The impact of AI on research productivity.

 

Certain observers have detected a possible decline in the productivity of research (OECD, 2023[69]), manifested by a slow-down in the number of significant discoveries and impactful inventions (e.g. the “Eroom’s law”, the reverse of Moore’s law that points to a doubling of the effectiveness of computer chips every 18 months). 

Factors often mentioned as affecting this possible decline include notably 

1) the increasing quantity of knowledge to manage in order to design new knowledge, which creates a “burden” for researchers; 

2) the increasing complexity of phenomena to be analysed by scientists (the simplest ones having been grasped already); 

3) a growing administrative burden as researchers have to do increasing amounts of paperwork; 

4) distorted incentives that orient researchers towards research topics that are “low risk, low reward”, or exploitation at the expense of exploration. 

Artificial Intelligence, notably GenAI, could alleviate the first two factors, as it can manage great quantities of knowledge and it can model extremely complex processes. It could also alleviate the third one, as it can help in paperwork (submitting grant applications, etc.). As for the fourth factor, its impact is uncertain, as explained above.