The use of Artificial Intelligence in companies: Where are we and what will the future bring?

Artificial intelligence (AI) - one of the buzzwords in IT in recent years. At the same time, AI applications are still rare and many companies lack the imagination of the functions in which artificial intelligence is used in practice. Pioneers, on the other hand, have already made AI investments and are using the technology to their advantage. More importantly, in which direction is AI developing? Do AI investments make sense at this point in time or have companies that do not yet use AI applications already missed the boat? Does AI possibly even pose risks for IT security or, on the contrary, is AI actually an enrichment for security?

What exactly does Artificial Intelligence mean?

Artificial intelligence is a computer system that makes certain decisions on its own. Another feature of this technology is that the systems are capable of learning. This is called machine learning. By performing recurring tasks, the software learns and works more precisely or efficiently over time.

Within AI, there are also different directions of development. Above all, there is the distinction between strong and weak AI. Strong AI is a machine on a par with humans that acts completely independently in every subject area. At the present time, such systems are still utopian. Weak AI, on the other hand, is designed for very specific task areas and applications. With structures that resemble human decision-making processes, weak AI solves recurring tasks with similar requirements. In this area, the use of AI is quite advanced today and concrete applications already exist.

Another area of artificial intelligence deals with neural networks (KNN). Artificial neural networks are quite complex systems in which the AI does not only make decisions based on individual factors, but rather runs through a process and analyses the task. Based on this, the KNN makes a prediction and develops a solution. Artificial neural networks are also characterised by the fact that parallel processing of processes takes place. In this respect, a KNN resembles the human brain and acts in contrast to computer systems that process tasks strictly one after the other.

The development of AI - from the idea to the first AI applications

The theory about artificial intelligence is older than many assume. As early as the mid-1950s, symposia were held in the USA to discuss AI applications. The first AI investments also date back to this time. Among other things, the Rockefeller Foundation financed a research project on artificial intelligence in 1956.

Artificial intelligence did not fulfil the first, high expectations. For example, the prophecy that a computer with artificial intelligence would win the world chess championship before 1970 did not come true. However, this prediction became reality in 1997. IBM developed the chess computer Deep Blue, which had simple AI mechanisms and defeated the world champion Garri Kasparov in six games.

In the years that followed, direct comparisons between AI and humans dominated the development of AI applications. In 2011, a programme called Watson managed to beat two well-known players in the quiz Jeopardy! In 2017, OpenAI competed with an AI application in a professional tournament in the real-time strategy game Dota 2 against the world's best players and won. According to the developers, it took the system only four months to develop these skills. Human players, on the other hand, train for years to reach the top of the world in Dota 2.

Much more interesting at present is the use of AI in research, production and IT security. Development is also progressing steadily in these areas. As early as the 1980s, researchers recognised that artificial intelligence has the potential to perform monotonous tasks independently and thus replace humans. In this context, machine learning is particularly important. However, the first systems of this kind often failed due to the so-called cliff-and-plateau effect. In this case, the artificial intelligence does not recognise the limits of its own competence and decides as usual. This was demonstrated in the 1970s with an AI application called MYCIN. Developed by Stanford University, MYCIN made diagnoses and therapy decisions for blood-borne infectious diseases. The system failed, on the other hand, when symptoms pointed to other diseases and continued to respond with suggestions appropriate for meningitis, for example.

AI applications received special public attention in connection with the term deepfake. Since around 2017, applications have existed that create moving images of human faces based on a photo and also generate matching mouth movements to spoken words.

The current status - what Artificial Intelligence does today

In the meantime, effective AI use is possible, even if it is not yet widespread. With Azure, for example, Microsoft offers a platform that enables access to AI applications. At the same time, the tasks that AI takes on are currently still very specific. We show a few examples below.

1) AI applications currently show their strengths wherever recurring, similar tasks are available on a large scale. AI applications are already replacing manual work in these cases. Practical examples can be found, for example, in the categorisation of digital information. Artificial intelligence is able to process documents and images based on their content. For example, software exists that recognises structures, shapes and colours on images and assigns tags or categorises the images based on this. One characteristic of artificial intelligence is that this software is capable of learning. The more similar tasks the software processes, the more efficient the categorisation.

