Image Source – Link here
When we look at a chair, irrespective of its shape and size, we know its for sitting.
Similarly, when we see a bird in air, we know that it can fly. In modern psychology, this ability of the intelligent human race to attribute a trait and action to a particular thing (living or non-living) by looking at it is known as ‘Theory of Affordance’.
To put in the words of the James J. Gibson, the father of this theory –
“when intelligent beings look at the world they perceive not simply objects and their relationships but also their possibilities. In other words, the chair “affords” the possibility of sitting” (source: Link here)
In recent research DeepMind is applying this concept into action.
Moving beyond just the regular 2D modelling that has been adopted in AI, it is trying to merge the concept of affordance so that the machine and robots are able to afford the purpose of things around them rather than learning from trial and error i.e.
- Identify the data points to be collected
- Then build hypothesis based on known 2D elements
- Anticipate behaviour to fine tune the model
- Then put training data set
- Then test with the real world data
- Repeat this till the model achieves the desired results
This process is expensive and is also prone to multiple failures. Consequently, the motivation to complete gets thwarted.
Affordances are quite popular (in conjunction with the signifiers) in the product development. It has a lot of application in the gaming world. Following video gives a good idea about this.
Further progress in this should surely help in building up a more intelligent ecosystem around the interface of real and artificial world.
With the concepts of digital twins gaining more and more prominence, this is an important experiment as this would make the virtual twin as real as possible.
In the field of healthcare, Philips is doing some pioneering work with digital twins. Given that the the twin is still a virtual replica of the human being, the affordance factor can help in fine tuning the ability of the twin to forecast things based on the human’s day to day habits.
Philips Healthcare’s Digital Twin Initiative
More research in this space is a certain evolution that the world needs to have AI do much more than what it is currently doing.
It is often said that with AI, we are still scratching the surface.
DeepMind’s venture into Theory of Affordances and a potential extended use case with the Digital Twins would define a significantly giant leap in making AI the central piece of corporate strategies in times to come.