AI for reducing inequality
A recent article (1) that indicates that universal access to generative AI tools such as ChatGPT increases productivity, differentially higher in those with lower skills, has policy implications. Philosophically, one could argue that productivity differentials observed in a population is fundamentally driven by initial conditions. The observation in the article further reinforces this idea. Universal access to internet is still lacking, albeit with many attempts at it by policymakers. This has led to inequity in education, healthcare and other areas. Now, generative AI, if not broadly available to the population, may enhance such inequity.
From the inception of humans, differential access to information and knowledge has been a problem. Information is a universal leveler and societies built on hierarchy used hoarded knowledge to increase status and to create and control lower classes. Generations reinforced this concept, with those access to information always starting ahead of those who do not, in the race of life. This sequential game played over 10K years resulted in such dramatic differences between people and populations in the status-quo, any policy other than complete access to all information for everybody, will be suboptimal.
To make matters worse, generative AI can not only aggregate, analyze and summarize information but also create knowledge from such diverse raw data. Those who have access to those tools can consistently over perform those who do not. The policymakers have to now worry about differential access to generated knowledge and not just raw information. Their inability to solve the simpler problem of information makes the harder problem even more difficult to tackle.
Universal access to information and knowledge is a necessary condition for societal development. Making generative AI tools and the internet broadly available to everybody could move us closer.
(1) Experimental evidence on the productivity effects of generative artificial intelligence | Science