Today, artificial intelligence relies on connections to external servers to perform complex processing: all (or almost all) of the research work aims to reduce the size and complexity of the process and processors. The target? Create artificial intelligence “operators” able to work alone, without connecting to the internet.
The latest study in this direction comes from the University of Florida and I'll put it here: shows how there is scope for voice assistants, robotics and other applications even in remote places on the planet (or on other planets). Artificial, autonomous and away brains!
Students without teachers
Today robots and neural networks have new demands. They need someone to tell them what job they have to do in an environment. In the near future there will be several robots that will have autonomous learning capabilities, without any supervision.
Think of activities like locating and rescuing people trapped in faraway places. Or constantly monitor a huge number of data (for example the health conditions of millions of elderly people) to independently request an intervention in the event of an emergency.
Memristori: portable brains at the forefront
Computer science inspired by the structure of the brain has grown tremendously in recent years: nevertheless, research is struggling to develop complex "neuromorphic" circuits (also called memristors) with standard materials.
The “portable mini brains” developed by the UCF are currently the most reliable: will they kickstart the breakthrough?
It would be interesting to see a memristore-based device that could perform the same function as an AI operator: in other words, as mentioned, a device capable of learning and working on its own without connecting to the Internet.
The researchers predict that these chips will be used in mass technologies within the next decade. Their paper already contains the next steps for the further development of this technology, which today is already incredible.