Rumours are already in circulation that SAP is once again missing the latest IT trend – and this time things could get dangerous for the ERP software producer. The first time, the IT scene laughed about SAP when it became evident that Walldorf was not giving enough attention to the internet.
“Is SAP sleeping through the age of the internet?“, was the question rustling through the trees. At the height of the dotcom bubble, SAP had no answers as to how the world wide web could fit with R/3. Walldorf’s first solution: sap.com became mysap.com and the SAP staff’s business cards were printed with very colourful mysap.com logos.
By now, in-house developments and purchases have adequately positioned SAP in the web community. SAP’s mobile computing conforms to the HTML5 web standard and is called Fiori/UI5.
So is it ꞌallꞌs well that ends wellꞌ? The Walldorfers were just about able to repair and make up for their late entry into the world wide web. Yet faced with the current IT megatrend of artificial intelligence / machine learning (deep learning), the picture is very different.
Here the megatrend could become a mega-meltdown for SAP. It is not only about being on a sidetrack in conceptual terms, one that is seen in Walldorf as the path leading to the solution; many experts also take the view that AI and machine (deep) learning constitute perhaps the last megatrend in information science: from now on, AI is expected to continue developing constantly, always acquiring higher performance as part of natural development, but there will be nothing else new after this.
The ꞌoath of disclosureꞌ was sworn by Dr. Tanja Rueckert, SAP Executive Vice President, at an AI conference in Berlin, organised by the weekly newspaper ꞌDie Zeitꞌ and featuring high-calibre speakers. There Rueckert spoke about intelligent systems based on mathematical and statistical functions, as are found by every Hana application user in the SAP framework PAL (Predictive Analysis Library). Her example was forward-looking maintenance undertaken by railway companies.
As much as two years ago, Bernd Leukert (SAP Board Member for technology) presented this scenario, taking the example of the state-owned Italian railway company: hundreds of sensors in the locomotives and wagons provide information about the state of engines and assemblies. Based on MTBF rates (Mean Time Between Failures), statistical functions such as the ꞌbathtub curveꞌ and further complex mathematical algorithms, forward-looking (and consequently less expensive) maintenance can be put into effect very well and successfully.
Maintenance and logistics systems of this kind can justly be characterised as intelligent; according to Bernd Leukertꞌs information, by now the Italian railway company is saving itself millions on maintenance costs. But, Dr. Rueckert and Mr. Leukert, just because a system is intelligent and can look into the future a little, that does not make it AI.
Artificial Intelligence means something different to being intelligent! Here SAP is comparing the famous apples with the equally famous pears. Hereꞌs the essential difference, immanent to the system: SAP’s predictive-analysis system for Italian state railways was programmed by humans and is understood by humans. Ultimately the variable service intervals could also be calculated by human beings – however, in view of the quantity of locomotives and the associated millions of units of sensor-data it is recommendable to use a Hana system.
Artificial intelligence is about a computer beating the world’s best player of Go. Here the human being solely understands the computer’s hardware construction (a very deep neural network) and also the initial algorithms that are set – the machine itself organised and learned everything else itself.
Similarly, the inventors of the machine AlphaGo and Go experts were not able to explain many of the machine’s moves when the contest against the world’s best Go player took place; yet AlphaGo won in four out of five games (the British firm, DeepMind, was bought up by Google for USD 600 m. and is acknowledged to be the creator of AlphaGo).
Some experts at the Berlin AI conference confirm this development: it is foreseeable that AI will soon be superior to human beings. Juergen Schmidhuber, Scientific Director at the Swiss AI Lab IDSIA, takes a positive view: AI will outperform humankind, establish settlements in outer space and leave us behind on planet earth.
By the nature of things, SAP is far away from universal problem solutions, neural networks with feedback capability, and mathematically rigorous, universal AI. However, Hana PAL would have at least one theoretical approach in the form of neural functions and a graph databank. In this way, SAP could still achieve the link-up to the future, yet Bernd Leukert has never mentioned anything about this during his Sapphire keynotes; likewise Tanja Rueckert kept silent on genuine AI at the Berlin event.
That said, and as Dr. Rueckert also emphasised, there will be no way around genuine AI in the realms of ERP, IoT, logistics, robotics and Industry 4.0. – the AI conference made this fully clear. SAP will really have to invest a lot of effort – perhaps there was already an SAP headhunter sitting in the audience?
PS: I attended the AI conference in Berlin as an invited guest. The travel costs and hotel costs were paid by the publishing house E-3 Verlag B4Bmedia.net AG.