top of page
  • Madeline Weiss, Director

Humans Plus AI Create Superminds


Recent articles have sounded loud alarms about AI usurping human jobs. No one can deny that reality among factory workers and lower skilled jobs. But another perspective, the benefits when computers and people think together, is beginning to get the attention it deserves in two new books: Superminds, by MIT researcher Thomas Malone, and Human + Machine: Reimagining Work in the Age of AI, by H. James Wilson and Paul R. Daugherty. Both books stress the benefits to organizations and society of leveraging AI to complement and augment human capabilities rather than replace them. Although they reach the same conclusions, the books take different approaches that complement each other and enrich executives' options and strategies.

In Superminds, Malone distinguishes between computers' specialized intelligence (the ability to achieve specific goals in a given environment) and people's general intelligence (the ability to achieve a wide range of goals in different environments). Malone predicts it will take at least 20 years for computers to emulate people's general intelligence. But when organizations properly harness the general intelligence and other specialized skills of people with the knowledge and specialized capabilities of computers, the result can bring more intelligence than any person, group, or computer has before. Malone offers many current examples. One example is the Crowd-Forge system, which managed the process of creating documents, such as an encyclopedia article on New York City. The system hired people registered to perform micro tasks on Amazon Mechanical Turk to create an article outline (e.g., attractions, brief history) and then hired others registered on the site to list facts related to the category (e.g., attractions, brief history). It then hired still others to write paragraphs using the facts. The system then organized the paragraphs into a coherent article that independent judges subsequently rated better than similar articles written solely by people. Another example is a competition among universities organized by the U.S. intelligence community to develop the best methods for predicting the answers to such questions as (1) Will Serbia be officially granted European Union candidacy by December 31, 2011 and (2) Will the six party talks on the Korean Peninsula resume before January 1, 2014. The winning team at the University of Pennsylvania sourced ideas from a crowd of thousands of online volunteers, then discovered through AI that some in the crowd were super forecasters. These super forecasters were assembled into small teams, which made predictions that were about 30% better than those of experts in the intelligence community who had access to intercepts and other secret data.

One can easily envision future superminds creating far-superior predictions, diagnoses of diseases, and solutions to intractable problems than the smartest humans alone.

In Human + Machine: Reimaging Work in the Age of AI, Wilson and Daugherty identify three crucial roles that humans must perform to benefit from human/AI collaboration: training machines to perform work through machine-learning algorithms; explaining machine conclusions to nonexpert users (e.g., disease diagnoses or rate offers on a credit cards or mortgages); and sustaining the proper functioning of AI systems. With such human assistance, machines help humans expand their abilities in three ways: amplifying human analytic and decision-making by providing the right information at the right time; interacting with customers and employees in novel, more effective ways; and embodying human intelligence in the form of a robot that works alongside humans.

How can organizations best combine the three crucial human roles with the three ways that machines can expand human abilities? According to the authors, executives must reimagine and redesign those operations that can benefit the most from greater flexibility (to customize products); speed (catching fraud situations more quickly); scale (selecting job candidates from a very large pool); decision making (predicting part failure and suggest actions); and personalization (anticipating cruise guests' preferences). Such redesigned operations then lead to developing employees with "fusion skills" - those that enable them to work effectively at the human-machine interface. With fusion skills, employees can delegate appropriate tasks to the new technology, as when physicians trust computers to help read X-rays and MRIs, and work well within AI-enhanced processes. In the longer-term, employees must be able to teach intelligent agents new skills.

Through their frameworks and examples, Malone, Wilson, and Daugherty suggest that human/machine collaboration is only limited by the imaginations of humans.

(Note: Thomas Malone presented his superminds ideas at the June 2018 Advanced Practices Council meeting.)

#ArtificialIntelligence #superminds

31 views0 comments
bottom of page