What is Augmented Reasoning

A new and emerging form of artificial intelligence (AI), augmented reasoning combines an advanced ability to learn patterns using traditional machine learning, with an ability to reason.

Rather than teaching humans to learn how to ‘talk’ in computer language, augmented reasoning can help us make computers better at understanding people and our needs, through more natural conversation and interaction. Augmented reasoning gives software an ability to solve some of the frustrations and problems that we all experience with current computers and technology.

Why augmented reasoning

CAR will research and develop new augmented systems and improve the application of machine learning technology across a range of applications, these might include:

  • machines that work with data analysts in companies to optimise business processes 
  • machines that can question people in ways that are more natural and easier than filling in forms
  • robots that can understand and follow instructions from people
  • websites that interact with people to solve their problems and answer their questions
  • factories where people and machines work seamlessly together without the need for constant reprogramming of software
  • machines and staff working more effectively to deliver what customers really want.

The result will reduce the need for structured interfaces between humans and machines (like keyboards and command lines) and enable machines and humans to interact in a more natural way (for humans).
 

young man working at a computer terminal with a robot arm attached to desk

Augmented reasoning will help us build machines that can learn faster, with less information, through greater interaction with their environment.

Examples of augmented reasoning

Examples of augmented reasoning include (but are not limited to):

  • visual question answering
  • visual language navigation
  • learning from priors
    (such as geometry + NLP, neural priors, implicit functions and deep priors)
  • machine learning and reasoning
  • causation and machine learning interface.

The intention is to build machines that can learn faster, with less information, through greater interaction with their environment and learn from prior knowledge.

Freed from keyboards and command lines, computers will be able to be more useful and less demanding to interact with. By being able to reason, computers will be able to link our language to its ability to process, operate and predict. We are looking particularly for examples whereby augmented reasoning can democratise technology, giving many more people the ability to program their own functions and operations, without needing to be experts in AI or coding.