Google’s Machine Learning Software Has Learned to Replicate Itself
The AutoML platform has begun to write code itself that is more effective than human developers; As part of the process, she began to replicate herself to speed up the action.
However, there is no fear of uncontrolled or problematic action on its part at this stage
Google's artificial intelligence recorded a significant achievement: AutoML, the company's AI platform, was able to independently write a series of software, also with the same technology, that is more efficient and better than Google's own.
AutoML developed by Google answers the lack of super-talents in artificial intelligence.
The company explains that it is not easy to find expert developers to meet the demands of the various projects.
The AutoML platform runs thousands of simultaneous simulations to determine which areas of the code require improvement, automates changes in algorithms,
and is able to continue the upgrade process until it achieves the required achievement.
As it improves the software, AutoML has done something to scare anyone who once saw one of the "Terminator" movies: it has developed clones of itself.
This replication process takes only a few hours using the smart algorithm, which would take months for a human key.
This is not yet full artificial intelligence, since code generation and smart algorithm optimization is not equivalent to creating an artificial neural network.
Such a process would require several breakthroughs in AI, as human developers continuously monitor the operation of the software and act as gatekeepers.
AI programs tend to find shortcuts and links some of which are wrong and even ethically prohibited, such as age discrimination, ethnic identity, or gender.
This is because they rely on past cases, ie, software and databases built by human developers whose attitudes and perceptions are embodied in the software.
