The AI voice assistant is getting more and more advanced, but it will read it, are you afraid?
Editor’s note: This article is from WeChat public account “Netease Smart” (ID: smartman163 ), the author of NetEase smart.
Since Siri first appeared on the iPhone 4S in 2011, the voice assistant has changed from a flashy gimmick to the foundation of smart speaker technology. What kind of height will the future AI voice assistant evolve into?
Today, smart speakers have penetrated one-sixth of the US family.
Tom Hebner, innovation director of cutting-edge artificial intelligence voice technology developer Nuance Communications, told Digital Trends: “Before Siri appeared, when I talked about (what I did) things, people’s eyes It’s all confusing. People will say, ‘Are you developing terrible mobile phone systems? I hate you.’ At that time people responded to voice technology.”
But today the situation is completely different. According to market research firm eMarketer, by 2020, nearly 100 million smartphone users will use voice assistants. Although artificial intelligence assistants are no longer new, we are still in the early stages of evolution of this type of technology. To fully realize the potential of voice assistants as a product category, there is still a long way to go.
At present, AI helpers can be smarter and more efficient in five ways to help us lead a productive life. Let them call them “predictions” or “wish lists”. After all, they are all challenges that need to be solved.
01 Enriching the professional knowledge base
Amazon’s AI voice assistant Alexa can tell you a variety of things, such as the weather conditions in Kuala Lumpur, Malaysia; a total of 720 South African Rand can be exchanged for $ 720; how to spell the complicated and complicated words. But in essence, consumer-grade artificial intelligence assistants are equivalent to a person in the digital world who has a complete set of the latest encyclopedias. You get the right information from them, but they don’t have the expertise at the professional level.
Herberner is accepting DIn an interview with igital Trends, “The challenge for smart devices in your home is that they try to solve things that are too broad for you.”
This is a difficult problem to solve, but once it is resolved, it can have a disruptive effect. Nuance has developed a number of expert-level systems for specific use cases, such as helping airline customers answer customer questions or helping doctors take notes. Not only does this mean that these systems can get more detailed information in depth, but it also means they can generate more intelligence. “People are very excited about computers that understand the meaning of words, but if the computer doesn’t know how to respond, it doesn’t necessarily make sense,” Hebner said.
He cited an example: a Nuance system not only understands the list of potential drugs that doctors talk to patients, but also identifies potential conflicts between those drugs. This goes far beyond the capabilities of most consumer-grade artificial intelligence assistants.
However, AI assistants have more specialized and detailed knowledge in different areas – Alexa skills suggest this – and may be revolutionary. On the surface, it may sound crazy to seek legal or medical advice from your smart assistant. But the industry has made extraordinary progress in areas such as legal robots. A recent report also shows that Apple hopes that by 2021 Siri will be able to conduct health-themed conversations with users.
The Expertise Assistant’s Expertise map is a sci-fi story. The recently released Voicebot.ai report also shows that the virtual assistant’s skill set is rapidly expanding. As they enter various fields of expertise, we will benefit a lot!
02 Strong personalization features
The personalization of today’s smart speakers is still in its infancy. You can change the voice assistant’s accent and gender performance, add or remove skills for it, and enter some information, such as your name and work location. In some cases, you can set up multiple voice profiles so that Google’s Google Home can identify individual members of your home.
Amazon Echo Show Smart Speaker
But smart assistants have a lot of room for personalization – but efforts in this area are still worthwhile. Mattersight has developed an artificial intelligence call center technology called Predictive Behavioral Routing that analyzes the caller’s speech patterns and matches them to a character-compatible human operator. According to the company, the human operator who matches the caller’s personality is conducive to smooth communication, and the call time can be cut in half compared to operators who match the conflicting personality.
Using a similar approach may make the AI Assistant talk to you the way you like. For example, it provides simple functions like matching the accent and volume of a conversation object. Or, the smart assistant can change its expression, using some of the more emotional words to express feelings, rather than using more detailed information as other users. Maybe some people want to have a long conversation with the voice assistant, while others just want the voice assistant to convey the necessary information in the most concise way. The AI Assistant should have both capabilities.
