The definition of artificial intelligence – AI for short – is disputed. The ability to solve problems independently is often cited as a key feature of intelligence. Another characteristic is that in the course of a learning process, the solutions to a problem are constantly reviewed and improved. This is why computer programmes that work with the concepts of deep or reinforcement learning, for example, are called AI programmes or algorithms. Some scientists do not consider machine learning (ML) to be an intelligent process, while others regard ML as a subcategory of AI.
The global economic expectations for artificial intelligence are enormous (source: statista).
A classic example of ML are the various image recognition programmes that detect and categorise patterns. Other kinds of software analyse repetitive tasks and are then able to perform, adjust and optimise these tasks by themselves. At best, Artificial Intelligence is based on its “live template” – in other words, the neural networks that connect to each other in the human brain. Computers connect information in a similar way, for as long as it takes to recognise a pattern from individual strokes.
Agents, as they are called, try out different ways to achieve a defined goal. If there are errors, the process starts over, as “punishment”. If the ideal result is achieved, this serves as “praise” for the system. It saves the way it has discovered and has thus learned for the future. The better way with the better result is also pursued going forward. This is known as reinforcement learning.
To exploit the full potential of artificial intelligence, a lengthy start-up phase is necessary. First, the problem must be pinpointed and the right algorithm identified, which can cost a lot of time. Yet the machine programmes will learn more and more over time, like in a kind of continuing education.
The acquisition costs can be significant, particularly if the IT infrastructure is not geared towards AI. One solution is collaboration, as is the case with HHLA Sky, which has entered into a partnership with Deutsche Telekom for the 5G campus networks.
There is a shortage of skilled workers in most industrialised nations. This shortage can be mitigated by having AI applications handle specific tasks. Skilled personnel can become better qualified and focus on areas such as research and development, but also maintenance and support. This leads to the creation of new careers and exciting jobs.
As with autonomous driving, humans can no longer make errors since the task is performed by a machine. On the other hand, humans cannot prevent errors when the machine operates autonomously. A balance must be found.
AI is most often used for the optimisation in the digitalised transport chain. With its help, very accurate forecasts of traffic density or passenger frequency can already be made today. Many projects deal with predicting the estimated time of arrival (ETA). Buyers can thus recognise impending delivery bottlenecks in good time and producers can optimise their shipping. In principle, it is even possible to better integrate production processes with logistics. However, quantum computers would have to be used for this, as the number of possibilities often overwhelms "classical" computers.
Image recognition captures invoices or simple contracts to relieve the accounting department. The utilisation of various assets projected with AI enables dynamic pricing on the one hand, and energy and other resources can be saved on the other. The technology of digital twins is also an essential area of artificial intelligence. With the help of such models, processes or plants are simulated in advance, for example to test construction risks or materials. Since AI is still "in its infancy", it will still open up many areas of application that were previously difficult to imagine.
The question of where AI can best be used in the maritime industry is also addressed in our HHLA Talk with Nils Kemme from Hamburg Port Consulting.
At a HHLA terminal, for example, container transporters always follow the same route. AI programmes optimise such routes or the automated block storage management. They test how containers should best be placed in order to achieve optimal utilisation. Thanks to Deep Learning, AI can make predictions about which modes of transport the boxes are most likely to be delivered to at Burchardkai. In the COOKIE research project (COntainer Services Optimised by Artificial IntelligEnce), the new technology should detect damage to containers more quickly. Image recognition is also used for the inspection of container gantry cranes.
The HHLA rail subsidiary Metrans wants to use artificial intelligence in train dispatch to automatically check more than 100 different conditions and damage to freight wagons. This will identify damage more quickly and trains will be back on the rails sooner. The more efficient use of technology and the associated conservation of resources will, of course, also contribute to the desired climate neutrality in the end. More details can be found in the page about our “Balanced Logistics” sustainability strategy.
IT Strategy & Digitalization
Phone +49 40 74008-0
Mobile +49 175 9932038