There is no uniform theory on the birth of artificial intelligence, and some scholars began with the construction of the world's first neural network computer by Marvin Minsky, father of artificial intelligence in 1950. In the beginning, Three stages of evolution and advancement Let's review the whole process together and take a look at several technological advances in artificial intelligence speech.
Artificial Intelligence Voice 1.0: A question and a answer
Nowadays, artificial intelligence speech technology is widely used in the field of digital, electronic and industrial manufacturing and has become a popular topic for all, but it still has many immature places when it was developed. There are many discussions on the AI era, At the very level, the artificial voice application that really entered the trial phase of the user is probably about after 2010. The artificial intelligence voice technology of this stage formed the perception state mainly based on voice interaction, which we will call the first phase for now.
At this moment, the artificial intelligence voice application realizes natural speech recognition and semantic understanding at the technical level through the evolution of the algorithm and the input of a large amount of data, and performs data matching on the conversational content and then retrieves relevant topics to achieve simple content one-way question Answer: The first phase of the self-learning, logic, there is a great lack of thrust, can not be the same dialogue on the content of in-depth interaction, horizontal development and vertical development can not meet the needs of users. For example, how do you ask today's weather? Then you ask what is the weather like tomorrow? He calls for tomorrow's weather forecast. Today's weather and tomorrow's weather are only independent answers, can not be connected to form a logic.
As for the research of artificial intelligence speech technology, foreign companies are investing earlier and developing more comprehensively. Like Nuance, they have performed well in speech recognition and are by far the largest speech technology company in the world with the number of patents and market shares far away Leading.After that, including Amazon, Apple and Google, are digging intelligent voice, launched their own voice assistant.According to incomplete statistics, the current global professional artificial intelligence voice company has thousands, as more and more rising star, involving Such as Cortical.io, headquartered in Vienna, Austria, which mainly offers new natural language understanding (NLU) solutions. The KITT.AI company is funded by the Allen Institute for Artificial Intelligence and the Amazon Alexa Foundation to develop Custom hotword detector and chat engine ChatFlow. San Francisco's MindMeld offers interactive AI platforms and more.
Artificial Intelligence Voice 2.0: Questions and Answers
As early as 2000, Bill Gates once proposed that "the next 10 years are the era of voice." We are talking about smart voice now, and academics call it 'natural language processing' (from Baidu) .Industry believes that intelligent voice technology will Will become one of the mainstream human-computer interaction modes, and the development of computer technology and artificial intelligence will bring hope for the development of intelligent voice interaction.
The deeper stages of artificial intelligence voice interaction - dialogue, that is, there are questions and answers, contains the logic of the context.For example, it may be better understood that if a course was not opened the previous year, for this course last year, how many students did not Through such a question, the machine is to answer 'Did not pass' or 'Did not open this course last year?' At the same time the machine also need to store in advance 'did not open this course last year' message, think we did not understand when we were young Now know the knowledge and information, which is unimaginable amount of data. For example, 'I want to hear Jay's song', the machine will only show Jay Chou's songs, and can not be played directly. Things need to give the machine a clear order, otherwise, it will make people very upset, because you have to keep talking.
Of course, at the "dialogue" level, it still stays in the stage of "man-machine dialogue." The machine still stays in the phase of accepting large amounts of data input by humans. Compared with the question and answer level, the flow of information and data is larger and the machine still can not Deeper understanding of human meaning, unable to achieve self-learning, self-growth, voice communication with the machine can not be as natural as people.
Artificial Intelligence Voice 3.0: Natural Interaction
Artificial intelligence into the third stage, the biggest progress is the progress of interaction.Not only have questions and answers, not only contains the logic of the context, artificial intelligence hardware to more integration of various environmental information, to make different decisions or recommendations. Said that in the interactive process, the machine has more initiative, to provide more and better help to make people's lives more convenient, safer and more interesting.
There are two typical scenarios, one is a smart car, one is a smart home.
In 2010, Google's secret research unmanned vehicle project was "New York Times" reporter exposure, was caused no small shock.UAV car or driverless car, also known as smart cars, mainly through the vehicle sensing system to detect the road Environment, and rely on the car's artificial intelligence system automatically planning driving directions to reach the intended destination.
When driving on the open road, Google's driverless car will recognize and analyze the environment information of vehicles, pedestrians and pavements, and provide the basis for the system's decision-making judgment. After long-term learning, the artificial intelligence has the nerve The yuan will become more proficient and efficient, and form a 'feel' similar to people, able to handle unexpected situations in driving.
In January 2016, the U.S. National Highway Traffic Safety Administration (NHTSA) posted an open letter on its website stating that the artificial intelligence system used by Google's driverless cars could be considered 'driver' under current federal law. This is seen as a milestone in the development of artificial intelligence, and also marks that 'intelligence' can sense, judge, reason and make decisions as 'human', allowing for both active control and better human-computer interaction and synergy.
In the field of smart home, the development direction of artificial intelligence is also human-computer interaction, simple operation, big data can make independent decision-making.Chang Hong recently introduced new artificial intelligence TV Q5K, the main semantic understanding and fuzzy search. With clear instructions, AI televisions can recognize and understand the semantics of the user, such as when the user says' too dark ', the TV understands what that means and knows what to do, rather than having the user say' Another example is the recent hit a TV series "that year blooming moon," many names too long to remember, but as long as you say 'I want to see the blooming moon', Changhong AI TV also Can find this film .If even these words can not remember, saying, 'I want to see Chen's TV series', it will also recommend the film based on the hit video data.Even when it comes to' I think Look at Chen Yan Xi husband's TV series', this TV can understand, you said the 80% success is this starring Chen Xiao, recently hit "that year blooming moon" and played for you.
This is equivalent to having the machine really have 'brains' so that it can sense, interact, understand, and make decisions, and thus better serve humanity.
The most cool new media technology information
Sweep won the prize
comment
* User statements are not the site's position, the site is not recommended any comments in the online store, dealer, beware of being deceived!