Workforce Development and Intelligence Analysis for National Security Purposes Proceedings of a Workshop (2018) / Chapter Skim
Currently Skimming:
4 Trends in Workforce Development
4 Trends in Workforce Development
Pages 43-54
The Chapter Skim interface presents what we've algorithmically identified as the most significant single chunk of text within every page in the chapter.
Select key terms on the right to highlight them within pages of the chapter.
From page 43... ...
Traditionally, he observed, people have focused on "identity" diversity, with inclusiveness being valued only because it is the right thing, a societal good. Others, he explained, have mistakenly viewed team diversity as akin to diversifying an investment portfolio, where the benefit comes from spreading risk, with the outcome of achieving the average return between stocks that do well and those that 43
|
From page 44... ...
He noted that diversity of perspectives among crowds with little to no knowledge does not lead to better prediction. He cited research that examined the economic predictions of 28,000 professional economists over a 40-year period and found that the combined crowd prediction of these economists was 21 percent better than the prediction of the average economist.2 Moreover, he observed, improving prediction does not require such large "crowd" numbers, noting that the predictions of two average economists are 9 percent better than those of one economist, adding that even the best economist (10 percent better than the average economist)
|
From page 45... ...
All three individuals in the group depicted in the figure provided their ideas for uses of blockchaining beyond cryptocurrency. Each was assigned a creativity score for the number of ideas generated and a Shapely value for each unique idea provided.
|
From page 46... ...
He pointed to numerous headlines and warnings about the threats posed by AI and automation to jobs, adding that others speculate about the pace of the development of AI and when it might match or overtake human intelligence. He reported that results of a 2016 survey of 352 machine learning researchers predicted that AI had a 50 percent chance of outperforming humans on all tasks and for less cost in 45 years.4 In Ysursa's view, with this lead time in mind, developers should focus on controlled, safe deployments emphasizing transparency, openness, and utility.
|
From page 47... ...
Innovations and technological advances, such as the personal computer, the Internet, mobile technology, the cloud, the Internet of Things,6 big data, AI/ machine learning, and automation, are transforming everyday lives, he said. He noted that the power of computing doubles every 18 to 24 months, and the resulting growth in the amount of data available provides the fuel for machine learning and AI.
|
From page 48... ...
He also mentioned that AlphaGo and other new versions of machine learning begin "from scratch" instead of learning from past human games, and reported that humans are improving by learning from the machine. Ysursa prefers the term "augmented intelligence" to "AI" because it signifies how humans and machines will work together to produce abilities "greater than human or machine." He pointed to examples of this augmented intelligence, such as use of computer vision in pathology to identify potentially abnormal cells for additional assessment by a human pathologist, noting that augmented intelligence would have the capability to compress the time required to obtain results from days or weeks to hours, potentially saving lives.
|
From page 49... ...
ARTIFICIAL INTELLIGENCE AND THE PROBLEM OF REPRODUCIBILITY IN SCIENCE Brian Uzzi, Northwestern University, focused his remarks on how to improve the scientific enterprise. He explained that AI could potentially help solve the problem that many scientific studies produce results that cannot be replicated when the studies are repeated by others, and suggested that the application of AI to the reproducibility problem has broad and potentially useful implications for the analyst workforce.
|
From page 50... ...
The method does not work well for some types of problems, he explained, is expensive, and is applied after a paper has already been published, which means that using the wisdom of crowds can cause damage before a problem is detected. Uzzi and his colleagues trained a computerized neural network to recognize papers documenting replicable studies, using 96 studies known to be replicable as a training set.
|
From page 51... ...
In Uzzi's view, human–AI partnerships can expand capabilities beyond those of humans alone, "expanding human consciousness." He also believes the techniques applied in his research could be extended to identifying retractions in science, "fake news," people who are using aliases on the Internet, and propaganda, all difficult problems that often elude human detection. DISCUSSION Following the presentations, workshop participants continued to discuss the implications, uses, and potential drawbacks of AI and their intersection with the information presented on diverse teams.
|
From page 52... ...
Often when a patient presents in an emergency setting with a complex problem, he elaborated, a group of doctors will see the patient sequentially rather than in a group, and will then meet to present and discuss their findings and conclusions to determine the strongest argument. Eric Eisenberg, University of South Florida, observed that the processes of moving toward consensus and preserving diversity of thought for sensemaking are in tension, especially when it is necessary to avoid acting prematurely or waiting too long to act.
|
From page 53... ...
Uzzi believes explanations will be derived as humans learn to work with AI and understand the tacit knowledge it possesses. However, Page believes that ultimately, algorithms will be developed to examine neural networks and provide explanations.
|
From page 54... ...
He suggested that advancing the process of selection, recruiting, and training may be attainable in the short term, while understanding teams and technology's role in teams is another important next step. He identified understanding augmented intelligence (humans plus AI)
|
Key Terms
This material may be derived from roughly machine-read images, and so is provided only to facilitate research.
More
information on Chapter Skim is available.