This week - turning dead spiders into robots; DNA-based artificial neural networks; 3% of new code at Google is AI-generated; and more!
More Than A Human
Here is an interview with Tilly Lockey - a teenage girl who lost both arms as a child and now wears a pair of fashionable bionic arms. She shares her life as a child amputee, the prosthetics she had over the years, how she got her bionic arms and how they have changed her life and how people see her.
In this blog post, researchers from Google AI describe their efforts in making AI provide useful code suggestions for Google engineers. There is one sentence there I'd like to highlight: "Currently, 3% of new code (measured in characters) is now generated from accepting ML [machine learning] completion suggestions".
Massive artificial neural networks are famous for being extremely complex and it is practically impossible to understand what is going on inside this black box. But there are researchers who try to make tools to understand the inner workings of neural networks. This article describes what are explanation methods and how they work.
Researchers took a bunch of dead spiders and gave them a second life as robotic grippers. Since spider legs are basically hydraulic grippers, researchers were able to control spider's legs by injecting air into its limbs and turning them into a claw gripper.
During a chess tournament in Moscow, a robot playing against humans grabbed a finger of a seven-year-old child and broke his finger. The child did not want to wait for the robot to finish its movement and then make its move. In the confusion, the robot mistakenly grabbed the child's finger and broke it. According to the tournament organisers, the child is not traumatised and the company behind the robot are “going to have to think again”.
Microsoft launched Project AirSim - a platform running on Microsoft Azure designed to safely build, train and test autonomous aircraft through high-fidelity simulation. The plaftorm allows AI agents controlling the drones to train on each phase of flight, from takeoff to cruising to landing. It will also offer libraries of simulated 3D environments representing diverse urban and rural landscapes.
If you have ever wondered how many sensors a driverless car has, then here is an article for you. It lists all sensors installed on Waymo's taxi and there is a lot of them.
Researchers from China made an artificial neural network not in software but with DNA. Apart from being interesting, this DNA-based architecture that could inform the design of new molecular computing systems. In the future, their approach could be used to create various molecular diagnostic devices for biomedical applications.
DeepMind's Alphafold has deciphered the structure of virtually every protein known to science, paving the way for the development of new medicines or technologies to tackle global challenges such as famine or pollution.