Who would not love to have personal assistants who not only keep track of your commitments and appointments but gets to know you better over time? And what if that assistant could simultaneously manage a range of tasks in your household? This scenario is no longer fiction. Smart home devices like Amazon Alexa have turned the dream of the all-seeing, all-knowing virtual assistant into a reality.
Alexa certainly makes use of advanced technologies to make her owners’ lives easier. The explosive growth in connected products is only one indicator of the success of digital voice assistants. But does Alexa represent artificial intelligence (AI) technology? Here is a closer look.
What Is Alexa?
For those readers who are not using Alexa yet, let us introduce this smart voice assistant. Alexa is nothing other than voice-controlled digital assistant software. She answers questions through Google searches, can order items online, or accept voice commands to create to-do lists.
Alexa also helps you set reminders for important appointments. As a tool, this virtual assistant is now closely integrated with other Amazon products, like the company’s smart speakers Amazon Echo and Dot. Alexa integrates with other smart home devices, too, allowing owners to control anything from lights to room temperature, and their favorite music through voice commands. Those are known as Alexa-enabled devices or Matter devices and include Amazon Fire TV.
Alexa AI Technology
Alexa is not the only smart home assistant available to consumers today. Like her smart device ‘colleagues,’ Google Assistant, and Siri from Apple, Alexa responds to user requests and executes those. To do that, all virtual assistants depend highly on AI technology. They also utilize automation and machine learning (ML).
AI is a large area of computer science. The term covers different fields that have one thing in common: they are home to scientists aiming to create systems that mimic human behaviors, such as learning. ML is a subfield of AI that is specifically concerned with enabling computers to learn without repeated human intervention. Machine learning models allow computers and other devices to learn from interactions between machines. Thanks to AI, computers are now able to complete tasks that require reasoning and perception without additional human involvement.
Since its inception, Alexa has become gradually smarter. She now understands follow-up questions in the context of previous questions and can handle different requests simultaneously. As a system, Alexa has become better at understanding its boundaries and knows when to ask for human help. As a result, error rates have dropped, and Alexa has become more usable.
These improvements are especially noticeable in Alexa’s automatic speech recognition and natural language processing (NLP) ability. In short, Alexa is becoming the perfect virtual assistant, making the technology a little more indispensable every day. Gone is the need for an artificial language. That change has automatically improved the customer experience.
Is Alexa AI?
With all that said, the question remains whether Alexa and her virtual assistant colleagues can be considered AI. They are certainly impressive examples of AI applications that are transforming our everyday lives.
Just a few years ago, we simply did not have access to technologies of the same level of sophistication. Computers that fit into a home were not able to process the amount of data AI needs to be effective. Thanks to unprecedented advances in computer technology, areas like natural language questions, natural language understanding, and generation are now some of the more commonly used subsets of artificial intelligence.
Does that make Alexa an AI? Not necessarily. Instead, Alexa is a system whose ability to function as an efficient assistant to its owners thoroughly relies on several forms of AI.
How Alexa Uses Artificial Intelligence
One of the biggest advantages of artificial intelligence technology over human intelligence is the capacity to process and analyze unimaginable quantities of data precisely. When it comes to processing information, filtering relevant from irrelevant parts, and then modeling scenarios, technology simply works better than people.
Alexa uses these capabilities to understand human requests and answer them appropriately. Her ability to respond to a request is not limited to a certain number of possible replies. Instead, she utilizes NLP to process human speech, put that voice command into context with previous inputs and reply appropriately.
One of the ‘skills’ that made these developments possible is Alexa’s relatively new ability to carry context from one query to another. At the same time, the system has also learned how to separate tasks and pursue multiple tasks at the same time.
Alexa’s Advancing AI
Alexa’s abilities are growing constantly. One of the major transformations was triggered by the introduction of transfer learning. Thanks to this development, Alexa’s error rate dropped by 25% over the course of 12 months. Transfer learning is essentially a form of deep learning.
Deep learning is one of the more recent advancements in machine learning technologies. Deep learning technologies are modeled on the neural networks of the human brain. These artificial neural networks are able to model different domains and then transfer what they have learned to a new skill or area of knowledge.
Another recent change in how Alexa handles queries and uses AI was the launch of self-learning technology. As the name suggests, self-learning allows the Alexa system to correct itself based on the clues the virtual assistant receives from the context of a query.
Here is a simple way to picture that skill: let us assume you are asking Alexa to play a certain TV channel. The request fails because Alexa stores the channel under a different name. When you are asking again, this time using an alternative name, Alexa understands that both are the same and changes the way its catalog works. The next time you ask using the first name, Alexa can respond correctly.
In 2020, whilst much of the world was preoccupied with the coronavirus pandemic, Amazon used the time to push ahead with more transformational changes to how Alexa was working and integrating with other Amazon devices. The complete list of compatible devices continues to grow. Talking to Alexa became more natural, requiring fewer “Alexa” commands, and Alexa started employing natural turn-taking.
The system also improved its conversational experiences and learned how to speak to multiple people at the same time. Alexa also gained the skill to learn from its users, contributing hugely to improvements over the following years. By having a conversation, users and Alexa effectively engage in interactive teaching.
Out of all recent advancements, conversational AI stands out among the rest. We may all have gotten used to touchscreens and mouse clicks but nothing beats a simple conversation when it comes to human-machine interaction. That is why voice technology and voice-user-interface services like Alexa Voice Services continue to grow in popularity. Using them simply feels effortless. Some of the key components of these interactions include variety, context, engagement, tone, and memory.
Conversations are more than an exchange of pre-scripted questions and answers. Humans talk to each other naturally. Their conversations change direction, and they use different phrases to express similar concepts.
Today’s Alexa can participate in these natural conversations and customer interactions.
Our everyday conversations take place within a certain context that goes beyond simple voice recognition. We expect that the person we talk to understands this context and integrates it into their reply. Common contextual areas include knowing why, when, and where something is meant to happen. Recent developments in Alexa’s abilities have enabled the assistant to add context to requests.
Human language is inconsistent. We would not naturally choose to structure our requests in a certain way. Here is an example: imagine you are asking Alexa to make a reservation at a certain restaurant at a specific time tomorrow for a set number of people. Chances are you include all this information in one sentence and expect Alexa to pick up on all of it.
If that happens, Alexa appears engaged in the conversation. However, if the virtual assistant were to reply with repeated questions covering the same ground, it would soon lose some of its usefulness.
Face-to-face conversations contain far more than the words people are using. There is emotion, facial expressions, and surprises that even video calls struggle to capture. Being able to integrate these human specifics into replies has made Alexa seem more natural.
If you are talking to a friend regularly, you expect them to remember what was said last time. After all, conversations do not happen in a vacuum. Alexa has got better at ‘retaining’ previous information, making the system a more sophisticated conversationalist.
Amazon AI is not only changing the way we organize our lives. It is also transforming computing experiences. As machine intelligence and customer feedback improve machine intelligence-driven devices like Alexa, more commercial device makers are likely to enter the market.
By putting the customer at the heart of each development, Amazon Science’s machine learning team is changing the way we use technology. Customer-centric thinking does not stop at the language aspect but includes the creation of devices that do not rely on a user’s programming experience.
These topics, as well as privacy implications as a result of connectivity, will be part of the discussion at the Project Voice event in Tennessee in April 2023.