Research on Machine Learning and Artificial Intelligence
Over the years, the field of Artificial Intelligence (AI) has seen a number of fundamental paradigm shifts. The first period of Artificial Intelligence was based on approaches that used symbolically expressed human knowledge for inference, for instance in rule-based expert systems.
In contrast, the successes of the current era of Artificial Intelligence are in large part brought about by Machine Learning (ML) systems that have become capable of using very large amounts of data to fit extremely high-dimensional models, such as Deep Neural Networks.
© Quardia Inc. –
Knowledge and Context
Current AI systems thus ignore two powerful sources of intelligence: knowledge and context. Passing knowledge to learning systems will be the key to AI systems that solve general tasks. Such knowledge can consist of learned representations or partial models, simulations or laws of the physical world, or explicit knowledge provided by humans.
Secondly, many current AI algorithms are created based on the assumption of an abstract and self-contained digital computer architecture. In reality however, there is abundant context. Such context may consist of awareness of the available resources and hardware architecture, of access to the world for active sensing and experimentation, or of interactions with humans.
The strategic direction of research at the Lamarr Institute is fundamentally shaped by our conviction that we need yet another paradigm shift toward a third generation of AI systems that we call AI³, or Triangular AI.
The three dimensions of Triangular AI:
data, knowledge and context
The Lamarr scientists therefore commit to the central vision and mission to bring forward AI3 – Triangular AI combining data, knowledge, and context by examining these three elements and their interrelationships, creating algorithms that can make use of them, and with the long-term goal of ultimately joining these together in an AI system exhibiting general intelligence.
Key Research Areas
The five areas into which we organize research at the Lamarr Institute reflect our commitment to AI³ – Triangular AI. These research areas are interrelated and address questions on how to integrate data, knowledge and context to build AI solutions which operate in a resource-efficient way and deliver powerful yet robust, explainable and trustworthy results.
Hybrid Machine Learning
Hybrid Machine Learning means the integration of data, knowledge and context, thus forming the basis for much research at the Lamarr Institute.
Integrating data, knowledge and context into Machine Learning is a promising approach for creating ML solutions that are more efficient, robust, explainable, and trustworthy. Moreover, hybrid systems that integrate knowledge representation with statistical learning techniques are expected to be less biased and require fewer training data. Therefore, researchers at the Lamarr Institute are currently working on creating such hybrid ML systems that bridge the gap between data- and model-based approaches.
© Fraunhofer IAIS
© Fraunhofer IAIS
Resource-aware Machine Learning
Resource-aware Machine Learning aims to adapt technologies to save energy, memory and computational resources.
Researchers of the Lamarr Institute are dedicated to developing sustainable and environmentally friendly Machine Learning solutions that save energy and computational resources. For this purpose, we study the connection between hardware and Machine Learning. It is our goal to make Machine Learning available even on devices with restricted computing power and limited energy and memory resources.
Human-centered AI Systems
Human-centered AI Systems are designed to interact with humans and deliver explainable and comprehensible results.
At the Lamarr Institute, we are developing human-centered approaches for bridging the gap between ML methods and human minds. On the one hand, human-centered AI systems adapt to human goals, concepts, values, and ways of thinking. On the other hand, these systems take advantage of the power of human perception and intelligence. Visual Analytics play a key role in combining human and machine intelligence. Thus, ML models are developed with involvement of human knowledge and then use this knowledge in generating explanations.
© sittinan –
Trustworthy Artificial Intelligence
Trustworthy AI builds on robust, verifiable procedures and forms the basis for certifying AI applications.
Trustworthiness has many facets and concerns, apart from computer science, diverse subjects, ranging from psychology and philosophy to economy and law. Consequently, the scientists from the Lamarr Institute who are working on trustworthy applications of Artificial Intelligence are members of an interdisciplinary team. For their research, they consider the overall pipeline of data gathering, storing, accessing, sampling, preprocessing, model selection, modeling, adapting, and applying the model to a process with a certain outcome and impact.
Embodied Artificial Intelligence
Embodied Artificial Intelligence refers to AI that is embedded in physical systems, such as robots, and can interact with the surroundings.
In contrast to classic ML in robotics, embodied AI encapsulates all aspects of interacting and learning in an environment: from perception, via understanding, reasoning, and planning to execution respectively manipulation. Just as human learning is based on exploration and interaction with the environment, embodied agents must improve their behavior from experience. Thus, embodied AI brings together multiple fields, such as computer vision, environment modeling, and prediction, planning, and control, reinforcement learning, physics-based simulation, and robotics.
© Michael Neuhaus –
Application Fields and Interdisciplinary Research Areas
While fundamental research is at the core of the Lamarr Institute and takes center stage, we believe that the full potential of Machine Learning and Artificial Intelligence in science can only be realized with active engagement and interdisciplinary research work that does not view other disciplines as pure application areas, but as research questions in their own right.
Together with our Lamarr colleagues from other disciplines, we have identified five application fields and interdisciplinary research areas in which we have already demonstrated the applicability of Machine Learning and which will play key roles in our future research:
- Planning and Logistics:
AI solutions help optimize transportation and mobility processes and save resources.
Using ML methods, large data quantities are analyzed and evaluated thus providing new insights into the physical world.
- Industry and Production:
AI allows the use of machine models and production data for intelligent control and increased efficiency.
- Life Sciences:
AI is revolutionizing medicine, accelerating drug development, and assisting in the study of biological systems.
- Natural Language Processing (NLP):
Understanding spoken and written language is the basis for many AI applications that facilitate our everyday lives.
© Anke Liepertz-Peter -