Artificial Intelligence: from cloud to dust?

Artificial Intelligence is experiencing a revolutionary paradigm shift, which will make it much more pervasive in all our activities and in our everyday lives. Several technology enhancements, unthinkable until a few years ago, are making the impossible possible. The topic is complex and extremely broad. Nevertheless, in this new “AI populated world” Artificial Intelligence is expanding in three different realms: the “Cloud”, “Things” and “Dust”, each with different laws, applications, and horizons.

Artificial Intelligence in the “Cloud”

Until recently, Internet of Things devices have been mostly “dumb”, made of sensors that collected data and were remotely operated from intelligent algorithms. It seemed to be a general law that all (artificial) intelligence had to be in remote data-centers (i.e. in the so-called “cloud”), as AI processing required a lot of storage and computational power, not available on small local devices.

Data centers and the “cloud” are currently experiencing an extraordinary growth in number, capacity, and AI effectiveness, pushing AI toward new boundaries. As an example, researchers have already created AI technologies in the cloud that can do many “human-type” tasks nearly as well as a person, including identifying images and recognizing words in a conversation.

On August 20th 2017, Artificial Intelligence reached a new level where it recognized words in a conversation as successfully as professional human transcribers. It is worth noticing that reaching human parity has been an AI research goal for the past 25 years and, whereas a couple of years ago this still seemed a very distant goal, today AI experts are already working on systems that can do even more complex tasks, like autonomously reading text and answering questions about it.

AI experts are developing what we might call literate machines, that can understand text and learn how to communicate it, both in written or oral form. Once this is achieved, AI will be able to read all information available in the world, especially on the Internet, turning information into knowledge and opening completely new scenarios for its application. Furthermore, when AI is used in conjunction with new mixed reality scenarios, completely new worlds arise, as seen in the video below.

Artificial Intelligence in “Things”

While AI continues its steep growth in the cloud, the “dumb” Internet of things paradigm is also profoundly changing, thanks to a new breed of artificial intelligence chips designed to run deep-learning software that leverages new developments in compressed AI algorithms.

This new frontier allows AI to address many new scenarios that will transform the world and which could not be solved with previous IoT devices and with AI in the cloud. Pushing AI to edge devices, in fact, reduces bandwidth constraints and eliminates the limitations related to network latency, i.e. the time it takes for data to travel to the cloud for processing and back to the device. On-device machine learning also limits battery drain from communication with the cloud and protects local autonomy.

The new AI “Things” will still be part of the so-called Internet of Things, or IoT, except that these things will be smart, or intelligent, even without the “Internet”. If you’re driving on a highway and there isn’t connectivity there, you don’t want AI to stop working. In fact, that’s where you really need it the most.

Interestingly, these new Artificial Intelligent “Things” are often inspired by nature and use self-organizing principles. As an example, AI researchers are testing in the Nevada desert a system that uses artificial intelligence to keep a type of glider known as a sailplane in the air without using a motor, by autonomously finding and catching rides on naturally occurring thermals, similarly to how many birds stay aloft.

Artificial Intelligence in the “Dust”

We are moving into a new world that will be composed by an intelligent “cloud” (at the center of the Internet) and by an intelligent network of smart “things” (at the edge of the Internet). This is a quantum leap toward realizing the promise of a future populated by tiny intelligent devices at every turn, embedded in our clothes, scattered around our homes and offices, and deployed to perform any sort of task, including predictive maintenance, everywhere, even taking care of our health.

Already today, many machine-learning models are deep neural networks, inspired by the biology of the human brain. In the future, a variety of techniques will enable AI to run on processors about the size of a red-pepper flake. New AI chips and new AI software are being embedded onto bread-crumb size processors that can run machine-learning algorithms on microcontrollers as small as the one pictured below.

This Artificial Intelligence will be distributed in models 1.000 to 10.000 times smaller than today and it is sometimes called “computer dust”. All sorts of intelligent particles could be created with this new paradigm, from smart soil-moisture sensors deployed for precision irrigation to brain implants that warn users of impending seizures so that they can get to a safe place and call a caregiver. At least in a first phase, “computer dust” tiny devices will be extremely focused on performing optimally specific scenarios.

In conclusion, Artificial Intelligence will be everywhere in our future world, from clouds to dust, from things to beings.

Are you ready for Artificial Intelligence?