Estimated reading time: 3 minutes, 5 seconds

For decades, artificial intelligence has been seeking to mimic humans and nature. The main goal here is to create systems that have the ability to make decisions like humans. As such, most AI systems are brain-inspired, and scientists behind it have always attempted to mimic neural networks, a term borrowed from biology. With computers now able to think like humans, computer scientists are now focused on a different goal as they seek to revise artificial intelligence by transitioning it into smarter and more efficient systems. Initially, most of the research was devoted to training AI through machine learning to enable algorithms to learn how to recognize objects and how to perform basic tasks. Now that this has been achieved, these techniques are being deployed to create more advanced systems that can work autonomously without human intervention.

Google has been working on AutoML (automatic machine learning system), a smart machine learning system that can be mutated into a next-generation AI. This new system has been enhanced through the addition of the Darwinian evolution concept that allows building of AI products that can improve themselves. The product of this improvement is called AutoML-Zero. While this system can be worrying to many, if used well, it can open doors for the development of much smarter systems that can mimic human brains in development and decision making through enhanced evolution.

Going back to Charles Darwin’s concept of evolution, slight or random changes in the genetic makeup in an organism are either advantageous or disadvantageous. If the mutation allows the organism to survive and reproduce, the characteristics are then inherited by the future generation. Otherwise, the mutation dies with the organism. In AI, this phenomenon is known as neuro-evolution. It is a process in that neural networks recreates that built parts of the brain to allow only the strong or smart to survive.

An AutoML system makes it easier for applications to use machine learning and other automated features, unlike AutoML-Zero, that requires human input. It uses a three-step process that is setup, predict and learn to improve itself. The system begins with a selection of 100 algorithms by analysis of their mathematical operations. It then identifies the best of these algorithms through a trial-and-error approach. The best ones are retained and tweaked for the next round of trial-and-error while those found to be less effective are dropped. This is basically a mutation of neural networks. The process goes on and on until the neural network is highly efficient.

The new code is more effective and can efficiently identify things that have always been hard for AI to differentiate, such as the difference between objects. These best performing algorithms are kept in a similar manner as survival for the fittest fronted by Charles Darwin. According to researchers at Google, this Darwinian concept of neuroevolution can lead to the creation of an algorithm with 94% accuracy. Various studies saw an improvement in the algorithm’s image recognition skills.

The evolution of the machine learning algorithms, according to researchers, has led to the creation of more artificial intelligence systems that can work more like humans and without bias, that humans are known for. The mutation of algorithms will see the emergence of systems that can perform several tasks, including complex tasks that have for long relied on human expertise.

Despite the complex nature of the development of AI systems and advanced research that is needed to deliver highly efficient systems, Google’s AI mutation approach has shown to be a giant step towards making cutting-edge AI systems that can be helpful in the future. It is through such initiatives that we are likely to see systems that can outsmart humans in every aspect sooner rather than later.

Read 529 times
Rate this item
(0 votes)
Tagged under
Scott Koegler

Scott Koegler is Executive Editor for PMG360. He is a technology writer and editor with 20+ years experience delivering high value content to readers and publishers. 

Find his portfolio here and his personal bio here


Visit other PMG Sites:

click me
PMG360 is committed to protecting the privacy of the personal data we collect from our subscribers/agents/customers/exhibitors and sponsors. On May 25th, the European's GDPR policy will be enforced. Nothing is changing about your current settings or how your information is processed, however, we have made a few changes. We have updated our Privacy Policy and Cookie Policy to make it easier for you to understand what information we collect, how and why we collect it.
Ok Decline