Today in Science: An AI that thinks like a human

Science, insight, discovery April 10, 2024: What neuroscience is teaching AI developers, free-floating planetary bodies in the Orion nebula, and take a turn operating a fish doorbell.  —Andrea Gawrylewski, Chief Newsletter Editor TOP STORIES Theories of Mind Neuroscientists and developers of artificial intelligence are on a sort of shared journey. They both seek to understand how intelligence works, with neuroscientists hoping to understand consciousness, and AI researchers hoping to build truly intelligent machines. A prominent theory of consciousness inspiring AI developers is called global workspace theory, which posits that the human mind is composed of modules that operate individually; under this theory, the modules regularly come together and ultimately yield to one module, giving it precedence in the "workspace" in that moment. Why this is important: Current AIs can't compete with the human mind and regularly make mistakes or hallucinate. They have billions or trillions of parameters, enough to absorb vast swaths of the Internet, but tend to get caught up in the weeds and fail to extract the larger lessons from what they are exposed to. So researchers are hoping to model AI functionality after the human mind, which has the ability to filter vast data perceived from the world through a narrow funnel. What the experts say: As much as AI researchers are hinging their work on neuroscientific conjecture, neuroscientists hope that AI researchers can provide some clues to them about whether the global workspace theory is an accurate representation of human consciousness. "The workspace is more an idea; it's barely a theory. We're trying to make it a theory, but it's still vague—and the [AI] engineers have this remarkable talent to turn it into a working system," says Stanislas Dehaene, a neuroscientist at the Collège de France in Paris. Free-Floating Planets Last year, the James Webb Space Telescope spotted 42 pairs of Jupiter-sized planetary bodies swirling around each other in the Orion nebula. Researchers have now discovered radio-wavelength signals from one of these so-called Jupiter-mass binary objects (JuMBOs). Their radio signals suggest they are 100 times brighter than any other planet. The signals also suggest the pair isn't moving quickly through the nebula, which would mean it might have been born alone rather than blasted away from a star system, for example. Why this is interesting: The new data only adds to the mystery around these objects. Astronomers have long thought that planets can only form in orbit around stars–not arise free-floating as are the JuMBOs. Plus, no ordinary planet's magnetic field can sustain such a dazzling glow as suggested by the radio signals. The technique: If such pairs can really form without a host star, says Luis F. Rodríguez, an astronomer at the National Autonomous University of Mexico, "it means there are probably a zillion planets in our own galaxy that we haven't accounted for." Part of the Orion Nebula shown in infrared. NASA, ESA, CSA/Science leads and image processing: M. McCaughrean, S. Pearson (CC BY-SA 3.0 IGO) TODAY'S NEWS • The nonprofit Astronomers Without Borders is collecting gently used eclipse glasses through Warby Parker and other sites to enable people to see future eclipses. | 3 min read • Peter Higgs, the Nobel Prize-winning particle physicist who predicted the Higgs Boson particle, died this week. He was 94. | 3 min read • Rural Americans are dying at increasingly higher rates than city dwellers. | 5 min read • Online viewers of an underwater live-cam at a dam in a canal in Utrecht can watch for fish crowding at the dam gate and send a virtual alert to the dam operator to open it and let the fish pass through. | 3 min read Credit: Gemeente Utrecht More News EXPERT PERSPECTIVES • Forensic genealogy screens hundreds of thousands to millions of distinct variations across a genome, a person's entire genetic map, capturing connections to their near and distant relations hidden across generations. It has helped solve cold cases and bring closure to many victims' families, and yet is misunderstood by those who say it violates data privacy laws, writes Nancy La Vigne, director of the National Institute of Justice. "When traditional methods have been exhausted, forensic genetic genealogy has been invaluable in establishing familial relationships and identifying unknown individuals. The possibilities of this promising new tool should not be overshadowed by fear and misunderstanding," she says. | 4 min read More Opinion WHAT WE'RE READING • How cryptographers solved the Zodiac Killer cipher. | 404 Media • The tallest building in Taiwan has a giant steel ball at its core to stop the building from swaying during an earthquake. | The Washington Post • A review of a new book about the scientists who are unlocking the mysteries of why we sleep. | London Review of Books You might wonder, after reading about AI developers modeling AI networks after the human mind, if we are setting up these machines to achieve consciousness. In the opinion of Anil Seth, a neuroscientist at the University of Sussex in England, being conscious is much more about being alive than anything else. "Consciousness is not a matter of being smart," he says. "It's equally a matter of being alive. However smart they are, general-purpose AIs, if they're not alive, are unlikely to be conscious." For now, we don't have to fear conscious robots.   Check out the photos below of Monday's eclipse by reader John Bychowski, who snapped these stunners in West Frankfort, Illinois. In the first, he captured an incredible view of magenta solar plasma spewing off the sun's surface. Breathtaking!  Reach out any time by emailing: newsletters@sciam.com. Until tomorrow. —Andrea Gawrylewski, Chief Newsletter Editor Credit: John Bychowski Credit: John Bychowski Subscribe to this and all of our newsletters here. Scientific American One New York Plaza, New York, NY, 10004   Support our mission, subscribe to Scientific American here   Unsubscribe Email Preferences