At A Glance – Broad AI

Multitalented, flexible artificial intelligence

Currently, our AI systems tend to be built for a specific purpose. This is Narrow AI – artificial intelligence that is designed to complete one highly defined task. Common applications for Narrow AI include tasks such as image recognition, automatic video recommendations on sites such as YouTube and Netflix, and the natural language processing ability of the Amazon Alexa digital personal assistant.

Whilst Narrow AI systems can excel at executing highly complex tasks, their abilities are limited. A Narrow AI programme that is extremely advanced in its field will be unable to solve tasks outside of that domain – even if they are technically much simpler. For example, a facial recognition system that can identify one specific person from millions of others will probably be unable to tell the difference between a dog and a cat.

The opposite of this highly defined, Narrow AI is Broad AI. Broad AI systems can execute tasks across a variety of different fields. This is the kind of technology that would be necessary to create the robots that are at present confined to the realm of science fiction – digital assistants who could walk your dog, answer your email, and then make you breakfast. The techniques involved in Broad AI aim at replicating the traits of human intelligence, but it is important to note that they don’t yet exist. We don’t know if it will ever be possible to create truly broad AI systems, how they would behave, and whether or not they would have human like consciousness.

At the moment, efforts are focused on developing various different Narrow AI systems and combining them together to solve more complicated tasks. This is the technology at work in applications such as the self driving car – where different individual AIs execute tasks like image analysis, trajectory and speed calculation, speed limit identification and navigation, resulting overall in an autonomous vehicle.

To understand more technological terms, sign up to our free weekly newsletter.