invisible arms and hands of your virtual servant’s smart device brigade are enabling real-world actions, all throughout your home, so that a friendly and frictionless interface is always on call.