The Human Aspect of AI

Overview

Usually, when we think of AI we picture some machine, usually a robot, doing its own thing. Those more knowledgeable in the field might see an algorithm or some application when they hear about AI. However, few people consider a crucial component of every AI project: the human aspect of it.

As much as we like to think of AI systems as being autonomous and being able to handle the tasks they undertake with superhuman efficiency, this is just the tip of the iceberg. Behind every successful AI, there is always a team of hard-working individuals who make it possible. That’s why before investing in an AI system or RPA infrastructure, it’s important to consider the human aspect of AI which is generally three-fold.

The human expert

It's no state secret that an AI requires specialized know-how to come about. Of course, there are AI experts out there who claim they can "spin an AI system" on the fly or that they can develop it in an hour or so, but truth is that these claims don't represent reality. Most AI systems, especially those developed for an organization that's new to data science and AI tend to require more time. Sometimes a team of experts is needed to make it possible for an AI system to be developed and tested. With all this human effort and know-how, an AI system would be mediocre at best or non-existent at worst.

The human users

An AI system is usually as good as the context it is aware of. Everyone can gather a bunch of data, from various data sources, and put them into a model. If that model is AI-based, he can call this an AI system. However, without the context necessary this AI system may not yield sufficient value or integrate well with the other pipelines of the organization. Usually, it is human users who provide this context. There are times when the system learns through the guidance of humans, be it in a reinforcement learning setting or an active learning approach. In any case, many problems are too complex for an AI to tackle on its own, and without the context human users provide, it's next to impossible to do a decent job at them.

The human maintaining the system

Since no AI system is infallible (even the ones in sci-fi films have their Achilles’ heel!), it makes sense to keep someone in the loop in an effort to maintain the system. After all, as the data behind the AI gets stale or less relevant to the problem it is trying to solve, the system ceases being useful. To make sure that it remains useful, someone has to figure out when it’s becoming less accurate (through periodic testing). Once this is done, this person needs to get more data for it, particularly recent data or data from additional data streams that may complement the existing datasets. Then, the AI system needs to be retrained and re-evaluated. All this work, although it can be automated to some extent, is best done with a human professional in the loop.

Learning more about AI

AI is no doubt a vast topic and here we can only scratch the surface of it. This article is basically just a single slide of a webinar I'm preparing these days. Don't worry, it's not too technical since I find that the less jargon I use in a talk like that, the better the engagement of the audience. After all, we can always get more technical afterward, during the Q&A part. Should you wish to check out this webinar, I invite you to visit the organizer's website as well as the webinar's web page. Cheers!