How Deep-Learning AI help human programmers
The prospect of artificial intelligence is infinite. Home service robots, Amazon’s smart Echo, drone delivery, accurate medical diagnosis… With the help of AI, these smart devices are entering every household at a rapid rate. Any incredible application that people can imagine is probably related to AI.
However, you may ignore the subversion of artificial intelligence to another area – the development of AI software itself.
Imagine what would happen if computers don’t need humans to understand their language.
Most AI researchers today tend to develop artificial intelligence applications for specific uses, such as face recognition, smart driving, speech recognition, and even smart search. And if these algorithms can understand their own code without self-help, self-analysis and correction, this may be a breakthrough for artificial intelligence. Once the machine is self-programmable, there is an endless possibility – the robot will be more like humans, the future of medical care will grow faster, the phone will be smarter, the software bugs will be reduced, and even annoying bank frauds will disappear.
Don’t think that “computer self-programming” is far from us. In fact, there are already some technology companies doing this research, such as Google. Google has developed a bug prediction program that uses machine learning and data analysis to determine whether a line of code in a program has a vulnerability. Google engineer Ilya Grigorik has also launched an open source version of the bug prediction tool, which has been downloaded 20,000 times.
In addition, Siri’s successor Viv is also an example. According to a recent report in the US Wired magazine, Viv implements language recognition through a series of natural language processing, and builds complex self-adaptive computer programs based on the English vocabulary. This means that the code “self” can be used to write code (Code writing code). However, since this code is professionally trained by Viv and colleagues, the “self” programming in this sense is different from what we are talking about, but this is already a major advancement in artificial intelligence.
Similarly, in non-professional areas, Emil Schutte has a very classic provocative remark: “Are you tired of typing code? Clever, me too! Let Stack Overflow do it yourself.” Stack Overflow program can be programmed from a large scale The code is extracted from the database to provide itself with a complete functional code module. However, this extracted module is still based on code that has already been written.
As more and more such attempts are made, the ability of the machine to self-program will become more mature. Artificial intelligence has surpassed humans in many ways—visual processing, image recognition, video games… Perhaps in the future, even human programming is not required for computer programming, and they can do it themselves.
Computers can understand themselves and achieve self-programming. This goal may be a matter of time. And once they can do this, artificial intelligence will change dramatically in every area.