What is Artifical Intelligence (AI)?
Artifical Intelligence is not new, but currently extremely popular. Numerous useful AI functions are currently making their way into companies - including ERP systems such as SAP.
Frequently asked questionsSee all questions
Artifical Intelligence - Basic knowledge, application areas and trends
Experts agree that enormous potential can be tapped in this way. Errors can be identified before they occur. Customer relationships, logistics chains and goods flows can be optimized. Soon it will also be possible to automate complete processes, with machines making or at least preparing intelligent decisions.
But what exactly does the term Artifical Intelligence mean? Since when have AI applications been available and why have they recently become an absolute trend topic? What can AI already do today and where are the limits? How does SAP deal with artificial intelligence? What does "Machine Learning" and "Deep Learning" mean? And how will Artifical Intelligence change companies and society in the future? On this page we will deal comprehensively with these and other questions concerning AI.
Artifical Intelligence: Definition
Artifical Intelligence (AI) is a discipline of computer science. It deals with methods that enable machines (computers) to solve tasks in a way that a human being would do with his intelligence. Artificial intelligence therefore encompasses not only aspects of information technology, but has also been shaped by psychology, neuroscience, linguistics, communication sciences, mathematics and philosophy. Computer science can rather be regarded as a means to an end. It brings the research areas together and enables their implementation.
Basically, Artifical Intelligence has two goals:
- Imitation of human thought and behaviour (strong AI)
- Autonomous and automatic completion of tasks (weak AI)
The first point is much more demanding and can still be described as a vision from today's point of view. The second aspect, on the other hand, is already reality. Here, AI takes on clearly defined tasks. Solutions such as intelligent software assistants or speech recognition systems are already working reliably and are increasingly finding their way into our everyday lives. Although such systems have already achieved enormous performance, they are also referred to as "weak AI" in technical jargon. One reason for this is that weak AI solutions operate at a relatively superficial intelligence level and do not develop a deep understanding of problem solving.
The strong AI (also called Strong AI or Super Intelligence) has the goal to gain intellectual abilities of people or even to exceed them. A strong AI acts actively, flexibly and intelligently. So far it has not been possible to develop a "super intelligence". There is controversy in the research community as to whether the development of such an AI is feasible at all.
Difference between Artificial Intelligence, Machine Learning and Deep Learning
While "Artificial Intelligence" is a generic term that covers all technologies used to provide intelligence services, Machine Learning is a branch of AI. Deep learning, on the other hand, is a sub-area of machine learning. Let's take a closer look at the approaches below to work out the differences.
Machine Learning describes mathematical methods that enable a machine to generate knowledge independently from experience. Or put casually: Machine learning lets computers do useful things without programming them beforehand. From a technical point of view, machine learning uses programs that can independently recognize certain laws and patterns in data on the basis of self-learning algorithms. To do this, the software must first be supplied with extensive data. During the development phase, a programmer ensures that the machine learning model is continuously adapted and optimized. In this way, the algorithm becomes more "intelligent" from data set to data set until it can finally perform its task independently.
The main goals of machine learning are to recognize correlations, to link data intelligently, to draw conclusions and to realize precise predictions. In the business environment, machine learning applications have the potential to relieve employees of tedious, unproductive tasks. This creates resources for new areas and makes work more efficient and economical. For example, learning software can independently scan paper documents, recognize the text, initiate further steps and organize archiving. A more complex scenario in which machine learning is already used today is predictive maintenance. Here, the algorithms are able to detect possible damage to technical equipment and fault patterns in order to request maintenance if necessary.
Speech recognition on mobile phones, spam filters in e-mail inboxes and facial recognition in photo management are largely controlled by machine learning algorithms. Often we have contact with machine learning without knowing it. This is the case, for example, when personalised advertising is displayed to us.
Deep Learning is a part of Machine Learning. It is a special method that uses so-called artificial neural networks and large amounts of data to learn particularly efficiently. The way it functions is based on learning processes in the human brain. Based on available information, corresponding systems can link what has been learned with new content again and again and thus continuously learn new things. At a certain point, the machine is then able to deliver forecasts, make independent decisions and question them. If the outcome of the decision is not satisfactory, it shall be adjusted in a new attempt. Man normally does not intervene in this learning process. This is also the main difference to Machine Learning.
