Humans vs Machines: Intelligence, Learning, and Decision Making
Our advanced research program, Human-Machine Teaming: Applications, Issues, and Case Studies, invites students to explore the many different ways that machines can complement and scale human performance while reducing risk. Teaming is an increasingly popular subject in the halls of academia, and working together with AI and other robotic entities seems to be an inevitable future for mankind. But a pressing question on the minds of those in the know is this: how does human intelligence compare to computational intelligence?
For the latest Scholar Launch webinar, titled The Human Mind vs. The Computational Mind: A Faculty Roundtable on Intelligence, Learning, and Decision-Making, we invited three of our Faculty Advisors–all distinguished professors from top universities–to satiate our curiosity about man versus machine. For this roundtable discussion, the professors dug into four topics that exist at the boundaries of artificial intelligence, neuroscience, and more:
What makes human intelligence unique, and in what ways are machines catching up?
Artificial intelligence’s limits in replicating memory, learning, and judgment
How computational neuroscience bridges the gap between biology and technology
How high school students can study these questions via online research
The first question we asked our roundtable participants was simple, at least on the surface. What is the definition of intelligence?
How We Define Intelligence is Evolving
Dr. Bob Clark–a Professor of Psychiatry at the University of California, San Diego and a Chief Researcher for UCSD’s Neuropsychology Laboratory–was the first panelist to answer. In his worldview, human intelligence is measured as the extent and efficiency with which someone “can encode information, and have that information represented in a way that can be used flexibly in different situations, or to unite disparate concepts.” He continued, saying “I think the better you are at integrating your experiences into new things, and how efficiently that is done, is probably what I consider to be intelligence from a human perspective.”
When asked how she would define intelligence, the next panelist smiled, saying “That is a dissertation question.” Dr. Andrea Geana is a Postdoctoral Fellow of Computational Cognitive Neuroscience and an Assistant Instructor at Brown University. As an expert in the study of learning within the context of human brains and algorithms, she proposed that intelligence can’t be singularly defined. “There are a few things that I think contribute to intelligence, although I don't think they make a complete definition,” she said, aligning with Dr. Clark’s viewpoint, “and those things include the speed of being able to learn.” From Dr. Geana’s view, intelligence can be measured not only by a person’s (or machine’s) ability to process information and the speed at which this is done, but also by the extent to which existing knowledge can be used to infer information in unfamiliar situations.
In contrast, Dr. Ganesh Mani–a Distinguished Service Professor and Director of Collaborative AI at Carnegie Mellon University–offered an alternate perspective. According to him, regardless of the specific definition of human intelligence, what’s really important is that we bolster it through the use of technology. “Why do I reach for the GPS?” he asked. “Because I cannot do all the computations between Los Angeles and San Diego. Even if I had a paper map, we would [have to] pull over…So we need a human complement. That’s an indirect way of defining what intelligence would be useful to you. Some people call that co-intelligence.”
Co-intelligence. Does that imply that AI is equal to the human intellect? Not quite. On one hand, the human brain has evolved over thousands of years into a vastly complex system that researchers are still trying to understand, and artificial intelligence still hasn’t fully caught up with some our capabilities. However, AI can do things that far exceed our own skills.
Artificial Intelligence vs the Human Mind
“I'm going to get to what I think the human brain might still do a little bit better than AI. But I really want to emphasize what it doesn't do as well as AI, because I think it has important implications,” said Dr. Clark, whose research in neuropsychology and memory gives him a keen perspective on this topic. As he informed our webinar attendees, artificial intelligence has perfectly pristine and accurate memory. Many humans may claim that they do too, but we are simply no match for computational recall. Our biological brains undoubtedly possess impressive memory abilities, chief among them being our knack for retaining information across time and re-experiencing prior events. However, as humans, our memories can become re-encoded and change in subtle, often significant ways.
“I have learned to be suspicious of all my memories. I don't trust them,” Dr. Clark continued. “Innocent people have been sentenced to death by inaccurate testimony that resulted from investigators feeding crime details to witnesses. I want to emphasize, this is not done maliciously. It grows out of a failure to understand how human memory actually works.” Artificial intelligence’s computational memory, on the other hand, is never clouded by age, environmental context, emotions, external suggestion, or illnesses like dementia. It is perfect, all the time. Our organic brains, though, are leagues ahead of computers in pattern recognition. We’re old pros at detecting subtle patterns and flexibly assembling them into meaningful stories that we can use to guide future behavior. But how long until AI catches up with us there?
Further into the webinar, the roundtable began to broach the subject of how machines think and how artificial intelligence can be used to solve real-world problems. Dr. Mani, who’s spent decades studying the progression of this fascinating tech, named numerous examples of what AI is capable of. For example, generative AI can simulate personalities and, using what it knows about a person’s background, persona, and demographics, engage in persuasive rhetoric or build a profile on someone that rivals the work of the FBI. So it’s fair to say that there’s definitive overlap between computational and organic thinking.
“We are designing artificial systems that are inspired by the things we think our brains do really well,” added Dr. Geana, “because the reason we started doing these artificial systems, aside from just curiosity about what we can do, is to try to make our lives easier and solve problems.” Artificial intelligence is built by biologically intelligent beings, so in many ways, it’s a mirror of us. Its neural networks are designed to mimic our neurons, for instance. But according to Dr. Geana, we really need to turn our attention to our increasingly complex interactions with AI.
What initially began as a one-way relationship of human beings designing computational thinking to mimic our own, as we took advantage of machines’ superior processing and memory capabilities, has slowly morphed into a more symbiotic relationship wherein the machines are now impacting how our biological brains work. As humans become more dependent on AI, it’s influencing the ways our brains operate and learning more about us in the process. “So one of the things I am working on now is trying to understand this bilateral connection,” explained Dr. Geana.
This roundtable discussion covered so many in-depth areas, it’s impossible to capture the entire conversation in one article. You’ll just have to see for yourself what webinar attendees learned from our esteemed professors of psychiatry, AI, and neuroscience. To recap the panel discussion in its entirety, watch the replay. And to learn more about how the human mind and the computational mind compare and contrast, apply to one of the following advanced research programs this summer: