Voice assistants are becoming increasingly commonplace in our homes and pockets.
The fear that they are always listening to you and tracking is one that is also growing.
But let’s take a step back and first understand, how do these assistants are able to listen.
For a voice assistant to listen they first need to understand natural language.
Natural language refers to ‘human’ speech, the words people use to communicate with one another.
The concept of artificial intelligence through an understanding of natural language was first developed by Alan Turing.
What are voice assistants?
Alexa, Siri and google home are all examples of voice assistants.
These devices and technologies can answer questions for you, schedule memos and even book appointments on your behalf.
These assistants are activated by your voice (surprise surprise ;p) and are able to follow the instructions you give them.
Using your voice you can efficiently manage your schedule, and always know who that actress was in that film about that thing.
How do these technical assistants work?
Many different technologies go into making an assistant like this work.
I’m going to focus on the machine learning techniques that allow the assistant to understand what you are asking.
The two processes are:
- Natural language processing
- Reinforcement learning
How is artificial intelligence used to improve voice assistants?
Let’s first talk about natural language processing.
Natural language processing allows the voice assistant to take the instruction you gave it and then convert this into a code it understands.
This process involves breaking the language used down and running algorithms on it to understand the intent.
You can read more about the technical details of natural language processing here.
How is reinforcement learning to train the technology?
Reinforcement learning is the second step in developing a successful assistant. Once you have a system that can process language you then need to train it to understand speech.
Unlike language, speech allows a person to express the same sentence in a variety of ways, dialects, and accents.
This adds a new layer that a voice assistant is required to understand.
Do you remember when you first started using Siri on your iPhone, and it didn’t understand a thing you were saying?
Over time, however, Siri learned to understand you and was able to support you with all your questions about cats, the weather and who it was that sung that song.
This isn’t a process that happens by accident.
How the technology learns with reinforcement learning
Your voice assistant is using reinforcement learning to learn how to understand your speech so that it can be processed.
Whenever your voice assistant asks if they got it right, you are giving him/her a signal by responding yes or no that helps them to learn.
This is very similar to how a person learns.
You can learn more about the theory behind reinforcement learning here.
Which is the best performing voice assistant?
Now that we know how they work the next thing to understand is which assistant works best!
There are three principal voice assistants that people commonly use today. These are Alexa, Google Assistant, and Siri, from Amazon, Google, and Apple respectively.
I have not included Cortana, Microsoft’s offering, for now as they have openly said they are scaling back out of this market (link).
The review of voice assistants has taken feedback from my personal experience and dedicated review sites I have found.
So who is the best of the assistants?
Well, it really depends on what you’re after.
I know that sounds like a cop-out but let me explain why.
There are on average 3-4 main functions your voice assistant runs for you.
- Answering questions
- Performing admin tasks
- Streaming music
- Running your smart home
Depending on which of the above 4 uses you prioritize, will dictate which will be your perfect voice assistant.
The first and most important thing your assistant needs to do is understand you. This could substantially limit your options.
Understanding language – who does best?
Siri speaks ~20 languages. This makes Siri the most widely versed assistant at this time.
The second is google with ~4 (but many more coming) and finally Alexa with ~3.
Once your voice assistant understands your language, you are ready to start using it.
When it comes to answering questions accurately, Google consistently comes out on top.
There are plenty of stats available on which voice assistant performs best under questioning here but to summarize Google is about 85 – 90% able to answer questions and accurate.
The next area where you would want to use your assistant for is admin tasks.
Which assistant is best connected?
If you are wanting to complete tasks with your assistant that involve other service providers, then Alexa is your best bet.
Alexa is connected and compatible with far more devices than any of the other voice assistant.
Moving onto streaming music, Alexa has also been identified as having the best speaker.
Finally, when it comes to running your smart home, it’s pretty much a tie.
That is assuming if you use Apple Home Siri is your assistant.
Google and Alexa are both more than capable of keeping your nest snug and working with smart home enabled devices outside of Apple Home.
So that’s all on these assistants. I’m curious to know if you use one? If so, what for? Let me know in the comments.