A robot hand that can solve a Rubik’s Cube, artificial embryos grown in the lab, a “bizarre, brainless blob” that can heal itself and has a fondness for oatmeal — oh, and there’s persuasive evidence for life on Mars. In this week’s coolest scientific discoveries, the hills are alive with all sorts of weird stuff.
What is it? Programmers at San Francisco’s OpenAI trained a pair of neural networks to solve a Rubik’s Cube with a humanlike five-fingered hand.
Why does it matter? This isn’t the only bot to solve a Rubik’s Cube — see also a machine designed by a German inventor that accomplished the task in a record-setting 0.637 seconds. But what the OpenAI engineers are after is a general-purpose robot that, like the human hand, is dexterous and multifunctional, rather than bots designed with only a single task in mind: “We set this goal because we believe that successfully training such a robotic hand to do complex manipulation tasks lays the foundation for general-purpose robots.”
How does it work? The team used an old piece of hardware — their humanlike robotic hand has been available for 15 years — and paired it with new software: Using a method called automatic domain randomization, they trained the AI with “progressively more difficult environments in simulation” until it got the hang of it. Like all of us, the AI is still learning — it solves the Rubik’s Cube only about 60% of the time. But don’t forget that it’s doing the job one-handed.
Top image credit: OpenAI
What is it? A collaboration between researchers at Johns Hopkins’ School of Medicine and its Applied Physics Laboratory has enabled a patient, for the first time, to simultaneously control two prosthetic arms with his thoughts.
Why does it matter? In past, patients controlled one robotic arm with their thoughts. The new development could restore capability to patients with spinal cord injuries and neuromuscular diseases, said biomedical engineer Brock Wester, the principal investigator on the study: “We are trying to enable a person with quadriplegia to use a direct neural interface to simultaneously control two assistive devices and, at the same time, feel touch sensation when the devices make contact with objects in the environment.”
How does it work? The study is part of the program Revolutionizing Prosthetics, launched in 2006 by the Defense Advanced Research Projects Agency. The newest breakthrough builds off a previously developed device called the Modular Prosthetic Limb. Earlier this year, doctors implanted electrodes in the region of a patient’s brain that controls movement and touch, using real-time mapping of brain activity during surgery to determine the best place to hook up the wires. Dr. Matthew Fifer, of the Applied Physics Laboratory, said, “For the first time, our team has been able to show a person’s ability to ‘feel’ brain stimulation delivered to both sides of the brain at the same time. We showed how stimulation of left and right finger areas in the brain could be successfully controlled by physical touch to the MPL fingers.”
What is it? Researchers at the Salk Institute and the University of Texas Southwestern Medical Center have created a mouse blastocyst — the first 100 cells of a nascent life, which can become an embryo when implanted in the uterus — from scratch. No sperm or eggs were used in the making.
Why does it matter? The development has “profound implications,” according to the Salk Institute, for researchers working on pregnancy and infertility, and even for those seeking to understand diseases that develop later in life, like Alzheimer’s. Juan Carlos Izpisua Belmonte, from Salk’s Gene Expression Laboratory, said, “These studies will help us to better understand the very beginnings of life; how early on in life a single cell can give rise to millions of cells and how they are assembled in space and time to give rise to a fully developed organism. Importantly, this work avoids the use of natural embryos and is scalable.”
How does it work? Because it’s not a natural blastocyst, scientists are actually referring to the artificial version as a “blastocyst-like structure,” or a blastoid. Researchers took cells from adult mice and put them into a solution that encouraged them to turn into induced pluripotent stem cells — which can differentiate into any kind of cell in the body, including the three cell types found in embryos. They then put them in a culture medium that coaxed the cells to form connections with one another, where, said Salk’s Jun Wu, “they essentially did the job on their own — you could see the cells that would become the placental tissue moving to the outside while others that would form the fetus moved to the inside.” The study is described further in the journal Cell.
What is it? The news this week was awash with reports of a “bizarre, brainless blob capable of learning and eating oatmeal” — but enough about this author! In truth, this organism is a rare, strange and scientifically intriguing slime mold, on display at the Paris Zoological Park starting this week. In nature it prefers muggy environments, but the Paris researchers grew their own specimen in a petri dish.
Why does it matter? Officially named Physarum polycephalum, it’s neither plant nor (apparently) fungus nor animal — and scientists refer to it as “the blob” for short. Bruno David, director of the Paris Museum of Natural History, told Reuters, “The blob is a living being which belongs to one of nature’s mysteries.” Scientists think it’s been on earth for a billion years, though the first one wasn’t discovered till the 1970s — in a Texas backyard.
How does it work? Amazingly, though it’s not clear how: It’s a single-celled organism that crawls along the forest floor in search of bacteria and fungal spores to digest, despite not having a mouth or stomach. The Paris zoo fed its specimen oatmeal, “which it seemed to like,” according to the Smithsonian. It can heal itself in minutes if cut in half, and has more than 700 discrete sexes. Also per the Smithsonian: “It possesses a kind of intelligence — though it has no brain.” The organism can find the shortest way through a maze in search of food, and the trail of slime it leaves in its wake functions as a kind of “externalized spatial ‘memory,’” helping it avoid areas it’s already been.
What is it? Gilbert V. Levin, the principal investigator in a 1970s NASA experiment to determine whether there was life in Mars, writes in a new article that the evidence led him to “conclude” that the answer is yes.
Why does it matter? As NASA plans to return to Mars, it’s vital to know whether life there already exists, Levin writes: not little green or red men, but microbial life such as what he and his colleagues looked for decades ago. “Our nation has now committed to sending astronauts to Mars,” he writes. “Any life there might threaten them, and us upon their return.”
How does it work? Levin worked on the Labeled Release (LR) life detection experiment, part of NASA’s Viking program, which returned data that “signaled the detection of microbial respiration on the Red Planet,” Levin writes. ” The curves from Mars were similar to those produced by LR tests of soils on Earth. It seemed we had answered that ultimate question.” But when the Viking experiment failed to find evidence of “organic matter,” NASA decided that the LR findings were a “substance mimicking life,” but not the real thing — and shifted its focus to the separate question of whether Mars could be hospitable to humans. Levin is proposing that the agency pick up where he and his colleagues left off and “put life detection experiments on the next Mars mission possible.”