By: Mark Cummings, Ph.D.

In the last year, it has become clear that GenAI will have dramatic impacts on society. While having tremendous productivity benefits, these impacts are coming so fast, are so broad and so deep that they can produce difficulties. A taxonomy of these effects helps to clarify the situation. With this clarity, it can be seen that the GenAI effects are taking us into a difficult period of adjustment. The challenge is to maximize the productivity benefits GenAI is bringing while minimizing the difficulties in the transition. 

If the period of adjustment is not well handled, chaos can result. One way to meet this challenge is to enlist the scientists and engineers who are creating the technology to help build a positive vision of the transformed society. A vision that we can work towards. A vehicle that could help with this a Union of Concerned AI Scientists and Engineers. The following section focuses on the taxonomy and its implications, which are the motivation for an effort rather than a description of how such an effort might be organized. 

Taxonomy

Then a particular area appears amorphous and confusing, it can be helpful to divide it up into its constituent parts. Each part can then be considered separately. Then, in combination. This division into component parts only focuses on the parameters of interest. It does not seek to cover all aspects. Within the parameters of interest, it does try to be complete.  

Figure 1 - Taxonomy of the social impacts of Gen AI
click to enlarge

The taxonomy of concerning societal effects is graphically represented in the Illustration above (see Figure 1). It does not address the dramatic productivity benefits of AI. Also, in some cases, the concerning societal effects are a result of our current inability to figure out how to take advantage of these benefits in some domains. It is a representation of the concerning societal effects as they are known today. As we gain more experience with GenAI, others may become visible. Also shown are representative aspects of each effect along with likely implications for society as a whole. Below, each portion of the taxonomy is discussed.

Job Loss

GenAI’s broad-based productivity gains offer tremendous benefits, while at the same time creating very large potential job losses. What is particularly concerning is that these job losses appear to be fundamentally different than those from previous generations of automation. In those previous generations, lower-skilled jobs were automated. This allowed people to move up to higher skill jobs, and the number of higher skill jobs increased. AI, coupled with robotics, is automating both low-skilled and high-skilled jobs. For example, Waymo in San Francisco now has a higher market share than either Uber or Lyft. At the same time, AI code generation is making it difficult for Computer Science graduates to find jobs. Estimates have been as high as 80% of jobs being replaced by AI. Recent statements by several CEOs make these large numbers seem eventually plausible. Starting with the Industrial