In Today’s Science News, we learn about the improvement of AI learning by adding variability as it already exists in the neurons of our brain; a new wireless wearable that can be 3D-printed and is useful for gathering health data; the impact of the Australian wildfire on the natives bees; and lastly, a bonus article about space to celebrate the 10th Science News entry.
Article 1: Brain cell differences could be key to learning in humans and AI
SD-Date: 6th October, 2021
Et-Date: 9th October, 2021
ScienceDaily Summary: „Researchers have found that variability between brain cells might speed up learning and improve the performance of the brain and future AI.“
The brain is made up of billions of neurons which are connected by a vast neural network.
They allow us to learn about our environment and are also quite efficient when it comes to energy consumption. Bioelectric processes are used for signal processing (which are subdivided into gradual and all-or-none processes) which in turn are created by differing distributions of ions inside and outside of the neural cell. As it is explained by the NIH:
„Neurons are information messengers. They use electrical impulses and chemical signals to transmit information between different areas of the brain, and between the brain and the rest of the nervous system. Everything we think and feel and do would be impossible without the work of neurons and their support cells, the glial cells called astrocytes (4) and oligodendrocytes (6).“
Now to the study and why I’ve told you about the function of neurons in the first place.
The authors explain that neuron cells are like snowflakes – they look identical from afar, but on further inspections it becomes clear that no two are exactly alike.
That’s in contrast to artifical neural networks where each cell is exactly identical. Only their connectivity is varying. While their speed continues to advance, these artificial neural networks do not learn as accurately or quickly as the human brain (for instance: an AI has a harder time to recognize a fruit under various circumstances, whereas humans learn it through a few samples -> that’s also why AIs are highly specialized). Said variability is suspected to be the culprit of the slower and inefficient learning.
Method of Research
For the study, the researchers decided to focus on tweaking the „time constant“. „[…] that is, how quickly each cell decides what it wants to do based on what the cells connected to it are doing“. There are two types of cells: those which act quickly by only looking at what the connected cells have just done; and those which act slowly and base their decision on what cells have been doing for a while.
Then, after adding the varying time constants to the artificial cells, the network was tasked with performing some benchmark machine learning tasks: „to classify images of clothing and handwritten digits; to recognise human gestures; and to identify spoken digits and commands.“
Thanks to the ability of processing information similar to the human brain in a slow and quick manner, the artificial network was better able to solve tasks in more complicated, real-world settings.
Furthermore, the closer the variability of the network was to the brain’s variability, the better they performed. Thus suggesting that the brain has evolved „just the right amount of variability for optimal learning“.
Lastly, a statement from the researcher Nicolas:
„We demonstrated that AI can be brought closer to how our brains work by emulating certain brain properties. However, current AI systems are far from achieving the level of energy efficiency that we find in biological systems.
„Next, we will look at how to reduce the energy consumption of these networks to get AI networks closer to performing as efficiently as the brain.“
Article 2: Engineers 3D-print personalized, wireless wearables
that never need a charge
SD-Date: 8th October, 2021
Et-Date: 10th October, 2021
ScienceDaily Summary: „Engineers have developed a new type of wearable with several unprecedented benefits. Not only are the devices custom 3D-printed based on body scans of wearers, but they can operate continuously using a combination of wireless power transfer and compact energy storage.“
The devices that are currently available to monitor the body face a variety of limitations.
Smartwatches, for instance, have to be charged and can only gather a limited amount of data due to their placement on the body (the wrist).
Other wearable sensors that stick to the skin come off when the skin goes through its normal shedding process, or sometimes when a subject sweats.
Then there are the sophisiticated wearables used in clinical settings. While they are able to gather a lot more data, they face also issues such as not being wireless which in turn means limited mobility. It prevents the patient from going about their normal daily routines due to the bulkiness of devices such as the ECG monitor.
Solution of the Engineers
Gutruf and his team developed devices that can be 3D-printed and then wrapped around the body where needed. They are custom-fitted to the wearer and biosymbiotic.
It receives power from a wireless system with a range of several meters. A small energy storage unit ensures its function even if the wearer goes out of range or outside the house.
Neither do they need any adhesives to stick to the body.