Deep learning is particularly suitable for scenarios in which large amounts of data are to be examined for models and patterns. Application examples here are speech, object or face recognition. In speech recognition, for example, Deep Learning enables systems to expand their vocabulary independently with new words and word variants. Further areas of application are autonomous vehicles or robots, AI in computer games or a prediction of customer behaviour within the framework of CRM solutions.
Development and deployment of Artifical Intelligence
Since its inception, Artifical Intelligence has made enormous leaps in development. This applies in particular to the recent past. In the management of corporate groups, AI is currently a future topic in which the potential to fundamentally change companies and society is seen. There are opportunities to optimize one's own business and even to reorganize entire industries. But how did Artifical Intelligence come into being and what characteristics have corresponding systems actually reached to date?
Meinolf Schäfer, Senior Director Sales & Marketing
Do you have any questions? I will be glad to help you.
+41 41 418 45-24
The beginning of Artifical Intelligence
The origins of Artifical Intelligence go back to the 1950s. Already in this decade the term was used at a science conference in the USA. The scientist Marvin Minsky is regarded as one of the founding fathers of the AI. In 1966, he formulated the following definition: "Artificial intelligence is the science of making machines do things that would require intelligence if done by men." Freely translated this means: Artificial intelligence is given when machines do things for which human intelligence is necessary.
An early milestone of the AI was also the Turing Test. It was developed in the 1950s by the British mathematician Alan Turing and was intended to enable a person to communicate synchronously with other people and a machine via a chat software. In the 1960s the so-called "General Problem Solver" was presented. It was an AI system that could solve simple problems. In the same decade, the ELIZA software caused a sensation. At the time, the chat system made it possible to simulate therapy conversations. In 1997, a computer beat the then world chess champion.
In the following decades, the capabilities of Artifical Intelligence improved continuously. This was mainly due to steadily improving storage options and computing power. In 2011 the Watson software was introduced by IBM. She could win the quiz show Jeopardy against human opponents. AlphaGo is another AI program that has accomplished the previously considered impossible feat of beating a professional Go player.
Today's use of Artifical Intelligence
Artificial intelligence is now present in both our private and professional lives. A classic example is speech recognition, which has gained recognition through applications such as Siri and Alexa. Work is also underway on translation systems that translate conversations live into another language. In addition, for some years now we have been encountering so-called chatbots, which are used by companies to increase the efficiency of customer communication. It is software that can engage in an intelligent dialogue with users through extensive databases, linguistic knowledge and a full-text search engine. In English-speaking countries, chatbots already act as contact persons for medical questions.
Another important field of application for Artifical Intelligence is "Predective Analytics". Here, algorithms based on historical data from various sources create predictions for the future development of customer relationships, results and business processes. The financial sector uses appropriate applications for damage management and fraud detection. The planning of marketing campaigns can also be optimized with Predective Analytics. For example, algorithms can predict which customer will buy a certain product, at what time and at what location.
The first AI applications are also being used in Service Management. For example, Artifical Intelligence categorizes incidents and conducts support dialogues. In addition, the support staff can be supported in the analysis and solution of problems, provided that a suitable knowledge database is available.
AI has also led to significant improvements in IT security. AI-based solutions now make it possible to recognize attack patterns and prioritize security incidents. Security systems continually evolve with machine learning methods by collecting, analyzing and classifying threat data.
What does Artifical Intelligence mean for SAP?
The current orientation of SAP is evident: The Walldorf-based company wants to position itself as a software provider that makes intelligent companies possible with its products. With AI, SAP wants to automate processes and improve the user experience. The goal is, among other things, an ERP solution that can be operated entirely without a keyboard and instead by means of voice recognition. The integrated assistant called "CoPilot" has already been introduced for this purpose. It enables users to interact with the system via voice and, for example, to search for information. In purchasing and sourcing, for example, language-based elements are intended to speed up and simplify purchase requisitions. Sales employees, on the other hand, are enabled to convert quotations into orders via voice commands. A vision is also an intelligent assistant that is able, for example, to understand the agenda of meetings and automatically deliver the appropriate key figures. The corresponding software should learn from the habits of the user and adapt intelligently on this basis.
Another important area of application for artificial intelligence in the SAP environment will be the automation of processes. By 2021, the software provider already wants to automate 50 percent of all tasks that were previously performed manually. This results in enormous savings for companies.
