Alexa, Siri, and Google Don’t Understand a Word You Say

An echo point in front of a command-line prompt
Amazon

Voice assistants like Alexa, Google Assistant and Siri have come a long way in the last few years. But, despite all their improvements, one thing holds them back: they do not understand you. They rely too much on specific voice commands.

Speech recognition is only a magic trick

An echo saying "Hmmm ... I do not know it"
Amazon

Voice assistants do not understand you. Not really anyway. When you speak to a Google Home or Amazon Echo, it essentially converts your words into a string of text and then compares them to the expected commands. If he finds an exact match, he follows a set of instructions. If this is not the case, he looks for an alternative based on the information available to him. If it does not work, you get a failure message like "I'm sorry, but I do not know it. . "It's a bit more than a magical trick to make you think it understands."

He can not use contextual clues to guess, or even understand similar topics to inform his decisions. It is not difficult to trip the voice assistant. While you can ask Alexa "Do you work for the NSA?" And get an answer, if you ask "Are you secretly part of the NSA?", You get an answer "I do not know that one" (at least time of this writing).

Humans, who really understand speech, do not work like that. Suppose you ask a human: "What is this klarvain in the sky? The one who is arched and full of striped colors like red, orange, yellow and blue. Although klarvain is an invented word, the person to whom you asked the question could probably understand from the context that you are describing a rainbow.

Although you can claim that a human converts speech into ideas, a human can then apply his knowledge and understanding to the conclusion of an answer. If you ask a human if he is working secretly for the NSA, he will answer yes or no, even if that answer is false. A human would not say "I do not know that one" to a question like that. That humans can lie is something that comes with real understanding.

Voice assistants can not go beyond their programming

The voice assistants are eventually limited to the programmed expected parameters, and wandering outside of these will interrupt the process. This shows that third-party devices come into play. Usually, the command to interact with them is very difficult to handle, which is "telling the device manufacturer to order an optional argument". An exact example would be, "Tell Whirlpool to pause the dryer." Alexa Alexa Skill controls some GE ovens. A competency user must remember to "say in Geneva" and not "tell GE" and then the rest of the order. And even if you can ask it to preheat the oven to 350 degrees, you can not respond to a temperature increase request of 50 degrees. A human could however follow these requests.

Amazon and Google have worked very hard to overcome these obstacles, and it shows. Where you had to follow the sequence above to control a smart lock, you can now say "lock the door" instead. Alexa used to be confused with "tell me a dog prank", but ask for one today, and it will work. They have added variants to the commands you use, but you must always know which command to choose. You must use the correct syntax in the correct order.

And if you think that sounds very much like a command line, you are not wrong.

Voice assistants are an elegant command line

A command prompt with the search text

A Command line is narrowly defined to perform simple tasks, but only if you know the correct syntax. If you deviate from this correct syntax and type dyr instead of dir, the command prompt will send you an error message. You can use aliases to help memorize commands, but you need to get an idea of ​​what they were, original commands, how they work, and how to use them effectively. If you do not take the time to learn the nuts and bolts of the command line, you will never get much out of it.

The voice assistants are no different. You must know how to pronounce an order or ask a question. And you have to know how to create groups of Google and Alexa, Why grouping your devices is essential, and how to name your smart devices. If you do not follow these necessary steps, you will feel the frustration of asking your voice assistant to turn off the study, but you will be asked "which study" should be disabled.

Even if you use the correct syntax in the correct order, the process may fail. Either with a wrong answer or a surprising result. Two Google homes in the same house can give the weather to slightly different places, even if they have access to the same user account information and the same internet connection.

In the example above, the command "Set a timer for half an hour" is given. The Google Home Hub has created a timer named "Time," then asked how long it should last. And yet, repeating the same command three more times worked fine and created a 30-minute timer. The use of the "Set timer for 30 minutes" command works properly in a more consistent manner.

While the conversation with Google Home or Echo may be more fluid, voice assistants and under-the-hood command lines work the same way. You may not need to learn a new language, but you must learn a new dialect.

Close understanding of voice assistants will limit growth

A Google home hub and an Echo spot in front of a smart socket and a light bulb

None of this prevents voice assistants like Google Assistant and Alexa from working well enough (although Cortana is a different story). Google Assistant and Alexa are looking for online questions in a decent way, although not surprisingly, Google is better at researching and can answer basic questions like metric conversions and simple calculations. With a properly configured smart home and a trained user, most smart home controls will work as expected. But this has come through work and effort, not intellectual understanding.

Timers and alarms were simplistic. Overtime naming has been added, then the possibility to add time to a timer. They went from simplistic to more complicated. Voice assistants can answer more questions and each day brings new skills and features. But it is not a product of personal growth that stems from learning and understanding.

And none of this offers the inherent ability to use what we know to reach the unknown. For every order and question that works, there will always be three that will not work. Without a breakthrough in A.I., which gives humans a capacity for understanding, voice assistants are not assistants at all. These are just voice command lines – useful in the right scenario, but limited to the scenarios for which they were programmed.

In other words: machines learn things, but can not understand them.

RELATED: The problem of artificial intelligence: machines learn things, but do not understand them

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.