Eye on AI - October 7th, 2022
Welcome to Aigora's "Eye on AI" series, where we round up exciting news at the intersection of consumer science and artificial intelligence!
This week, we’ll be looking at newly developed algorithms that could significantly improve chaos predictability and robot maneuverability in the wilderness.
New Algorithms Better Train Models and Make Chaos More Predictable
Traditionally, chaos has been nearly impossible to predict with any consistent degree of accuracy. The problem is twofold: 1) collecting the massive amount of data needed to make such predictions, and 2) the computing power needed to identify telling patterns within those massive amounts of data sets. Now, according to the article “New Approach to Machine Learning Could Make Chaos More Predictable,” the computational problem may have been solved with the creation of a new algorithm.
“Developed through next-gen reservoir computing techniques… the new algorithm improves predictions of complex physical processes such as the global weather forecast,” writes ScienceAlert contributor David Nield. “Calculations of these processes – known as spatiotemporal chaotic systems – can now be done in a fraction of the time, with greater accuracy, using fewer computational resources, and based on less training data.”
Reservoir computing itself is nothing new in machine learning. It uses a reservoir of data randomly connected by artificial neurons, similar to how the brain identifies patterns amid the constant barrage of data we receive in real life. Up until recently, reservoir computing calculations required massive computational power to run. Through advances in reservoir computing, as well as new innovations in machine learning, researchers have found a new approach to using reservoir computing that allows algorithms to spot potential symmetries using less computational power in what is otherwise a chaotic mess of information.
“… researchers tested their new approach on an atmospheric weather model,” continues Nield. “Using a normal laptop running Windows software, they were able to make predictions in a fraction of a second that previously needed a supercomputer. In this particular case, the calculations were made 240,000 times faster than with traditional algorithms.”
Using this same technique with new algorithms, researchers believe fast, accurate predictions are possible in virtually all fields. In the future, these new and improved algorithms could be used in a wide variety of situations, such as identifying mining possibilities, predicting health issues, significant weather events, or even socio-economic outcomes well in advance. For another story on how algorithms are advancing learning capabilities, check out this article on how AI models can continually learn from new data on the internet.
Robots Improve Maneuverability in the Wild
Chaos prediction isn’t the only field where new algorithm advances are driving machine learning to new heights. According to the article “New algorithms help four-legged robots run in the wild,” they’re also advancing the field of robotics by improving maneuverability over unpredictable terrain.
“A new system of algorithms enables four-legged robots to walk and run on challenging terrain while avoiding both static and moving obstacles,” writes UC San Diego’s editorial team.
Using a dual system of vision and feel – cameras along with foot sensors – researchers at UC San Diego were able to train robots to maneuver safely through real-world settings, allowing their robot to maneuver autonomously and swiftly across wilderness surfaces, such as sand and grass, while avoiding obstacles along the way.
"Right now, we can train a robot to do simple motions like walking, running and avoiding obstacles,” writes the study’s senior author, Xiaolong Wang. “Our next goals are to enable a robot to walk up and down stairs, walk on stones, change directions and jump over obstacles."
In the future, robots using Wang’s system could be trained for search and rescue missions, field maintenance, or even indoor activities. More algorithms, more possibilities. Watch this space.
That's it for now. If you'd like to receive email updates from Aigora, click on the button below to join our email list. Thanks for stopping by!