The poster is too optimistic about the current development of artificial intelligence. At present, it is still unclear whether no matter how sophisticated artificial intelligence is, it has basic consciousness, let alone personality and emotions. The most widespread application of artificial intelligence is still classification and forecasting. Support vector machines are used to identify computer viruses (classification), and neural networks are used to predict the direction of the stock market (prediction).
Some artificial intelligence mechanisms have a theoretical framework (such as Bayesian inference) set by humans. This algorithm works better when the model highly simulates reality, but has poor ability to adapt to unknown environments. Another type of algorithm regards calculating the output of artificial intelligence as an optimization problem. On the basis of bionics, it reduces the difference between the algorithm output and the ideal output by continuously adjusting various parameters (usually automatically completed by the computer). This process is learning. A highly automated learning process can be said to be "self-growth." Mainstream algorithms include various backpropagation algorithms, genetic evolution algorithms, etc. The ultimate goal is to find a balance between the algorithm's mastery of existing data and its ability to apply to unknown data. For example, if you want to train an algorithm to determine age by looking at people's faces, tell the algorithm that a person with wrinkles and white hair is about 70 years old, and a person with no wrinkles and black hair is about 30 years old. When the algorithm encounters a person with wrinkles but black hair, it can output 50, which roughly proves learning. And if the output is 10 or 100, it means the learning failed.
In terms of implementation, those who are more interested in software engineering may use Java/C++/Python, and those who are more theoretically inclined can use R/Matlab, etc. There are many ready-made libraries, so it’s no problem to develop them yourself.
In fact, the fields of artificial intelligence are far more than these. How to model natural language, how to deal with noise in data, and how to discover the most obvious patterns in the least amount of data like humans do, are all hot topics in research today.
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
The poster is too optimistic about the current development of artificial intelligence. At present, it is still unclear whether no matter how sophisticated artificial intelligence is, it has basic consciousness, let alone personality and emotions. The most widespread application of artificial intelligence is still classification and forecasting. Support vector machines are used to identify computer viruses (classification), and neural networks are used to predict the direction of the stock market (prediction).
Some artificial intelligence mechanisms have a theoretical framework (such as Bayesian inference) set by humans. This algorithm works better when the model highly simulates reality, but has poor ability to adapt to unknown environments. Another type of algorithm regards calculating the output of artificial intelligence as an optimization problem. On the basis of bionics, it reduces the difference between the algorithm output and the ideal output by continuously adjusting various parameters (usually automatically completed by the computer). This process is learning. A highly automated learning process can be said to be "self-growth." Mainstream algorithms include various backpropagation algorithms, genetic evolution algorithms, etc. The ultimate goal is to find a balance between the algorithm's mastery of existing data and its ability to apply to unknown data. For example, if you want to train an algorithm to determine age by looking at people's faces, tell the algorithm that a person with wrinkles and white hair is about 70 years old, and a person with no wrinkles and black hair is about 30 years old. When the algorithm encounters a person with wrinkles but black hair, it can output 50, which roughly proves learning. And if the output is 10 or 100, it means the learning failed.
In terms of implementation, those who are more interested in software engineering may use Java/C++/Python, and those who are more theoretically inclined can use R/Matlab, etc. There are many ready-made libraries, so it’s no problem to develop them yourself.
In fact, the fields of artificial intelligence are far more than these. How to model natural language, how to deal with noise in data, and how to discover the most obvious patterns in the least amount of data like humans do, are all hot topics in research today.
I am learning artificial intelligence, everyone should encourage me, qq:282886636