This week on the ATG: The Podcast, we featured a conversation between Matthew Ismail, Editor in Chief of the Charleston Briefings, and Todd Carpenter, Executive Director of NISO and author of a chapter in the recent Charleston Briefings title Artificial Intellegence in Libraries and Publishing. This title was co-edited by Matthew Ismail and Ruth Pickering, Co-Founder and COO of Yewno and includes chapters from Daniel W. Hook And Simon J. Porter, Catherine Nicole Coleman And Michael A. Keller, James W. Weis And Amy Brand, Ruggero Gramatica, and Haris Dindo in addition to Todd’s contribution.
In addition to the conversation from the podcast, here’s an excerpt from Chapter 6 “Would a Google of AIs be Able to Predict the Future?: Where is Information Management Headed in a World of Artificial Intelligence?” by Todd Carpenter:
“Before we can understand the future of machine intelligence, we should certainly ask if we even have a common understanding or expectation of what human intelligence is. If one were to ask a random assortment of passersby to define intelligence, one suspects that there would be no common definition among them. And, indeed, we all know that various people have very different intellectual strengths. One person who is brilliant at mathematics or physics, after all, might be quite clueless in the world of practical affairs. The brilliant orator might not be able to balance his or her checkbook. An athlete able to read an opponent’s physical “tells” and therefore win championships but might not be able to speak eloquently. The most artistic pianist might not be able to remember where they left their car keys.
“The function and mystery of the mind has occupied the attention of countless philosophers, neurologists, psychologists, and linguists over many centuries. Tens of thousands of books have been written about the human mind and how it works. Given how intimately connected we are to our minds, the most amazing thing is that our minds are still poorly understood. And if our own minds remain so mysterious, it is not surprising that we have had limited success translating that vague understanding we do have into silicon, transistors, and computational switches.
“To be sure, narrow AI—that is, the application of AI toward single tasks or problem-solving—has made great strides in recent years. But this narrow form of machine intelligence does not necessarily translate to intelligence in other spheres—something that applies equally to humans as it does to machines. There may only be a few instances of machine learning that can perform in multiple spheres, but the true polymath is also a very rare human being.
“One of the errors in expecting the creation of a human-like AI is that we commonly underestimate the underlying complexity of many basic human activities. We mistake the narrow capacity of a machine in a very specific task or domain for a more general intelligence. There is not now, nor will there be any time soon, a generalized AI that is as adaptable as even a chimpanzee or a dog. Machines can be good—far better than humans in some cases—at a very limited set of activities that require rapid computation or the recognition of patterns in data. Some of the things machines do might even fall under the rubric of intelligent behavior. Yet “intelligence” is the wrong word for what a machine does in a very narrow context.” (Pickering, Ruth, et al. Artificial Intelligence In Libraries and Publishing. E-book, Ann Arbor, MI: Michigan Publishing Services, 2022, https://doi.org/10.3998/mpub.12669942.)
The Charleston Briefings are short, open access ebooks on the topic of innovation in the world of libraries and scholarly communication. Read Artificial Intelligence in Libraries and Publishing, edited by Ruth Pickering and Matthew Ismail.