If we ever want to get a glimpse of the world outside the porch of the current state of AI applications, we will need to figure out a way to solve the issue of collecting enough data. Data has shape - and the shape matters, now more than ever.
The not so distant future of AI will impact society beyond measures, from truly customized health care to autonomous industries. This we know since it has been declared throughout every other business report the previous decades. Before discussing how to get there, we need to settle the As-Is. As with any other technology there are stages of AI development – and as of today we have barely left the porch.
Our current stage of AI is Artificial Narrow Intelligence (ANI), where AI can outperform humans in specific and repetitive functions such as driving based on strict traffic rules and sensitive sensors or recommendations of fashion products based on demography, clicks, views and previous consumption. This is all done under supervision of people. Also, here AI is mapping a course of action that in its fundament is based on set rules, environmental observations and existing distinct patterns of human actions and the human mind. Yes, the human mind follows very distinct patterns even though we would not possibly surrender to the thought of that “theory”. Today, it is not like AI relates to you as an individual, it is merely observing your most evident behavior and starts conceptualizing your needs based on that observation.
In the case of truly autonomous industries, we are talking about the so called Artificial General Intelligence (AGI) or even Artificial Superintelligence (ASI), where AI has gone to autonomously learn and understand any intellectual task and even improvise solutioning without being provided a thorough manual by a human being. AGI is rather an integration of several ANI applications – being able to learn and perform different tasks with a multi-purpose. Imagine a system being able to deliver customized health care advice to one patient and seamlessly performing surgery on another whenever it suits the purpose – or from a business perspective – SAP morphing into Adobe Illustrator and Microsoft Excel simultaneously. ASI is kind of unimaginable for the human mind in itself. However, imagine an emulated human mind that is run on much faster hardware than the human brain – thereby surpassing the intelligence of us as people. But as with any technology out there – this is enabled by data.
Data has shape - and the shape matters
Data has shape and the shape matters. If we combine data consisting of the color of leaves on a certain selection of trees and its relation to climatological seasons, we can make a logical assumption about one tree’s color based on what climatological season we are in. If we stir around the shape of those data sets and add assumptions of when a certain season arrives in time – we can make a logical assumption of the color of a certain selection of trees at a given time point. An artificially intelligent recommendation system is also restrained by the shape of the data that it can analyze. Creating a recommendation for one specific user on an e-commerce platform is based on that the shape of the data from the user – observations of the user’s individual behavior on that platform along with, for example, geographical location. It is merely a way of narrowing down your customer segments into more precise spaces. But to make it even more intelligent and making the space even more precise – it needs to be fueled by more data. What type of data? Some data could initially be more interesting than other, for example demography. But certain types of data can have intrinsic value beyond its original purpose. You can study satellite imagery of a retailer’s parking lot to assess its economic performance. Was this a part of the purpose of launching satellites into space?
With enough data of different nature to act on, the ANI applications of today can start the process of acting with a multi-purpose. If we ever want to get a glimpse of the world outside the porch of ANI, we do not necessarily need to become much more intelligent in terms of advanced algorithms – we can come a long way with just much more data.