Where are smart speakers smart?

I believe that many people are familiar with the word "smart speaker", but what is the core value of smart speakers? Different people have different answers. From the perspective of skills, a smart speaker is a life and entertainment assistant that can listen to music, ask about the weather, and ask for routes. It is also a smart assistant that can control home appliances.

How far are we from Iron Man’s intelligent assistant Jarvis?

There are many smart speaker brands on the market, which can be classified from two dimensions: intelligence level and industry application:

1. There are three levels of intelligence: non-intelligent, scene linkage, and intelligent

1) Non-intelligent: Many Bluetooth speakers are sold very well under the banner of smart speakers. First of all, Bluetooth speakers do not have microphone arrays, so there is no "sound pickup" link. Therefore, users cannot interact with them and can only use them as speakers for mobile phones.

2) Scene linkage: The intelligent attributes of the scene linkage here are not strong, and it mainly relies on fixed instructions to achieve automatic response.

3) Intelligence: Intelligence refers to strong technology in speech recognition, semantic recognition, natural language processing, etc., and a long period of training. The speaker truly becomes "conscious" after multiple interactions with users.

2. Industry applications have two directions: consumer level (toC) and industry level (toB)

1) Consumer level: The main skills include listening to music, asking about the weather, alarm reminders, etc., as well as chatting when bored. It is positioned as a life and entertainment assistant.

2) Industry level: Customized functions and services in certain industries. The main skills include intelligent customer service, intelligent customer control, etc., such as finance, government affairs, hotels, real estate, etc.

Judging from the above classification, a true smart speaker is based on artificial intelligence technology, and its skills should not be too many but must be refined.

1. AI-based

The core of a true smart speaker is not the hardware itself, but artificial intelligence technology, namely speech recognition, semantic recognition, natural language processing and other technologies.

2. Massive data samples

How to make smart speakers smarter, in addition to using artificial intelligence technology, is to constantly teach them to speak. As long as there are massive data samples, the smart speaker's ability to understand will improve. This is the same principle as teaching a child to speak.

3. Skills must be refined

Focus on core skills and continuously strengthen them, rather than teaching many skills that are difficult to digest. This is the same principle as "you can't become fat in one bite" and "eating too much will cause indigestion."

Currently, most of the real smart speakers are made by large companies. The earliest ones are Amazon's Echo and Google's Google Home. Domestically, there are Alibaba's Tmall Genie, Baidu's Xiaodu speaker, etc. The price of the version with a screen is less than 1,000 yuan, and the version without a screen is less than 500 yuan. The cheapest one is already in the 100 yuan level.

The author has basically experienced all products that can be called "smart speakers". On the whole, the overall experience is average. It is normal to be called mentally retarded by users. Indeed, there are problems with understanding many times. However, now it can only be regarded as the "infant stage" of smart speakers. When they really grow up, their skills are mature, and they learn more, the experience will naturally be better.