2) Another area where AI is very advanced today is automated chat bots. Many companies have made AI investments in this technology and so these chat bots can be found on more and more websites. Visitors and customers use these chat bots to get information, give feedback or when help is needed. The chat bot is built around ready-made questions and answers. By analysing the questions, the artificial intelligence recognises what information users are looking for. In this way, the chat bot quickly provides the appropriate answers or forwards the questions to the relevant people in the company, who then personally deal with the request. The advantage of such intelligent chat bots is that users get help around the clock. In addition, a chat bot saves resources.

3) AI is also used in data mining. In the evaluation and analysis of large amounts of data, the use of AI reveals further strengths of this technology. First of all, AI is significantly faster than a human employee. However, the real strength lies in the ability to precisely recognise anomalies within large amounts of data. For such tasks, AI already outperforms humans.

4) Thanks to this characteristic, AI is increasingly finding applications in IT security. Certain programmes in IT security have the task of searching for suspicious events in networks (anomalies) and logs. Here, a permanent evaluation of accesses, data queries and similar processes takes place. These IT security programmes sound an alarm when they detect suspicious actions. It is important for the effectiveness of IT security that these applications work precisely. The proportion of false positives must be kept as low as possible, and at the same time it is important to reliably detect all unauthorised access.

In this area of IT security, AI benefits from two properties at once. The first is the ability to evaluate large amounts of data in real time. This is the log data from the network. The other is machine learning. Processes in corporate networks are usually very specific and follow recurring patterns, for example logins at a certain time or actions from a specific IP address. The AI learns this over time. In this way, the number of false positives is reduced and the IT Security AI reacts even more precisely to network anomalies.

Prospects and future outlook for AI use

One thing is certain: artificial intelligence is far from having reached its maximum potential. Development in this area is constantly progressing and has increased significantly in speed in recent years.

In the development of self-driving vehicles, designers are faced with the challenge that the systems have to make decisions continuously. Sensors and cameras provide the information on the basis of which control takes place. Parallels to existing AI applications can be seen here. For example, in the categorisation of images. Here, artificial intelligence already works autonomously and reliably recognises certain objects in photos. Artificial intelligence works in the same way in a self-driving vehicle. It identifies people, traffic lights, obstacles or other vehicles and makes decisions based on this data.

The topic of artificial intelligence, unlike many other new technologies, will particularly affect small and medium-sized enterprises. Through IT service providers offering specific solutions, SMEs gain access to this technology. With automated solutions that require comparatively few resources, deployment is possible with relatively low AI investments. Thanks to this feature, AI has the potential to provide crucial competitive advantages to small and young enterprises.

Limits and risks in the use of AI

There are also concerns about AI. One of the most common concerns is transparency. Artificial intelligence-based software makes decisions autonomously and without human support or control. Especially when humans rely fully on these results, transparency is lost. In addition, the machine lacks a sense of empathy, at least until now. This is why there is an increasing argument against the use of artificial intelligence in the field of medicine, for example.

Some employees see AI as a threat to their own jobs or jobs in general. In fact, this technology has the potential to completely replace manual work in many areas. The use of artificial intelligence is particularly interesting for labour-intensive, manual tasks in order to perform them cost-efficiently. Some see a further risk in the area of IT security and cyber threats. Artificial intelligence opens up the area of deepfake in particular to cybercriminals. This is how identity theft succeeds in the digital space, as criminals forge photos as well as video and audio recordings at a very high level. Recently, hackers have already been observed to carry out AI-driven attacks. Especially in the area of targeted spear phishing, attackers are using the possibilities of intelligent attack tools.

In some areas, artificial intelligence is reaching its technical limits. This is especially the case where creativity is required. In fact, AI is now even capable of writing texts. However, AI fails when it comes to writing texts with an independent content, such as a novel. On the other hand, AI can produce surprisingly good texts from collections of facts, such as those found in product descriptions or stock market headlines.