If we look at the current status, SAP already has several AI and machine learning applications in its portfolio. Some of the things to mention are:
- SAP Leonardo: collection of applications and microservices for IoT, machine learning, blockchain, big data and analytics
- SAP Cash Application: Automating Receivables Clearing
- SAP Service Ticket Intelligence: Automate ticket capture, classification, routing, and resolution
- SAP Brand Impact: Analysis of advertising campaigns with automatic logo recognition
- SAP Customer Retention: Predicting customer behavior, predictive analysis
- SAP Predictive Service: Predictive analytics, predictive support for service personnel
- SAP Predictive Analytics: Predicting the Probability of Future Business Results
Artifical Intelligence and Machine Learning are already integrated in the SAP product SAP S/4HANA.
Ethical principles already defined
Artifical Intelligence is associated with enormous social changes and challenges in data protection. SAP has already dealt extensively with these "dark sides" of the AI. For example, the Software Group has developed guidelines to control the introduction and development of AI components. The overriding goal is "to improve the functioning of the global economy and people's lives". The principles cover the following aspects:
- Value-oriented action (respect for human rights and UN guiding principles)
- Focus on people and user experience
- Unbiased action for companies
- Transparency and Integrity
- Quality and safety
- Data protection and privacy
SAP also wants to address the societal challenges posed by AI. Aspects such as economic redistribution, economic development, social security and normative issues play a role here.
Ethical and moral issues in the field of artificial intelligence
Several ethical dilemmas arise in connection with artificial intelligence. Thus, it must be critically questioned whether decisions by autonomous machines may pose a threat to free will and the assumption of responsibility. In addition, developers can add tendencies to AI software that lead to exclusion or discrimination. This circumstance is particularly problematic if these tendencies arise unintentionally within the framework of machine learning. There is also the risk that people will distinguish themselves through algorithms. This could jeopardise cultural and political pluralism.
In addition, AI systems require huge amounts of data in order to be able to sensibly run through learning processes. This also includes personal data. The applicable data protection laws stand in stark contrast to this. The large amount of information, however, poses further challenges. For example, it is sometimes difficult to filter out correct and error-free information. If the database is not undoubtedly of high quality, software results and decisions cannot be trusted.
Outlook: What is the trend for Artifical Intelligence?
Artifical Intelligence is undoubtedly a highly emotional topic. Extreme positions often dominate the discussion. One camp sees AI as a threat to all humanity, while the other side sees technology as a panacea for all our problems. Whether one of these scenarios will occur, nobody can judge today yet. The fact is, however, that AI will become significantly more important in the coming years and decades. Experts agree that Artifical Intelligence is a key technology of the digital revolution.
For companies wishing to make progress in the field of digitisation, there is therefore no alternative to linking artificial and human intelligence. It is very likely that AI will relieve employees of routine tasks in the future. In particular, standard operations that are repeated at high frequency are the potential areas of application. In 2013 and 2016, the Institute for Employment Research collected data on the development of digitisation and changes in job profiles in Germany. During this period, the potential in the area of machine learning has increased enormously, particularly in the area of company-related service occupations. In the manufacturing technology segment, the proportion of activities that could potentially be performed by robots was already 65% in 2013.
However, machines will not only automate processes in the coming years. An enormous potential is also hidden in the analysis of large amounts of data. In contrast to classical approaches, which are based purely on the evaluation of past values, AI enables a look into the future. On this basis, precise decisions can be made and new business models and smart services such as predictive machine maintenance can be implemented. The optimisation of goods flows and logistics chains is also a field of application in which AI is likely to prevail.
And where is man? Will it soon be completely replaced in the working world by robots with Artifical Intelligence? Certain job profiles could actually disappear through AI, but at the same time new jobs could be created. Human abilities such as creativity and empathy could come to the fore again. There is freedom to concentrate on one's own strengths again and to develop innovations. This also has a positive effect on employee satisfaction. The intuition factor will also continue to be in demand. Decisions will be based to a much greater extent on data from intelligent analyses. However, if the scope of the decision is high, man will still have the last word until further notice.
Conclusion: Opportunity and risk at the same time
There is a chance that AI will develop positively and empower people to better solve the problems of a modern society and achieve more. The risk, on the other hand, lies in allowing AI to act beyond the limits of meaningful control. For companies, such an approach would not only be unacceptable in terms of reputation and ethics. In the event of failures, the management would possibly delay innovations or even stop them completely. The level of security and control that Artifical Intelligence provides could therefore determine whether intelligent machines become a curse or a blessing for us.
Meinolf Schäfer, Senior Director Sales & Marketing
Do you have any questions? I will be glad to help you.
+41 41 418 45-24