„These devices are designed to require no interaction with the wearer,“ Gutruf said. „It’s as simple as putting the device on. Then you forget about it, and it does its job.“
A video from the the University of Arizona, where the device is explained, was also linked:
The device’s ability to monitor various parameters was tested, among the parameters were temperature and strain while a person jumped, walked on a treadmill and used a rowing machine. In the rowing machine, multiple devices were worn to track exercise intensity and the way muscles deformed with fine detail. Its accuracy was high enough to measure body temperature changes induced by walking up a single flight of stairs.
Article 3: Wildfire bees on the brink
SD-Date: 4th October, 2021
Et-Date: 10th October, 2021
ScienceDaily Summary: „The number of threatened Australian native bee species is expected to increase by nearly five times after the devastating Black Summer bushfires in 2019-20, new research has found. With 24 million hectares of Australia’s land area burnt, researchers say the casualties are clear among bee fauna and other insects and invertebrates after studying 553 species (about one-third of Australia’s known bee species) to assess the long-term environmental damage from the natural disaster.“
The wildfires in Australia in 2019/2020 destroyed approximately 3,000 homes and killed or displaced an estimated 3 billion (3,000,000,000) animals. 24 million hectars of land were burned.
As the ScienceDaily summary already pointed out, about one-third of Australia’s known bee species were subject to the study to assess the long-term environmental damage from the natural disaster.
Method of Research
The study at hand was a collaborative study and included researchers from Flinders University’s Laboratory of Evolutionary Genetics and Sociality, the South Australian Museum, University of Adelaide, Curtin University, University of Sydney, University of Melbourne, Murdoch University and Charles Darwin University.
553 species (ca. 1/3rd of Australia’s known bee species) were studied to determine the damage that was caused by the wildfire. It also revealed the casualties among other insects and invertebrates (animals with neither a backbone nor a bony skeleton).
- 11 Australian bees (2% of those analyzed) are recommended to be set as priority taxa for listing as IUCN Threatened species (9 were assessed as vulnerable and 2 as endangered)
- The study also warns about widespread wildfire and forest fire damage around the world which have a catastrophic impact on biodiversity
- Additionally, the study serves as a foundation to assess the likely impact of natural disasters on poorly studied species
- Preservation of the native bees in Australia is also important, because of their very important role in pollination and the still unknown role of most others
„Our research is a call for action, from governments and policymakers, to immediately help these and other native populations most in danger,“ says lead author Flinders University PhD candidate James Dorey […].
Mars‘ surface shaped by fast and furious floods from overflowing craters
SD-Date: 29th September, 2021
Et-Date: 10th October, 2021
ScienceDaily Summary: „On Earth, river erosion is usually a slow-going process. But on Mars, massive floods from overflowing crater lakes had an outsized role in shaping the Martian surface, carving deep chasms and moving vast amounts of sediment, according to a new study.“
(Note: You can interact with Mars on the NASA website, for everyone who’s interested)
Method of Research
By using remote sensing images that were taken by satellites orbiting Mars, scientists were able to study the remains of breached Martian crater lakes. Before, – according to lead author Tim Goudge – the crater lakes and river valleys had mostly been studied on an individual basis. In this study, they investigated how the 262 breached lakes across the planet shaped the Martian surface as a whole. Moreover, the research entailed a review of a preexisting catalog of river valleys on Mars and classified the valleys into two categories:
„[V]alleys that got their start at a crater’s edge, which indicates they formed during a lake breach flood, and valleys that formed elsewhere on the landscape, which suggests a more gradual formation over time.“
With this data, they compared the depth, length and volume of the different valley types.
Following Researchers were involved: lead author Tim Goudge, an assistant professor at the UT Jackson School of Geosciences; study co-author Alexander Morgan, a research scientist at the Planetary Science Institute; as well as other co-authors: Jackson School postdoctoral researcher; Gaia Stucky de Quay and Caleb Fassett, a planetary scientist at the NASA Marshall Space Flight Center.
- Nearly 1/4th of Mars‘ river valley volume was created by crater lake breaches (the furious floods mentioned in the title) despite making up only 3% of the total valley length
- The median depth of a river valley created by a crater flood is 559 feet (170.5 meters)
- Those created gradually over time have a median depth of about 254 feet (77.5 meters)
- Moreover, the unique topography of Martian river valley may also be the result of these strong crater floods
From the findings of the study can be concluded that lake breach river valleys played an important role in shaping the Martian surface. Goudge says that it is a lesson in expectations, as well: „The Earth’s geology has wiped away most craters and makes river erosion a slow and steady process in most cases. But that doesn’t mean it will work that way on other worlds.“