There is still a long way to go, counting the ten crimes of artificial intelligence television

Artificial intelligence has continued to heat up. Since last year, it has penetrated into the field of smart TVs. In the first half of this year, artificial intelligence televisions broke out. More and more televisions are tagged with "artificial intelligence," as if the traditional smart TVs evolved overnight. For artificial intelligence TVs, however, artificial intelligence is just another concept of being over-consumed.

What is an artificial intelligence TV? At the 2017 Artificial Intelligence TV Industry Summit held in Beijing some time ago, the “2017 Artificial Intelligence TV White Paper” released jointly by the China Electronic Chamber of Commerce and Jingdong Home Appliances gave a definition of “accepting user instructions through sensors and relying on basic applications and basics”. The data platform recognizes and understands the user's purpose, makes feedback and processing, and relies on sound technical logic and big data operations, combined with deep learning algorithms, analyzes user habits, and masters user preferences and behavioral logic to achieve more precise interaction and personality. Content recommendations." From this definition, it can be seen that the algorithms, hardware, and technology required by artificial intelligence televisions need time to accumulate. There are certainly many problems with so many artificial intelligence televisions in such a short time.


1, still staying at the level of voice interaction
Throughout most artificial intelligence TVs, voice functionality is a must, but AI TV reception instructions do not rely solely on speech recognition, but also in-depth interactions such as somatosensory recognition, face recognition, and emotion recognition. Some artificial intelligence TVs only have the ability of voice recognition. For smart TVs, this is not a fresh feature.
About 2-3 years ago, some TVs began to have voice capabilities. However, the concept of artificial intelligence was not so hot today. The voice function is not the selling point of the main push. With the rise of artificial intelligence, TV manufacturers have emphasized the importance of intelligent voice technology. Although they have all been labeled as "artificial intelligence," the perceived intelligence and previous smart TVs have not significantly improved.
2, many voice features are not practical enough
Although each of them is actively enriching the voice function, the most obvious improvement of the TV control experience after the voice is added is only in search. In the past, inputting on the TV with the remote controller was very tedious. Now, it is OK to say what you want to see. As for most other functions that are optional, they look very good, and they are basically useless.
For example, search for “×××” Episode X. Where can I remember the last episode? Adjusting the progress of the play and the volume level is actually better than using a remote control, or setting a reminder on the TV. Why don't I use a mobile phone? Chatting with a TV conversation is fresh but it loses interest after being used several times.

3, speech recognition accuracy rate is affected by external factors

With regard to some artificial intelligence TVs that have been experienced before, voice recognition capability cannot actually achieve significant gaps. The accuracy rate can reach at least 90% or more. If you are talking, basically all TVs can understand and can be controlled in a realistic use scenario. There are many, such as dialects, environmental noise, the speed of speech, etc., will have a certain impact on accuracy.

For example, our country has complex dialects and local accents. However, only some televisions currently support a few dialects. Even if it is putonghua, the use of other people's speeches in the process will have a greater impact on the accuracy of recognition. , It is not suitable for the whole family to sit and chat while voice control TV.

4, there is much room for improvement in semantic understanding

Voice recognition is the most basic in the process of intelligent voice interaction. Excluding external interference factors, the accuracy rate is relatively high. To tell a TV the name of a specific movie or TV show, the TV will automatically start searching after recognition. If you start playing, you can ask questions such as "Dong Dongying's TV series" and "Oscar's Best Picture last year." This is the test of TV's semantic understanding ability. It requires precise analysis and in-depth understanding to give a satisfactory answer. With high technical requirements, there is much room for exploration.

5, full voice control can not be achieved

Constrained by the unity of the content platform, smart TVs always have limitations on content and have to install third-party applications to expand content. However, voice function does not work for them. Content-based search ultimately targets only the built-in video platform of the TV. The content, after entering the third-party applications, all the operations still need to be completed through the remote control button; the same is true after the signal source switching, such as when watching traditional cable TV, the common channel switching operation still depends on the remote control set-top box, voice The function has become a decoration.

Therefore, there are many limitations in the use of smart voice currently, can not be used throughout, it is difficult to cultivate the habit of voice control, even if it is to support the far-field voice of the TV is also the case, want to completely get rid of the remote control is unlikely, making the usefulness of voice features Greatly reduced.

6. It is difficult for the whole family to have thousands of people.

Compared to mobile phones with personal attributes, television is shared by the whole family. As an interactive artificial intelligence TV, combined with the definition at the beginning of the article, it needs to “analyze user habits and master user preferences and behavioral logic in order to achieve more accurate "Interactive and personalized content recommendation", however, different family members have different viewing habits, and TV content recommendation based on the captured viewing habits is certainly not accurate enough, perhaps relying on voiceprint recognition technology to achieve the "thousands of people" of the family. .


7, will use more intelligence
Artificial intelligence TVs need to be evolved through constant learning. For example, the above-mentioned content recommendation “Thousands of People” requires a period of continuous use to enable televisions to possess such capabilities. In other words, each use is actually data. Accumulation process, then TV will be based on the viewing habits of accurate personalized content recommendation, as to how long it takes to be aware of it's smart recommendation is not easy to say, and television is a life cycle, after the device replacement TV can Whether or not the accumulated data can be migrated is not certain.
8, there is always a short board
No matter what stage the artificial intelligence television develops to, the content is indispensable, after all, the television must satisfy the demand that sees first, but now there is not a television that can include video, music and other content of all platforms.
For example, searching for a movie just happens to be a single video broadcast by TV, and the intelligence level of TV is too high to be able to do anything. The artificial intelligence experience eventually felt will also be greatly reduced. Therefore, for the television industry, it is necessary to enrich content resources while advancing artificial intelligence.
9, higher artificial intelligence is still in the laboratory
Having said so much, most of them are actually focused on voice interactions and still remain in the primary stage of artificial intelligence TVs. What kind of artificial intelligence TVs are still exploring experiments in the future?

Millet TV mentioned in the 4A series of new product launches that an artificial intelligence computer vision team has been established to carry out research on face algorithms, image understanding, etc. The scene uses the movie “From Your World Passing” as an example to demonstrate some of the application scenarios. During the viewing process, you can ask questions such as who is this, where is it, and what mobile phone is.
Microwhale TV mentioned that in depth cooperation with Microsoft and Baidu in the field of artificial intelligence will be in depth in face recognition, voiceprint recognition, emotion recognition, voice response, image recognition, video detection, etc. Explore, artificial intelligence services will continue to superimpose in the future.
10. Become an obstacle to entry of smart homes With the in-depth development of smart homes, TVs with large screens naturally have the advantage of smart home entrances, and TVs can be interconnected with other smart devices in the home, such as televisions, air conditioners, refrigerators, and water heaters. However, this requires the support of a unified control protocol. Judging from the current development situation, each major appliance company is fighting each other. The equipment that can be accessed is very limited, and the advantage in smart home entry competition is not obvious.

Artificial intelligence TV has just emerged. For now, the combination of TV and artificial intelligence is far from expected. Although artificial intelligence has injected fresh blood into smart TVs, the power to bring about changes in the TV industry is still very limited. The more televisions that use "artificial intelligence" tags, don't expect too much. The real artificial intelligence TVs still have a long way to go!
Smart TV/box information can focus on smart TV information network sofa butler (http://), China's influential TV box and smart TV website, providing information, communication, TV boxes, smart TVs, smart TV software, etc. Answering questions.

Ego E cigarette

Suizhou simi intelligent technology development co., LTD , https://www.msmvape.com