During our internship in Japan at Advantech, we got the opportunity to visit NextTech Week’s AI Expo.
Hot Topics
The venue was filled with tech-companies trying to promote and sell their products and services. It was interesting to see which direction Japanese AI-focused IT-companies are going towards.
LLM, RAG-implementations
RAG stands for Retrieval-Augmented Generation, a lot of Japanese companies are interested in this solution as it will decrease time spent on searching for information significantly. This is how it works:
- User asks a question to a chatbot.
- Instead of the LLM answering based on static information, it references (given) documents.
- LLM uses these referenced documents to supplement it’s static knowledge and answers.
Essentially, RAG improves LLM performance by incorporating information retrieval before answering. This can help reduce AI Hallucination, as well as reduce employees their search time by providing the LLM with private information.
AI-Inference
Another topic we saw coming back lots of times was Live- and Video-inference.
AI-Inferencing is a trained AI-model detecting things on camera or video. We can divide AI-Inferencing into three main tasks:
- Classification = assigning a label or value to an entire input.
- Detection = identifying and locating objects within the input.
- Segmentation = dividing input into regions or segments.
This was very popular at the venue because of it’s Wow-factor and it’s ask. Companies want the lowest-cost, but best-performance solutions. This gives a very competitive edge to the companies competing with each other.
Conclusion
This Expo was very meaningful in the way that we got to see what Japanese-AI technology is focussed on. We got to see all sorts of high-tech demo’s and informative booths.
Although most of them were Japanese-only, it was still a learnfull experience for us interns.