AI < / > self-service or expert system

Everyone will use AI, as a business and even as a private person.

This statement is being published these days in numerous magazines, blogs and presented at conferences. Along with that and with the advent of Watson analytics, self-service AI seems to be the future. In fact, there is no doubt that AI will have a dramatic impact on our lives and on how we interact with IT. Processes will continue to be further automated while humans will be able to concentrate more on a high-quality job rather than doing repetitive tasks. Yet there is still an opinion that not everyone will use AI, however, nearly everyone will be able to build AI. As Business intelligence (BI) shifted some years ago from the IT department to the end-users’ fingertips, so will AI, – some experts and analysts claim. However, this second statement, as certain experts put it, is quite risky and controversial.

Experts from the industry and university researchers insist that some of the promises that tech companies make are hardly to be delivered. Algorithms are still too complicated for everyone to build or use them out of the box. Even after decades general software development is still a field that non-experts barely enter, so is AI.

Even though BI area is often claimed to be a self-service field, on the close scrutiny it turns out to be quite the opposite. Analysts and decision makers can build reports and dashboards, click around and drill down, however, the backend construction of the required data model is still done by the experts from that field. With AI and machine learning algorithms it is even more risky to let the average person put hands on it.

This becomes especially true when the algorithm should be used not only for playing around and a nice showcase, like it used to be with the good old linear regression model that nearly everyone performed at least once during their university time. To make a critical business decision that will impact customers or operations in general, models require a substantial amount of analysis and checks before being released and used in production. Of course, automatic checks and frameworks can help here but the devil lies in the detail.

Results interpretation may also depend on the business problem and various problems may require different min. levels of accuracy. This is also revolving around the required knowledge areas and skills a data scientist should have.

But while we are at it saying that self-service is not the way to go with AI, how will the future look like? Will it remain the same and we won’t see any development? Certainly not. As it has always been in technology and especially with IT, the field is moving rapidly. Claim that the most wanted job of the 21st century at some point may cease to be so highly demanded, unlike now, might line up with reality. One of the reasons for such a decrease may be that the industry structure will mature, like it already happened in other fields of IT. As with ERP systems: previously mature companies were making their own systems and hiring their own staff for that.

Nowadays, everyone buys the industry standard system with some customization from the core vendors. In their turn, vendors buy the underlying technology like databases from db-providers etc. It is very likely that we will also see similar development in the AI field.

Conclusions

Whether we will see a general-purpose AI in future beyond the systems we have right now – no one can tell for sure. But on the other hand, this is an exciting thing about the future that we do not know everything.

You might also like…