If you’ve watched Hollywood movies about the world being taken over by artificial intelligence, then you might be scared of the news that Google has currently released to the world. The creation of an AI system is usually made by machine learning experts who spent countless years in mastering their art. But how will these creators feel when Google asks their AI system creations to develop the next generation AI that will one day be better than its predecessors in their line of work?
Source: Giphy
Google has taken the “train your own replacement” concept and integrated it into the modern age. Machine learning experts are a dime a dozen and usually costs a lot of money to employ, but soon, the need for them will definitely come to an end. The new age where software will inevitably be able to learn on its own marks the nearing end of these experts’ careers.
The research teams behind the Google Brain, DeepMind, and OpenAI, along with various university research departments are creating machine-learning systems capable of developing themselves. This means that these developed AI can technically “reproduce” on their own. The team behind Google Brain was able to design software that made an AI model to gauge how well software can process language. The results were astounding as the software created by the AI was marginally better than that of the manmade counterpart.
Jeff Dean, Google Brain team lead. Source: TechHive
Jeff Dean, leader of the Google Brain team, views automated machine learning as a great path to take for him and his team. As per Dean, problems are only solved with expertise, data, and computation. But is the elimination of the vast amount of machine-learning expertise really possible? An set of experiments made by the Google DeepMind team shows that this is a possible solution for lessening the amount of data tp finish a specific task.
Though the point of software being able to create AI isn’t particularly new, they weren’t able to produce practical results in order to be classified as efficient. Yoshua Bengio, who was one of the first to explore the subject in the 1990s, called it an “exciting” endeavor.
Source:
