Perception centric AI: ‘Project Rear-view Mirror”!

Joining AI and Business through the Missing Link of Perception

We all may remember the warning “objects in mirror are closer than they appear” in the rear-view mirrors of motor vehicles. This captures the heart of a wide and complicated phenomenon called “perception”. The fact that makes perception so interesting is that “no two humans are equal” in terms of how they perceive anything. In the project rear-view mirror, the aim is to investigate and model how AI and Perception interact. This not only bridges the domain of AI research with the domain of customer-centric application; it also links pure computer science-based research to humanities-based research.

Figure 1 shows a block-diagram level representation of some of the ways AI and Perception get connected with each other.

Figure 1. Linking AI with Perception

Let us explore some of the links.

1. The first link is the way AI directly affects industries. AI is being used and shall be used by various industries to various degrees. The major aim would, of course, be to maximize profit. “Profit” is an easy-to-measure quantitative variable. However, more and more industries are also showing interest in increasing their “impact”. This is an abstract concept and is, often, very difficult to define leave alone measure. However, most business gurus agree that impact affects the profitability of a business and vice-versa. Hence, when we consider the effect of AI on businesses, we have to give profit and impact equal importance. Measuring impact mostly involves understanding the perception by a certain group of people of certain events, products or investments.
2. The second link shows how AI research and innovation (R&I) give rise to newer algorithms. For their wide and “proper” adaptation in the industry, we need to make sure that the algorithms and their modus-operandi should be understandable by people from the business domain. This is a particularly complicated challenge given the facts that 1) most AI algorithms heavily depend on statistics, and 2) the intuitive understanding of statistics by humans is naturally very poor. This brings perception back into our equation.
3. The third link between AI and perception is an obvious one. This is about presenting the output from the algorithms in a format which creates expected reaction from the user through the use of proper UI/UX interfaces. This is an emerging field of research especially the work on modelling perception and using that in the training of the AI algorithms. However, proper exploitation of this link needs extensive co-development between AI and business experts. This also thrusts on the fact that most AI systems will need substantial customization when being ported into products. This is one of the reasons why sometimes porting AI algorithms do not give the level of product appreciation as professed by AI experts.
4. The last link is regarding how studies on ‘perception’ can affect AI algorithms making them better and more useful. This is when perceptual feedback is properly modelled, measured and used in the training of AI algorithms.

Project “rear-view mirror” is mostly about studying “perception” and investigating the four links discussed above. The author believes that for AI to become more fail-proof and to give expected benefit to business these pieces of research shall prove invaluable.

It is needless to mention that this is highly multi-disciplinary research where we would need as many interactions with psychologists, sociologists and the blue-collared worker from the factory floor as with the business managers and AI scientists.