The augmented age of AI

This TedX presentation of intuitive AI is already two years old but it is one of the most powerful concept I have seen on Ted. One which will completely transform the way we live and work in short order.

The video is 15mn. Rather long for a TED speech but well worth the time.


So what will be the consequences of augmented AI?

When we look at AI software today, we are very much where operating software / systems were in the 1970s just before the advent of Windows. If you want to implement a AI solution today, you must build it ad-hoc specifically to answer one problem. It is expensive, time consuming and whereas the AI is supposed to input flexibility related to the problem, the solution itself is everything but flexible.

This is what will change soon. We will see the emergence of out of the box AI, very simple systems at first, which job will be to assist us in taking the right decisions.

The technology is already here for engineering problems because the parameters are usually well understood and easy to model (often because we have a long experience in doing so) as explained in the video with the example of the drone exoskeleton. Less so with soft skills decisions where intuition is more important.

This is where the second idea presented in the video is so revolutionary as it explains an easy way out that almost any company can apply in short order.

Intuition is and will remain one of the most complex nut to crack for AI systems. We will most certainly get there eventually but it will require new breakthroughs beyond Machine Learning and backward propagation. Nevertheless what AI cannot yet do on its own, can be solved with data or rather data feedback.

This is where the IoT with its connected capabilities, but not only the IoT, any system with an inbuilt connected loop feeding back usage information into the designing process, can improve an AI system with the appropriate information, replacing pure intuition with actual usage information.

With this simple idea, we can overcome the "intuition" or "creativity" bottleneck and reset the question as a choice between options generated by the data fed back from the system as explained with the barbie doll example.

This is a new paradigm which changes everything. To make sense and become useful, the data generated by any product, system or service must be appropriately measured, recorded and stored in order to make it accessible to humans, a AI system or both to improve the original input whatever it is.

This is an extension of the original Japanese concept of kaizen (continuous improvement) with the twist that it can be automated and applied to almost anything provided data can be generated.

There is still the concern of personal information (PII) to solve but here in most cases, we are looking at interaction data more that purely personal data as such. This data can therefore easily be anonymized and used for the purpose intended.