Quick Thoughts on the Higher Order Effects of AI
Predicting trends for new technologies is tough. We are bound by what we know at any point in time. We can extrapolate from use cases, but customer acceptance, technology advancement, regulation, societal backlash, and unintended consequences can throw so many discontinuities that there could be many future trajectories. Moreover, while short-term trends could be logically visualized, predicting the 2nd and 3rd order effects is usually foolhardy. For instance, when the automobile replaced the horse and buggy, there were immediate first-order effects. People spent less time traveling to work. The second-order effects were the building of better roads and a highway system. Third-order effects included people moving away from downtown and the emergence of suburbia and malls. Later, more businesses started relocating to these suburbs. What seems logical in retrospect was, of course, tough to predict prospectively.
When we look at predictions regarding AI today after generative AI tools have created broadly accessible demonstrable value…. we see predictions as somewhat bipolar. People are in utopian or dystopian camps…perhaps reflective of the current hype cycle. However, is there any logic that can predict some higher-order effects of AI? For instance, can we project an increasing scope of AI impacts from individual tasks to jobs to organizations to industries? While the technology evolves at a furious pace, propelled by data network effects and the motivation of tech companies that need to stay ahead of the game, the most evident first-order effect is the use of AI as a tool – enhancing individual productivity and convenience. It’s easy to get generative AI to write letters and essays, summarize documents, draw out pictures, and see it get better with more precise prompt engineering. It is also easy to see AI increasingly incorporated into products and services, making them even smarter.
But then, what are the higher-order effects? Once AI is better integrated with redesigned processes, then are we going to see more movement from the tool effect to the substitution effect? It is easy to see domain-specific jobs be automated – like accountants that take raw information and fill out tax forms or create accounting statements according to generally accepted accounting principles. But from there, can we envision broader automation of coordinated jobs that require knowledge and judgment? If AI can deal with uncertainty and equivocality — access knowledge relevant to a strategic decision and analyze the consequences of alternatives, what is the role of the C-Level Team, including the CEO? Considering even greater scope, will there be a point where AI can disinter-mediate entire industries? For instance, the central role of a bank is to manage the risk of money on the supply side and the demand side. AI can discriminate good loan candidates from bad ones, manage investments, and essentially replace a bank. (I have written about a completely automated company that buys and rents apartment units without any human involvement). Similarly, if AI can accurately predict the health outcomes of individuals or their driving risk, what happens to risk pooling (where low-risk individuals subsidize high-risk individuals), which is the foundation on which the insurance industry is built?
Of course, as I said earlier, predictions are subject to many forces that create discontinuities. Much of the current AI hype is based on language models, which is the cumulative effect of human expression. Clearly, a collective human “bounded rationality” is being captured through such expression. However, humans also learn through other modalities – vision, smell, taste, touch. When we can create AI training sets that fully incorporate such modalities – we can speculate endlessly about the possibilities of artificial general intelligence. Such predictions are nothing more than speculative fun – until they happen.
Varun Grover
George and Boyce Billingsley Endowed Chair and Distinguished Professor, Walton College of Business at University of Arkansas
