Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: robotics/AI – Page 1,751

Watch Cassie the bipedal robot run a 5K

And it did so on its own without a tether.

Cassie, a bipedal robot that’s all legs, has successfully run five kilometers on a single charge, all without having a tether. The machine serves as the basis for Agility Robotics’ delivery robot Digit, as TechCrunch notes, though you may also remember it for “blindly” navigating a set of stairs. Oregon State University engineers were able to train Cassie in a simulator to enable it to go up and down a flight of stairs without the use of cameras or LIDAR. Now, engineers from the same team were able to train Cassie to run using a deep reinforcement learning algorithm.

According to the team, Cassie teaching itself using the technique gave it the capability to stay upright without a tether by shifting its balance while running. The robot had to learn to make infinite subtle adjustments to be able to accomplish the feat. Yesh Godse, an undergrad from the OSU Dynamic Robotics Laboratory, explained: “Deep reinforcement learning is a powerful method in AI that opens up skills like running, skipping and walking up and down stairs.”

The team first tested Cassie’s capability by having it run on turn for five kilometers, which it finished with a time of 43 minutes and 49 seconds. Cassie finished its run across the OSU campus in 53:03. It took a bit longer because it included six and a half minutes of dealing with technical issues. The robot fell once due to a computer overheating and then again after it executed a turn too quickly. But Jeremy Dao, another team member from the lab, said they were able to “reach the limits of the hardware and show what it can do.” The work the team does will help expand the understanding of legged locomotion and could help make bipedal robots become more common in the future.

NASA Is Using Artificial Intelligence To Calibrate Images Of Sun

Researchers at the National Aeronautics and Space Administration (NASA) recently announced that it is using artificial intelligence to calibrate images of the Sun.

NASA launched its Solar Dynamics Observatory (SDO) back in early 2010 to conduct research and capture high-definition images of the Sun.

The new artificial intelligence-powered technology is now helping scientists to precisely calibrate captured images at a quick pace in order to generate accurate, usable data. NASA uses the Atmospheric Imagery Assembly (AIA) present at the SDO to capture the Sun’s images across various wavelengths of ultraviolet light every 12 seconds.

Hiding malware inside AI neural networks

A trio of researchers at Cornell University has found that it is possible to hide malware code inside of AI neural networks. Zhi Wang, Chaoge Liu and Xiang Cui have posted a paper describing their experiments with injecting code into neural networks on the arXiv preprint server.

As grows ever more complex, so do attempts by criminals to break into machines running new technology for their own purposes, such as destroying data or encrypting it and demanding payment from users for its return. In this new study, the team has found a new way to infect certain kinds of computer systems running artificial intelligence applications.

AI systems do their work by processing data in ways similar to the . But such networks, the research trio found, are vulnerable to infiltration by foreign code.

This little robot is cleaning up our beaches, one cigarette butt at a time

Cigarette butts are a common type of litter for marine environments but AI-powered robot litter pickers could be the solution.

It seems many people leave behind more than just sandcastles when they go home after a trip to the beach. Beach litter is a recurring issue, and it is damaging our coastal environments and wildlife.

And there is one small item that is causing a big problem: cigarette butts. They may only be a few centimetres long, but they are full of microplastics and toxic chemicals that harm the marine environment. They don’t easily decompose, and when they come into contact with the water, harmful substances can leach out.

Unfortunately, they are also the most common type of litter, with an estimated 4.5 trillion discarded annually.

Bipedal robot developed at Oregon State makes history by learning to run, completing 5K

CORVALLIS, Ore. – Cassie the robot, invented at Oregon State University and produced by OSU spinout company Agility Robotics, has made history by traversing 5 kilometers, completing the route in just over 53 minutes.

Cassie was developed under the direction of robotics professor Jonathan Hurst with a 16-month, $1 million grant from the Advanced Research Projects Agency of the U.S. Department of Defense.

Since Cassie’s introduction in 2017, OSU students funded by the National Science Foundation have been exploring machine learning options for the robot.

DeepMind: Generally capable agents emerge from open-ended play

I’ve been suggesting for a long time to drop these Ai’s into open world games.

EDIT: Also see paper and results compilation video!

Today, we published “Open-Ended Learning Leads to Generally Capable Agents,” a preprint detailing our first steps to train an agent capable of playing many different games without needing human interaction data. … The result is an agent with the ability to succeed at a wide spectrum of tasks — from simple object-finding problems to complex games like hide and seek and capture the flag, which were not encountered during training. We find the agent exhibits general, heuristic behaviours such as experimentation, behaviours that are widely applicable to many tasks rather than specialised to an individual task.

The neural network architecture we use provides an attention mechanism over the agent’s internal recurrent state — helping guide the agent’s attention with estimates of subgoals unique to the game the agent is playing.

Robots are making progress on space exploration, along with billionaires

What i would suggest is landing Atlas robots in waves on the Moon, the first wave builds a solar panel farm for power, the second repairs the first wave, the third joins the first two to begin building large scale runways, the fourth joins the first three to begin building permanent structures.

The Moon is close enough for teleoperations, and in the 2030s, when we actually do Mars, the AI could repeat the whole thing there.

Before they explore Mars, the robots explore Martian-like caves on Earth first.