Artificial intelligence helps dumb ERP

Rumors are already starting to surface that SAP is once again missing out on the latest IT trend - and this time it could be dangerous for the ERP software manufacturer.

The first time the IT scene laughed at SAP was when it became obvious that too little attention was being paid to the Internet in Walldorf.

"Is SAP Sleeping Through the Internet?", it rustled through the press.

At the height of the dot-com bubble, SAP had no answers as to how the World Wide Web could fit into R/3.

First solution from Walldorf: sap.com became mysap.com and the business cards of SAP employees were printed with very colorful mysap.com logos.

In-house developments and acquisitions have now sufficiently positioned SAP in the web community. SAP's mobile computing obeys the HTML5 web standard and is called Fiori/UI5.

All's well that ends well?

The Walldorf-based company was just able to repair and compensate for its late entry into the World Wide Web.

The current IT megatrend of artificial intelligence/machine learning (deep learning) is a completely different story. Here, the megatrend could become a mega-GAU for SAP.

Not only is it conceptually on a sidetrack that Walldorf believes is the way forward, but many experts believe that AI and machine (deep) learning are perhaps the last megatrend in computer science:

From now on, the AI is to develop continuously and steadily, naturally becoming more and more powerful, but there will be nothing new after that.

Dr. Tanja Rückert, SAP Executive Vice President, performed the oath of revelation at an AI conference in Berlin, which was organized by the weekly newspaper Die Zeit with top-class speakers.

Ms. Rückert spoke there about intelligent systems based on mathematical and statistical functions that every Hana user finds in the SAP framework PAL (Predictive Analysis Library). Her example was predictive maintenance at railroad companies.

Two years ago, Bernd Leukert, SAP's Chief Technology Officer, presented this scenario using the example of the Italian state railroad company: Hundreds of sensors in locomotives and rail cars provide information about the condition of engines and aggregates.

Based on MTBF rates (Mean Time Between Failures), statistical functions such as the "bathtub" curve and other complex, mathematical algorithms, predictive and thus less expensive maintenance can be implemented very well and successfully.

Such maintenance and logistics systems can rightly be called intelligent and, according to Bernd Leukert, the Italian railroads are now saving millions in maintenance costs.

But, Ms. Rückert and Mr. Leukert, just because a system is intelligent and can see a little bit into the future does not make it AI.

Artificial intelligence is something other than being intelligent! Here, SAP is comparing the famous apples with the equally famous pears.

The essential and system-immanent difference: SAP's predictive analytics system for Italian Railways was programmed by humans and is understood by humans.

Ultimately, the variable service intervals could also be calculated by humans - but given the volume of locomotives and the associated millions of sensor data, it is advisable to use a Hana system.

Artificial intelligence is when a computer beats the world's best Go player. Here, the human only understands the hardware structure of the computer (a very deep neural network) and the set initial algorithms - everything else organized and learned by the machine itself.

Even the inventors of the machine AlphaGo and Go experts could not explain many moves of the machine on the occasion of the competition against the world's best Go player, nevertheless AlphaGo won in four out of five games (the British company DeepMind was bought by Google for 600 million dollars and is considered the creator of AlphaGo).

Some experts at the Berlin AI conference confirmed this development: It is foreseeable that AI will soon be superior to humans.

Prof. Dr. Jürgen Schmidhuber, Scientific Director, Swiss Research Institute for Artificial Intelligence (IDSIA), takes a positive view: AI will surpass humans, colonize space and leave us on Earth.

By its very nature, SAP is far from universal problem solving, feedback neural networks, and mathematically rigorous universal AI. But Hana PAL would at least have a theoretical approach in the form of neural functions and a graph database.

SAP could thus still catch up with the future, but Bernd Leukert never mentioned anything about this during his Sapphire keynotes, and Tanja Rückert was also silent about real AI in Berlin.

But - as Ms. Rückert also emphasized - real AI will be inevitable in the areas of ERP, IoT, logistics, robotics and Industry 4.0, the AI conference clearly revealed.

SAP will have to make an effort - maybe there was already an SAP headhunter sitting in the audience?

Peter M. Färbinger, Editor-in-Chief E-3 Magazine

PS: I was at the AI conference in Berlin by invitation. The travel and hotel expenses were paid by the E-3 publisher B4Bmedia.net AG