Time：12/5/2019 6:59:20 AM
AI Multi-modal Technology is the key in the area Strong artificial intelligence.The three development stages of industrial intelligenceComputational intelligence: Machines have the memory and computing power of humans to store and process vast amounts of data. Can help humans to complete large amounts of storage and complex calculations. This step is the basis of perception and cognition.
Perceptual intelligence: Machines have the same perceptual ability as human beings. Through organs such as "eyes", "ears" and "skin", machines also have the perceptual ability of vision, hearing and touch. To help humans do the simple tasks of seeing and hearing.
Cognitive intelligence: The ability of machines to learn and think like humans, to make decisions and take action on their own, and to replace humans in part or in whole.Artificial intelligence is at the stage of weak artificial intelligence
At present, the development of artificial intelligence is still in the category of weak artificial intelligence, mainly solving the problems of computational intelligence and perceptual intelligence, but there is still some way to go before it reaches the era of strong artificial intelligence that can think independently and make decisions.
Weak ai is mainly specialized ai that is good at a particular field, such as speech recognition and image recognition, which focus on computing power and data accumulation.
The single mode interaction technology is a typical representative feature in the era of weak artificial intelligence.Limitations of single mode interaction techniques
At present, perception technology, as the most important part of artificial intelligence, has been widely used in recent years. Although the single-point AI sensing modal capability of machines is gradually improved, voice interaction, machine vision and sensor intelligence technologies have achieved great development. However, single-point AI modal perception has a great limitation in product application.Single mode signal input, lack of initiative service ability, let the machine and misjudgment instructions, mistakes often wake up, all affect the quality of the user experience ascension.
For example, smart speaker, as the hottest intelligent machine at present, after a wave, people will find that simple voice interaction can no longer meet the higher demand of users for intelligent machines.The trend of multi-mode interaction under strong artificial intelligence
Strong artificial intelligence mainly refers to the general intelligence that can learn knowledge, think about problems and solve problems. Machines have multi-channel perception and consciousness, and can match humans in every aspect.
If artificial intelligence needs to have a better experience in life, we want it to be able to work together with eyes, nose, mouth and ears, and provide services actively, just like the human body.Multi-mode interaction technology is the key to realize the above requirements.
Today there are three main AI perception modes:
Voice interaction (including voice command control, semantic understanding, multi-round dialogues, NLP, accurate speech recognition, etc.);
Machine vision (including natural object recognition, face recognition, body movement recognition, etc.);
Sensor intelligence (including AI's reading and understanding of heat, infrared capture signals, spatial signals, machine vision, and sensor intelligence).
In the era of strong artificial intelligence, multi-mode can combine these three modes of perception, and Internet of things devices can observe with cameras and judge with sensors in addition to simply listening and speaking. By integrating different signal modes, the multi-dimensional perception and thinking of the equipment are realized, and the intelligent machine moves towards humanization and activeness of service.Actions technology's first multi-mode interactive chipAfter half a century of development, in the last decade, artificial intelligence technology has stepped into the fast track of development. High-performance computing chips and even AI chips, massive data accumulation and excellent software algorithms are pushing artificial intelligence to further development.Shensi technology has many years of experience in the design of single mode interactive chips, related products are widely recognized by the market. In the face of the era of artificial intelligence transition, we are also in the development of further technology extension.
This time, torch core launched its first multi-mode interactive chip, ATS 3609D, multi-mic intelligent voice, light intelligent image, finger read dual mode recognition input solution. ATS 3609D integrates voice interaction, machine vision and sensor modes to provide a feasible solution for multi-mode interaction under strong artificial intelligence.△ATS 3609D （reference object：A dime）
ATS3609D Key Feature
● dual core architecture supports higher computing power
● DSP NN LIB, improve the efficiency of the algorithm
● voice, finger, LCD feedback, touch panle assistance, multi-mode interaction
● full-format picture decoding, mainstream SWF format decoding, improve audio and video player, support video call live, specially designed for electronic educationThe future and prospect of multi-mode interaction technology in the era of strong artificial intelligenceWe can find that more and more multi-modal interaction technologies are being applied to various scenarios, breaking through the traditional single-point perception of intelligent machines, realizing multi-dimensional perception and thinking of devices, and enabling intelligent machines to move towards humanization and activeness of services.AI multi-mode interaction technology has become an inevitable trend of artificial intelligence interaction. We also expect that the ATS 3609D chip of torch multi-mode interaction platform can be applied in more scenarios, endow the machine with stronger intelligent properties, and bring better services to everyone with more human-like intelligent machines.