The research areas in which AI can help is as long as impressive, according to the FAZ. Here are two examples:
- Algorithms can interpret economic activity in regions for which there is actually too little data by using high-resolution satellite photos. If, for example, a quadrant has more roads and factories than fields or forests, this tends to indicate higher economic performance.
- According to the FAZ, one area in which machine learning has been used for some time now is the analysis of texts: This is worthwhile, for example, in the publications of central banks, which generally choose every word carefully in order to anticipate changes in their policies – but also in journalistic media. The computer can recognise linguistic patterns that humans would probably not notice in thousands of pages of text.
However, according to the FAZ, computers cannot replace humans in other central questions of economics. Two examples here as well:
- The distinction between correlation and causality can only be made by those who have more understanding of the content than the AI can glean from the data. According to the article, the AI only recognises common patterns in the data – but not how these patterns come about. This makes the algorithms basically very good at predicting future events – but also vulnerable to anything that does not fit into this pattern.
- According to the FAZ, the size of data sets in research also limits the usefulness of AI for economic analyses. Find out why here:
