Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

 Artificial intelligence (AI) and machine learning

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs? 

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

Machine learning processes allow large amounts of data to be analyzed and processed quickly and precisely. Machine learning and artificial intelligence (AI) show great potential in application areas such as autonomous driving or Industry 4.0, but also bring new challenges for safe and reliable use. That is why the Fraunhofer Institute for Cognitive Systems IKS is researching to comprehensively secure artificial intelligence so that it can also be used in security-critical areas.

 

What is Artificial Intelligence? What is machine learning?

Artificial intelligence is a branch of compt science. It mimics human cognitive abilities by recognizing and sorting information from input data. This intelligence can be based on programmed processes or generated through machine learning.

 

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

In the past few years, great progress has been made, especially in the area of ​​machine learning. This is mainly due to the increasing availability of large amounts of data and high computing power, which are a basic requirement for the complex calculations of machine learning.

 

An algorithm learns to perform a task independently through repetition. The machine is based on a specified quality criterion and the information content of the data. In contrast to conventional algorithms, no solution is modeled. The computer learns to recognize the structure of the data independently. They are only told from where and to where to transport the objects. How exactly the robot grips, it learns through repeated trial and error and through feedback from successful attempts.

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

Neural networks and deep learning

Neural networks are a sub-area of ​​machine learning. These learning algorithms are inspired by nerve cell connections in the human brain. The brain processes information through neurons and synapses. Similarly, artificial neural networks consist of several rows of data nodes that are networked with one another with weighted connections.

The neural network is trained by repeatedly presenting it with data. Through this repetition, the neural network learns to classify the data more precisely each time. This works by constantly adjusting the weighting for the individual connections between the neuron layers. The model generated in the learning runs can then also be applied to data that Artificial Intelligence has not yet learned about during training.

If neural networks have hidden neuron layers that are not directly coupled to the input or output layer, they are called "deep neural networks" . Deep neural networks can have hundreds of thousands or millions of layers of neurons. In this way, more and more complex problems can be solved in so-called "deep learning".

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

Diverse areas of application for machine learning processes

Machine learning processes are used in very different areas:

Image recognition: Machine vision algorithms can be used to recognize and categorize images. So a lot of data can be processed at lightning speed. Machine vision is used, among other things, in medical diagnostics or face recognition, but it can also be used to translate handwritten characters into print. Image recognition is also crucial for autonomous driving.

Speech recognition: Recognizing and interpreting verbal language can also be learned using machine learning processes. These algorithms are used in voice assistance systems, for example.

Semantic speech recognition: written text can be semantically interpreted using machine learning. This allows context-related translation applications or catboats that independently generate meaningful solutions.

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

Pattern recognition: Machine learning processes can also be used to recognize patterns in sequences of events that humans cannot recognize due to the large number of data points, variables and dependencies. For example, an AI can learn error patterns in vehicle electronics from data and compare these anomalies with the behavior in operation. Anomalies are recognized more quickly, which means that countermeasures can be taken early on. For example, by replacing a component before it actually causes a fault.

Process optimization: The recognized patterns can also be used as an information base for optimization processes.

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

Artificial intelligence as a driver of innovation

Automated machines must be able to react quickly and reliably to their environment. These skills are reinforced through machine learning. However, AI applications do not function properly per se. Errors in the selection of suitable training data, in data generation and processing, can lead to dangerous malfunctions of the system, which the AI ​​itself cannot recognize and prevent. The focus of the Fraunhofer Institute for Cognitive Systems IKS is to secure AI-based technologies through extended and adaptable software architectures.

 

Particularly in safety-critical applications, it is important that the system in which the artificial intelligence is built works absolutely safely and reliably. The challenges in securing machine learning from data differ greatly from the challenges with conventionally programmed software. The training data used play an important role in the quality of the generated neural network. If the data is not representative of the multitude of situations the system is later confronted with, the model is not good enough and makes bad decisions. So that the model also applies to data that has not been learned, the model must be robust and abstract. So it must not be too close to the training data, which results in over fitting and the model is not abstract enough for new data. On the other hand, there must be no under fitting, i.e. a model that is too simple and that does not describe the structure of the data with sufficient precision.

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

Safe AI: safeguarding machine intelligence in autonomous driving

For example, precise data analysis and process control are essential for autonomous driving. The vehicles must be able to recognize their surroundings, interpret them accurately and then optimize their actions.

In contrast to classic algorithms, machine-trained programs have the problem that the individual learning steps cannot be interpreted by humans. Due to the automatic adjustment of the weightings in neural networks, only the input and the result remain accessible for human control. A research goal under the catchphrase »Explainable AI« is therefore to design neural networks in a comprehensible manner. Because since the AI's decision-making process is opaque, the security and reliability of the AI cannot be assessed without further ado.

So far, the machine vision (perception) of the AI ​​has not yet been so reliable that it is suitable for safety-critical use in autonomous vehicles on public roads. First of all, ways must be found to make the uncertainties of the AI ​​quantifiable in order to be able to evaluate the perception in a meaningful way. The Fraunhofer Institute for Cognitive Systems IKS is working on solving these challenges and creating verifiably reliable systems by quantifying the uncertainties of AI. In this way, the previously non-in transparent classification of AI can become manageable. Because only if AI systems are understandable for humans are they safe enough to be used, for example, in autonomous driving on public roads.

One approach of the Fraunhofer IKS is to supplement the AI ​​with an extended software architecture. This monitors the artificial intelligence and checks the decisions made for plausibility. At the same time, the dynamic safety management approach gives AI more freedom than classic safety approaches, which always start from the worst-case scenario. In this way, the advantages of fast data processing through machine learning can be used and at the same time possible wrong decisions can be caught. This is particularly important if, as with autonomous driving, wrong decisions by the AI ​​would put human lives at risk. 

Artificial Intelligence (AI) and Machine Learning, Robot is better then Human? Robots are safe to live with? Robot are stealing our Jobs.

Machine intelligence in Industry 4.0

Even in the digitalization of industry, the so-called Industry 4.0, machine intelligence enables optimized planning and better predictions. Automated and networked machines recognize their environment and can independently adapt their actions to it. The human-robot collaboration without protective fence is only possible through machine learning and artificial intelligence. Here, too, the AI ​​must be secured so as not to endanger human life.

In the joint project “REMORA - Multi-Stage Automated Continuous Delivery for AI-based Software & Services Development in Industry 4.0”, Fraunhofer IKS is working with other partners, for example, on the simple integration of AI services in Industry 4.0. The aim is to simplify the integration of AI for real-time machine data analysis and to create tools for high-quality and dynamic machine data