Technologies like Google Duplex show how accurate synthetic speech and conversations generated by artificial intelligence are becoming. As AI involves more complex tasks than responding to song play requests and setting food timers, this technology is expected to play a more important role in our daily lives.
This is expected to be achieved with breakthroughs in speech recognition technology. Hebner pointed out that Nuance’s technology can identify users with only one second of audio. “In the past, speech recognition technology took 10 seconds to know who you were in order to get an accurate signal.” He said, “Today, this technology is very powerful.” This technology can identify users with a small amount of speech. It solves the password problem, which means you can choose to use the voice assistant to get relatively subtle confidential information.
03 Active from passive
Excellent assistants will do things for you when you ask for them, and great assistants don’t need your request, but can proactively help you solve problems. Currently, the artificial intelligence assistant is still in the first stage. Users can get the songs they want and get the reminders they need, but usually only after they have explicitly requested the voice assistant. As people become more accustomed to using voice assistants, they have a great opportunity to move from a simple passive device to a proactive device.
How do you think the artificial intelligence assistant makes decisions on your behalf? These suggestions can be: when someone says theyWhen you feel cold, turn the thermostat on; or because you are late, reschedule a lunch meeting; urge you to exercise more, or save money. As more and more smart devices enter the home, the things that voice assistants can control will increase dramatically.
This is a social problem to a certain extent, that is, how safe people are to make decisions on the machine. Are people willing to hand over certain tasks to the machine? There are still big questions about this. Think of it as giving your credit card and house key to your human assistant. The downside is that you have to give up a certain degree of control. The potential benefit is that you will have more free time. Of course, this is a huge technical challenge…
04 Feedback Mechanism
Hebner points out the challenge of giving voice assistants the initiative: How do our machines know if they are right or wrong? Speaking of the concept of “excellent assistants” and “excellent assistants”, a great assistant will take all your documents before a large meeting without asking for your request. But what if the file is not right? A big problem with making home artificial intelligence assistants more proactive is that there are only limited ways to reveal whether the information we get is the information we need.
Pepper Humanoid
“If I go to the house every day and listen to the same song, then one day I walk into the house, the smart assistant will start playing. How does it know that I have chosen the right song?” Hebner said, “If I If you don’t stop playing, does that mean it is right? If I say “stop playing” with it, does it mean that it is wrong, shouldn’t play the song again? The feedback mechanism is missing, it is you One of the reasons for not getting a more proactive smart system.”
This is a challenging issue for engineers. Anyone who has had an experience with interns asking for guidance and feedback on each task knows that sometimes it is easier to do a job yourself than to delegate someone else. AI assistants should be able to make your life work smoother; instead of letting you do dozens of small surveys every day to see if it works right. This will need to be addressed in a way that does not compromise the user friendliness of these devices, and does not require extensive pre-training until the system understands your preferences.
05 New interaction method
The science fiction film “2001 Space Walk” (2001: a Space Odyssey) has such a scene: the ferocious HAL 9000 – disturbingly, it is still the most famous fictional artificial intelligence assistant in history – revealing that it Not just use the microphone to judge what others have said to it. When the two crews tried to choose a place where they knew HAL 9000 couldn’t hear, HAL 9000 revealed that by reading their lips, they could still know what they said.
It’s terrible? Future artificial intelligence assistants may be able to do the same.
The voice assistant should be limited to the idea of voice interaction, reducing the potential ways in which they interact effectively with us. With the increasing popularity of facial recognition and emotion tracking technology, more and more biometric data is being collected, and reading technology is even possible. Artificial intelligence assistants can use a large number of different signals to draw conclusions.
After 10 years, we will only be able to use voice to control the ideas of these artificial intelligence assistants. Just like when we saw a personal computer in the early 1980s, we thought, we will never have another keyboard. Computing device. (Lebang)