Sophisticated Quant models run billion-dollar portfolios. Autonomous vehicles are on the road. AIs write music. AIs read maps and satellite images for the military. So when may an AI realistically lead a company as CEO?
If one considers a large company such as Microsoft (tkr: MSFT), CEO Satya Nadella does not spend his days writing code. Mr. Nadella has created value for Microsoft shareholders since his being named CEO in 2014 by taking a drastically different approach than his predecessor Steve Ballmer. For example, Mr. Nadella favors partnerships and co-opetition to sharp-elbowed competition. Nadella bet Microsoft’s future on the cloud (broadly defined) and AI (the “intelligent cloud”) whereas Ballmer’s strategy was..? Mr. Nadella has advocated for a common design language across Microsoft products.
Mr. Nadella’s operating principles could be observed by a machine learning platform and incorporated into the AI version of Satya Nadella. However, what about Mr. Nadella’s interactions with his direct reports, investors and other stakeholders? Many of these interactions and Nadella’s communication style could also be observed by deep learning models/neural networks but they would be next to impossible to replicate given the inputs that would be required to train the models, lack of computational power and so on. The AI would fail in a couple of key areas. 1.) The Long Tail: those events however significant or insignificant that rarely occur. A default set of operating principles would not be sufficient to address these rare events. Thus, the AI would lack the history/experience required to take appropriate action or to take action at all. 2.) Multi-discipline or “cross-domain” AI: we are decades away from an AI having the ability to carry a natural language conversation that covers multiple topics. Your guess is as good as mine (100 years from now?) as to when an AI may sit in a CEO chair. It is more likely that within the next 20 years CEOs will have next-level Augmented Intelligence where 5G applications powered by quantum computing will deliver data-rich intelligence on-demand that far exceeds the output generated by present day Internet queries or predictive analytics. This will mark the birth of the “cyborg CEO”.