This week - machine scientists help understand the world; DeepMind's new AI can help you clean up a virtual house; Spot fetches a dinner; protecting bioeconomy from hackers; and more!
More Than A Human
Brain-computer interfaces (BCI) have been multiple times used to restore movement or the ability to communicate for paralysed patients. The technology gets better every year, reporting breakthrough after breakthrough with more people and money entering the field. But there are still big challenges ahead, like creating implants that can last a lifetime, increasing the bandwidth or how to make these devices work without expert oversight for months or years, as this article points out.
Researchers say we’re on the cusp of “GoPro physics,” where a camera can point at an event and an algorithm can identify the underlying physics equation. All thanks to combining symbolic regression and genetic algorithms to find the equation that best describes the data set.
Two Minute Papers describes in this video the results of DeepMind's experiments in building interactive AI agents that learn to understand and interact with their environment. This new AI can learn by watching humans doing some tasks and then applying what it learned to interact with humans in a virtual house, like showing where the bathroom is, cleaning the room or starting a band.
New research led by scientists from Imperial College London suggests that computing with networks of nanoscale magnets could be a promising alternative for current computers or neuromorphic chips to deploy AI for small, low-power devices. In a paper published last week in Nature Nanotechnology, the team showed that by applying magnetic fields to an array of tiny magnetic elements, they could train the system to process complex data and provide predictions using a fraction of the power of a normal computer.
The Department of Justice and the Equal Employment Opportunity Commission (EEOC) each released a technical assistance document about disability discrimination when employers use AI and other software tools to make employment decisions. These documents explain how employers can make sure their recruitment process does not result in unlawful discrimination against people with disabilities.
Someone took Boston Dynamics' robot-dog Spot and taught it to fetch snacks from a restaurant. Just tell what you want and the robot will go get it entirely on its own. The future has finally arrived.
Researchers from NVIDIA presented a virtual robotic arm that learned what are the objects in front of it and how to manipulate them. After successfully teaching a virtual robot, they moved on to a real robot and it performs as good as its virtual counterpart.
A new partnership between the cybersecurity nonprofit Bioeconomy Information Sharing and Analysis Center and the Johns Hopkins University Applied Physics Laboratory, which works on emerging research with US government agencies, is highlighting the need for more resources to better secure biomedical, bioindustrial, and biomanufacturing entities.