"AI on AI" Robo-Report - At Robotic Online Intelligence (ROI), we have combined LLMs with our Kubro(TM) Information Engine for automation of market research, in order to generate this 'robo-report' - using AI to summarize the latest AI announcements and news directly picked up at source from blogs, press releases, and announcements from a selection of around 30 tech companies driving the current AI waves.
Such a simple output but involves a number of steps:
1. Research and encoding of what matters/sources of info (human effort)
2. Search/crawl and capture the info (automation)
3. Taxonomy/domain model (human effort)
4. Filtering/classification (automation)
5. Summarization of individual items (LLM + code)
6. [Optional] Summarization of the summaries (LLM + code)
7. Packaging into digestible format (automation)
It also served as one of our experiments in the practical deployment of LLMs in market research. We have been running robo-reports in different domains as 'search and alert tools' for years but now with the Large Language Models, so much more can be done.
Still, our take is that while summarization and data extraction from shorter bodies of text work well, for now, it's only one component in the workflow.