A company does not necessarily need a proven expert to be able to use artificial intelligence. A Fraunhofer study shows how small and medium-sized companies can proceed instead.
Artificial intelligence (AI) helps to optimize production processes and thus save money. However, small and medium-sized companies often lack the expertise to use this future technology. Although they can collect the necessary data, they fail to analyze it. This is where large cloud providers can help. They offer simple digital tools that process large data sets and deliver AI solutions. Experts speak of "machine-learning-as-a-service platforms". This means that any company can get started with artificial intelligence without a great deal of experience and have models developed that automatically detect faulty workpieces, for example.
Comparison of the most common use cases on four platforms
But which platform is suitable for which task? The Stuttgart-based Fraunhofer Institutes for Manufacturing Engineering and Automation IPA and for Industrial Engineering IAO have compared the approaches of the four largest providers - AWS, Google, IBM and Microsoft. They implemented solutions for four use cases that frequently occur in practice and cover four categories of data: Tabular data, text, image and time series data:
- Customer churn: it is beneficial for hotels to know early on which guests are at risk of canceling. There may already be an indication in the tabular booking data. AI can detect it and develop a corresponding algorithm.
- Text categorization: Texts can be assigned to different categories, such as culture, sport and politics. For example, a press agency can automatically maintain an archive.
- Image recognition: Image analysis plays an important role in production. Camera systems can be used to detect defects on the workpiece. AI helps to automate this inspection. The AI learns to recognize defects from a large number of annotated images provided with metadata.
- Tool wear: Replacing a milling head at the right time saves money. If you intervene too early, you give away material; if you intervene too late, you risk a long production standstill. AI learns to interpret the time series data of vibrations and power consumption in order to correctly assess the condition of the milling head.
As a rule, the following applies to AI solutions: The more data available and the better the quality of the data, the more reliably the model obtained works. When comparing the platforms, the Fraunhofer scientists always chose the most accessible solution. Often, all they had to do was upload the data sets and add an annotation: In the case of image processing, for example, this would mean adding the words correct or incorrect to each image. The platform then provided the desired model including the prediction accuracy.
Results
The Fraunhofer study showed that the solutions from all providers demonstrate strengths and do not require in-depth specialist knowledge. Of course, there are one or two differences. For example, some platforms are more intuitive to use than others. Also, some AI models only run on the provider's cloud, while others can also be exported and installed on a company's own servers.
The study "Cloud-based AI platforms - opportunities and limitations of services for machine learning as a service" shows which platform can be recommended for which use case. It is available for download at the following link: https://www.ki-fortschrittszentrum.de/de/studien/cloudbasierte-ki-plattformen.html
