Artificial Intelligence has influenced the interaction between humans and machines like no other technology before. However, with increasing capabilities, regulations are also required to make their decisions transparent, understandable and safe.
James Cameron’s 1984 film “Terminator” is a classic that still shapes many people’s perception of Artificial Intelligence to this day. In the film, a computer system developed by the military was to start a devastating war against humanity in 1997 to protect itself from being shut down. Now in the 2020s, the horror scenarios predicted in “Terminator” and similar films have not come to pass. Nevertheless, AI has now become a part of professional and private everyday life, enabling a new era of human-machine collaboration: chatbots respond flawlessly to questions, smart home devices are controlled by voice and cobots work hand in hand with humans. AI systems are the basis for many innovative Human Machine Interfaces like voice and gesture controls.
Attempt at a definition
But what exactly is meant by Artificial Intelligence? There is, in fact, no universally accepted definition in science or practice. The draft of the European “AI Act” defines artificial intelligence as “software that is developed with one or more (…) techniques (…) for a given set of human-defined objectives and generates outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with.” At its core, AI uses algorithms and complex mathematical models to analyse data and identify patterns. Like humans, AI is supposed to learn from experience, make judgments and solve problems independently – to be able to perform tasks increasingly better.
Several subareas
The term “artificial intelligence” covers several subareas, including Machine Learning and Deep Learning. In Machine Learning (ML), the system independently discovers connections based on example data. Thus, AI systems can learn from data and solve problems on their own without being explicitly programmed in the form of rules. Machine Learning is particularly suitable for recognising and generating so-called “patterns” from existing datasets – for example, the system can thus recognise which gesture a hand performs. Deep Learning goes a step further by automating further aspects of the learning and training process. Deep Learning algorithms can decipher unstructured datasets such as texts or images, so much less human intervention is required.
AI becomes creative
Thanks to ChatGPT, the latest technological developments in AI are currently a hot topic – the so-called generative AI and foundation models. They are capable of independently generating content such as software code, texts, images and music. This sets generative AI apart from “classic” discriminative AI, which is designed to differentiate and classify input but does not create new content. Compared to previous AI models, generative systems are particularly powerful as they are trained based on a very large amount of data. With this breadth and amount of information, foundation models can, for example, translate between languages and systematically work through tasks. In terms of HMIs, they offer advantages in voice control, among other things: the conversation with the machine is more natural, and responses can be given based on the context. Additionally, generative AI can process complex commands better. Instead of simple actions, users can give more detailed instructions: the AI can interpret these, ask follow-up questions if necessary and generate appropriate actions.
Legislative regulation
Despite all the advancements, artificial intelligence is still far less sophisticated than portrayed in Hollywood blockbusters. Nevertheless, regulations are needed so that AI can make the right decisions. An AI system is only as good as the database with which it was trained. There have already been several practical examples of AI systems having a certain “bias” because the database was not diverse enough. If, for example, a language model were trained only with a North German dialect, the later system would have problems understanding someone from southern Germany. To make AI safe and trustworthy, the European Union has introduced legislation to regulate its development and use: the AI Act. This regulation, which is expected to come into force in 2026, is intended to ensure that AI systems used in the EU are safe, transparent, understandable, non-discriminatory and environmentally friendly. AI systems should be monitored by humans and not by automation to prevent harmful results. Thus, the “Terminator” should remain pure science fiction in the future.
Boom in Edge AI processors
14.54 billion US dollars in 2022
54.38 billion US dollars in 2029
To realise AI functions in Human Machine Interfaces, so-called Edge AI processors are important as they can evaluate the data directly on-site and thus ensure a quick response to commands. According to Maximize Market Research, the market will grow by an average of 20.1 percent annually in the coming years.
