What is an AI assistant?
An AI assistant is an adapted version of a pre-trained language model that solves specific tasks through configuration rather than training.
Users can implement versatile applications by making simple adjustments such as adding context information, instructions or process descriptions.
Goal of our hackathon
The aim of the hackathon was to convey the Industry Core Standardsby asking the participants to develop a frontend using Gen-AI that works with two REST interfaces of the Item Relationship Services interacts.
The participants could choose between three approaches:
- Independent implementation without assistance from AI
- Use of any AI
- Use of the AI assistant provided by us
The implementation at a glance
The Hackathon was structured in three phases that built on each other and were each accompanied by an AI assistant. Participants could call in the assistant as required or use other methods. This flexibility made it possible to test the assistant or develop solutions independently.
Phase 1: Setup of the front end and connection to the interfaces of the item relationship service
The first phase involved setting up the front end and connecting to the REST interfaces of the item relationship service. The AI assistant provided support here by providing specific code snippets and explanations to make it easier to get started with the interface connection.
Phase 2: Visualization of a Tier 3 data chain
In phase two, a tier 3 data chain was mapped using digital twins. These twins and their relationships across a bill of materials were to be visualized as a graph, tree or similar form. The AI assistant provided suggestions for possible visualization methods and helped to interpret the existing data.
Phase 3: Enrichment of semantic models
In the final phase, the model was enriched with semantic data linked to the digital twins. The AI assistant made suggestions on how to integrate this additional information and further refine the visualization.
Conclusion
The increased efficiency should be particularly emphasized, as time-consuming research was reduced and the focus could be placed on communicating the Industry Core Standard with digital twins and the parts list. This made it possible to convey the complex topic in a reasonable amount of time, which would probably not have been possible without AI. The participants were enthusiastic about the modern technology and the rapid prototype development made possible by the direct provision of suggestions and code. The hackathon also provided a platform for discovering new applications for the AI assistant and trying out innovative ways of working in the automotive sector.
Nevertheless, there were also challenges. The solutions generated by the AI were often similar and offered little scope for creative approaches, as the users were strongly steered in one direction by the ready-made suggestions. Another critical point was the testability of the AI assistant itself. Validating its results proved to be time-consuming and automating such tests was difficult to implement, which can lead to increased costs in the long term.
A hackathon with AI assistants showed how much potential there is in this technology. I would choose this approach again, as the AI assistants worked efficiently and made it clear how much they can change software development. It is up to us to exploit this potential and use AI as a tool for more efficient processes.
