What Is a Chat Bot?

A chatbot is a computer program that talks via dialogue or text. Able to simulate human conversation and pass the Turing test.

A chatbot is a program used to simulate human conversation or chat. "Eliza" and "Parry" were very famous early chatbots. It tries to build programs that at least temporarily make a real human think they are chatting with another person.

Chatterbots have been used in the online interactive game Tinymuds. A single player can interact with a chatterbot while waiting for other "real" players. At least one company is currently making a product that will allow you to build a chatterbot that can be used to grasp relevant markets or other questions about your website. It's not hard to imagine two chatterbots talking to each other or even exchanging information about themselves. As a result, their conversation will become more complicated. (Looking at artificial intelligence will know what kind of panic this idea will cause if implemented.) Of course, they can use more general chat acronyms.

To put it plainly, chatbots

Early classic chatbots were ELIZA (1966) and PARRY (1972). Notable recent ones include ALICE, Jabberwacky, and DUDE. Although ELIZA and PARRY are only used to simulate written chats, many chatbots now include gaming and web search capabilities. In 1984, "Policeman's Beard is Half Constructed" was published, allegedly by the chatbot "Racter". ^[1]

TalkBot

Originally as an online chat system, TalkBot was written in javascript and PERL in 1998 by Chris Kwalt, and twice won the "Chatterbox Challenge" championship in 2001 and 2002.

Elbot

After the improvement of the German chat robot Charlie's program, Albert was born. The German version of Albert started online chatting at the end of 2000, and even the English version was available in 2001. Won the "Chatterbox Challenge" championship in 2003.

eLise

German speaking chat robot.

Chat bots are often integrated into conversation systems, such as

How to judge whether a robot is smart enough.

Most bystanders think that a robot can answer many questions and can be regarded as more intelligent. Regardless of the data structure or algorithm of the robot, as long as it has enough data, it can achieve a relatively accurate rate in answering questions. Obviously, the number of questions that can be answered just represents whether the database behind the robot is large enough.

And intelligence should not be compared to whose database is large enough, intelligence is the manifestation of multiple aspects:

1. Learning ability

This is the most fundamental and difficult to improve standard. A robot that can grow automatically but has a small amount of data (less questions that can be answered) is obviously more than a robot that cannot grow automatically but has a large amount of data (that can answer questions). More) practical. The essence of robots is to help humans and reduce the amount of human labor in various fields. If a robot needs to manually input all knowledge, it will increase the burden on humans, which is contrary to the original intention of manufacturing robots. This is also the current chat robot Although hotter, it is a root cause of less application.

Data screening capabilities

After possessing the self-learning ability, whether the robot fully accepts the knowledge or chooses to learn more correct knowledge is a criterion for further judging the intelligence of the robot.

If the robot can only collect all the knowledge, this learning ability is incomplete, and the robot should have more or less screening capabilities.

3.Self-upgrade ability

After the robot has self-learned according to the designer's data structure and algorithm, and has a certain degree of knowledge screening ability, the self-upgrading ability will become the next bastion.

The essence of learning ability is to structure the input data into the robot's data organization structure according to the designer's algorithm.

After "destructuring various types of input data according to the initial algorithm", after statistics, it is found that such rules cannot adapt to a certain type of knowledge organization or knowledge in a certain field (increased error rate), designers should consider how to make robots try to adjust Algorithm rules and data organization structure reduce the learning error rate in the knowledge organization form or field with higher error rate.

Many Chinese robots that appeared in the early years, because they use the most primitive "keyword matching" method, do not have the concept of data structure, so they should not be included in the category of "smart chat robots".

Here is a conversation between A and J:

1-A: Do you dream?

2-J: Do it.

3-A: Of course. So, can you remember your dreams?

4-J: I can recall them very clearly and vividly.

5-A: What is it like?

6-J: I don't want to flirt with a robot yet

Obviously, machines can have "memory".

If there is no long-term memory, according to the second sentence: do. How can I produce the third sentence?

If there is no long-term memory, A will not know what do in the second sentence means, but the third sentence will continue to discuss the topic of dream precisely, indicating that A has long-term memory of.

Fetion Assistant has not yet been able to go further in this regard. It can only respond quickly to a sentence, without coherent thinking and logical abilities. Far from real artificial intelligence, Fetion Assistant can only be described as a small trick. In essence, it does not have the ability to actively think about association and memory. As Roberts, a Bronze winner of the Lebner Award, puts it: "I don't believe in Turing's theory, and I don't believe Elbert can think." As the creator of Elbert, he likened, "If You know where a magic secret is and how it's done, it's no longer mysterious to you. "

However, it does represent a direction. In a certain sense, Fetion Assistant is equivalent to a web search engine, which is responsible for the automatic search, query, and processing of network information. However, it is more friendly and visual, and more in line with interpersonal communication habits. The appearance of Fetion's assistants also seems to have another meaning. Does the cleverness and ingenuity it shows mean that our human mind and language may not be so difficult to imitate? Does it mean that human language habits can be summarized and copied, and then this law can be applied to humans?

Imagine further, maybe one day, the scene in the movie "The Matrix" will really appear. Humans made robots, but robots rebelled, and war broke out with humans. Thinking computers control the human brain, and humans live under the deception of computers?

In 2014, the first domestic artificial intelligence company, Beijing Guangnian Infinite Technology Co., Ltd., launched the Turing robot, which is a more intelligent robot brain in the Chinese context, and is the world's leading Chinese semantic and cognitive computing platform.

Turing robots have an accuracy rate of 90% in understanding Chinese semantics, and can provide intelligent software and hardware products with artificial intelligence technology services such as Chinese semantic analysis, natural language dialogue, and deep Q & A.

Turing robot applications include smart customer service, virtual robots, smart watches, smart cars and smart homes. Among them, the virtual robot can be connected to WeChat, QQ and other platforms to build chat robots and communicate smoothly with users. Turing robots provide natural Chinese conversation ability, accurate Chinese semantic analysis ability, accurately judge user intent, and also have rich contextual scenarios and strong self-learning ability.