An argument has emerged that beyond simply emphasizing the collaboration between humans and artificial intelligence (AI), we want to accurately understand in what cases synergies occur. Contrary to vague expectations, it is commonly more practical for humans and AI to work individually.
The MIT Center for Collective Intelligence (CCI) recently published a report in Nature:When the mixture of humans and AI is helpful: A scientific review and meta-analysis.‘ published a paper.
The researchers first introduced, “The usage of AI tools is increasing not only in skilled fields but in addition in on a regular basis activities, and human-AI systems have tremendous potential given their complementary characteristics.”
“Nevertheless, increasingly more research is showing that human-AI systems don’t necessarily achieve higher results than the most effective humans or AI alone,” he said, citing reasons resembling communication barriers, trust issues and ethical concerns. I picked it.
Subsequently, it was revealed that this study is about when humans and AI complement one another and to what extent they complement one another. To this end, they conducted a literature review and meta-analysis to quantify the synergy of human-AI systems and discover aspects that specify it.
Most studies have focused on AI augmenting human capabilities, he noted. That is an ‘implicit assumption’, otherwise the higher of the 2 might be chosen. This implies the lack of human jobs or the uselessness of AI.
The researchers analyzed 370 results from 106 different trials published between January 2020 and June 2023. This included ▲cases where humans handled the duty alone, ▲cases where AI worked alone, and ▲human-AI collaboration.
The outcomes showed that, on average, human-AI systems performed higher than humans alone. Nevertheless, AI fell wanting handling the duty alone, and collaboration proved to be useless, especially in cases where humans were excellent. In other words, collaboration is meaningless aside from using AI to do things that humans are usually not good at.
He explained that that is as a result of the undeniable fact that humans ultimately have the facility to make decisions in all collaborations. In cases where humans are superior to AI, using AI is useful, but in areas where AI demonstrates superior capabilities than humans, human intervention only hinders.
The researchers also emphasized, “Although these results could appear disappointing, what is vital is that now we have identified specific aspects that do or don’t contribute to the synergy of human-AI systems.”
For instance, it was revealed that AI was not helpful in tasks related to decision-making, but made a major contribution in tasks related to content creation. The title of the paper, ‘When the mixture of humans and AI is helpful’ refers to this.
He argued that it’s time to transcend emphasizing human-AI collaboration and examine what varieties of work collaboration is helpful for. He emphasized that much additional research is required to bring out the synergistic potential of AI.
Reporter Lim Da-jun ydj@aitimes.com
