"Artificial Intelligence" and "AI" are going throughThird waveHas entered the public's vision. "AI double camera", "AI beauty", "AI translation", "automatic driving")......

AI seems to be omnipotent in some people's eyes, but it is not. In this paper, some principles will be given to determine the circumstances under which artificial intelligence (AI) is useful.

AI is not a "future technology"

Xiaoqiang who is not dead thinks:There is one product that can be called a "future product." He is the kind of product that can no longer be used after it has been used, and is destined to replace the previous generation of products. such as:

  1. iPhone4S has opened the era of retina screens. Since the retina screen was used, it has been unbearable to see the previous "blur" screen, and it is now difficult to see low-resolution mobile screens.
  2. Anyone who has used an electric toothbrush will feel that the non-electric toothbrush is weak. Except for the more expensive you can't pick any ordinary toothbrush. (Electric toothbrushes have an 30% or more cleaning effect than ordinary toothbrushes)
  3. After the emergence of iOS and Android, Saipan went to death. iOS and Android and Saipan are the future products. Similar to the above, there is a huge gap in experience. If you don't want to go back to the previous product, you will be a "future product." (not too much about cost)

But instead of using more advanced technology, the more cool products must be future products. such as:

  1. Compared with fuel vehicles, pure electric vehicles are not “future products”. Fuel vehicles are still used in electric vehicles because electric vehicles and fuel vehicles have different driving experiences.
  2. Compared with ordinary headphones, noise canceling headphones are not "future products". He only experiences better in some scenarios (such as on the plane). But the noise quality of non-noise-reducing headphones is better than that of noise-reducing headphones, so noise-cancelling headphones will not replace traditional headphones.

The reason why "future products" are meant to express a very important point: 

AI is not a "future product". It does not replace the original technology in many scenarios (for example, the simple four-step operation does not use artificial intelligence at all), but in some specific scenarios (the next question will be described in detail). What is the AI ​​scene?) Better than the original technology.

"Muscle T-shirt" or "Jianmei male" or "Kicker"

Muscle t-shirt  

With the fiery heat of artificial intelligence, many people, many companies regard AI as a means of flickering, and everything must be added to AI, so that it appears tall. This is the same as a thin man wearing a muscular T-shirt, without any dry goods.


Some AI applications are mainly used for shows, the most typical is AlphaGo. His marketing role is far greater than direct economic returns and social values. If it’s not Google’s wealth, it’s occasional to show off muscles, and it’s not going to do anything that doesn’t make money. This kind of application is very similar to "Jianmei male". Looking at the very tall, there is actually no use of eggs~


Some AI applications are how to use them, such as speech recognition. Speech recognition does not require any flower racks, it is not accurate, there is no special interaction, there is no special UI, just a word appears a bunch of words. (For example, during the meeting, it is not convenient to listen to WeChat voice, you can voice to text) This application is the "beating hand" of boxing to the meat, using technology to solve the real pain points.

Regardless of how AI is used, you can't escape the 3 class above, and you need to know exactly which type of application you are doing. 

The most feared is that I want to be a fighter in the first place, but in the end I only bought a muscle T-shirt from my comfort. I still have no self-knowledge. I really ran to the ring and the fighter, and the result must have been hit by the ground.

How to use artificial intelligence (AI) to make "future type" + "boxer" products?

There are some conditions for wanting to make a "future type" + "boxer":

  • Enough NB
  • Sufficient money, but in addition to its own conditions, what you do must conform to some objective laws, not in accordance with the objective laws of things, it is likely to take a long time, only to make a "Jianmei male" or "muscle T-shirt."

Xiaoqiang, who is not dead, summed up one"HBI Law"Used to determine if you really need artificial intelligence:

  1. High frequency
  2. Big data
  3. Irregular

High frequency 

It is not necessary to use artificial intelligence (AI) for things that are not high frequency. For example, at the end of the year, it is enough to summarize the 1 times with 1 times. "Manual" is enough. There is no need for "smart". You look at the most useful AI applications (beats) are all high frequency products:

  • Speech recognition (WeChat, translation pen, etc.)
  • Face recognition (security, etc.)
  • Speech synthesis (smart speakers, translation pens, etc.)
  • Image recognition (photographing, identification card, license plate, bank card, etc.)
  • Video recognition (photographing, video recommendation, etc.)
  • Machine translation (going abroad, reading foreign articles, etc.)
  • …… 

Therefore, high frequency use is a necessary condition for the "sniper" application. Things that don't have high frequencies don't need intelligence.

Big data 

Current artificial intelligence is based on big data, and statistics and probability theory are important theoretical foundations. So whether you can "continuously" get "effective" big data is the crucial foundation. For example, in the field of personal health, in the past, we went to the hospital to check the data before we got the data. If we don't go to the hospital for inspection, we can't know if there is any problem with the body. In this environment, AI can't talk about it. In the future, when the bracelet and other biometric products are mature, we can get all kinds of data of our body around the clock. In this case, AI can exert its value, and it may give an early warning before you get sick to prevent the disease from getting worse. 

Therefore, "continuous" access to "effective" big data is an important basis for AI to play its value. There is no artificial intelligence without data.


This is also the biggest difference between artificial intelligence (AI) and traditional computer science. If there is clear and fixed logic, then traditional computer science can solve it, and scenes like speech recognition and image recognition can be solved without a fixed logic. 

For a regular thing, AI is more than one thing, and only in places where there is no regularity, AI is "possible" to play a role.

Final Thoughts

First of all, we must give artificial intelligence (AI) a certainty: He is not a "future technology", and traditional computer science can never be replaced by artificial intelligence. AI is just something that traditional computer science can't achieve under special scenes and conditions. But in many scenarios, traditional technology is still the best choice.

Secondly, artificial intelligence does not have great value in all scenarios, but it can only become a "sniper" (practical and valuable) in certain scenarios.

Finally, what scenarios can AI be very practical and play a huge value? Be sure to comply with the "HBI Principles":

  1. High frequency
  2. Big data
  3. Irregular

Ok, today I say so much, Xiaoqiang, who is not dead, is also a beginner in the field of artificial intelligence (AI). I hope to know more about the AI ​​field practitioners. Welcome to add my WeChat (pkqiang49).