Connect with us

Technology

Generative AI brings robots one step closer to being able to do many things

blank

Published

on

blank

It makes sense that most of the news about humanoid robotics has been about designing the hardware. Because the people who are making them use the phrase “general purpose humanoids” a lot, the first part should get more attention. It will be a big change to go from single-use systems to more general ones after decades of them. We’re still not there.

Researchers have been working hard to make a robot with intelligence that can fully use the wide range of movements made possible by humanoid design that is based on two legs. A lot of people are also very interested in how generative AI can be used in robotics these days. According to new research from MIT, the second could have a huge impact on the first.

Training is one of the most difficult parts of making general-purpose systems work. We know exactly what works best when teaching people how to do different jobs. Robotics approaches are scattered, but some of them look good. There are many promising approaches, such as imitation learning and reinforcement learning. However, the best solutions for the future will probably be a mix of these approaches, with generative AI models added on top of them.

The MIT team says that one of the best uses would be to be able to combine useful data from these small, task-specific datasets. Policy composition (PoCo) is the name of the method. As part of a task, the robot has to do useful things like drive a nail home and flip things over with a spatula.

Researcher’s teach a different diffusion model—a plan, or policy—for finishing a single task using a single dataset, the school says. “Next, they put together the rules that the diffusion models taught them into a single rule that lets the robot do many things in different places.”

According to MIT, adding diffusion models made tasks 20% easier to do. That includes being able to do things that need more than one tool and learning how to do things that are new to them. It is possible for the system to put together useful data from various datasets into a sequence of steps needed to complete a task.

The lead author of the paper, Lirui Wang, says, “One of the good things about this approach is that we can combine policies to get the best of both worlds.” It’s possible for a policy trained on real-world data to be more dexterous, while a policy trained on simulation data could be more general.

It is the purpose of this work to create smart systems that let robots switch between tools to do different jobs. A lot of systems that can be used for more than one thing would bring the industry one step closer to the general-purpose dream.

As Editor here at GeekReply, I'm a big fan of all things Geeky. Most of my contributions to the site are technology related, but I'm also a big fan of video games. My genres of choice include RPGs, MMOs, Grand Strategy, and Simulation. If I'm not chasing after the latest gear on my MMO of choice, I'm here at GeekReply reporting on the latest in Geek culture.

Technology

Hold on tight! Tiny “scooters” are being driven by microscopic algae

blank

Published

on

blank

Biological microorganisms use little energy, multiply quickly, and move on their own. Because of this, they are perfect for powering biohybrid machines. Scientists have now made the first micromotors that run on algae. You thought animals driving cars were cool? Wait until you see this.

This is a green alga with only one cell. Its name is Chlamydomonas reinhardtii. They can swim 100 micrometers per second thanks to their two long flagella, which are the tails of cells. That’s pretty fast for something that’s only 10 micrometers long. This made student Naoto Shimizu and his co-authors want to try to use their skills more effectively by putting a harness on them.

Two tiny machines that Haruka Oda and her coworkers made will use C. reinhardtii as their power source. The machines have tiny baskets (10 micrometers across, or 1/100th of a millimeter across!) that catch the algae. The cell can still move its flagella because the basket holds it in place. The baskets are connected to each other, which makes several people work together without meaning to. This is needed to make enough thrust to move a mechanism bigger than a single cell.

blank

One is a “scooter,” and the other is a “rotator.” Four algae are in baskets at the ends of four arms in the rotator. The middle algae is pinned down. The algae could spin the rotator at 20 to 40 micrometers per second when they were swimming quickly.

 

GIF of microscopic green algae "driving" a rotating micromachine with four baskets around the outside

The harder part has been making a biohybrid machine that can move in a straight line. Ratchets were used to line up the motion in earlier designs that used different biological microorganisms, but this made it hard for the machines to move. That scooter just showed up.

It looks more like a Star Wars podracer. Two algae that face the same direction can move the small vehicle forward, if possible in a straight line. When the algae pushed the scooter to twist, turn, and tumble, it did even stranger things. This is likely because it’s not attached to anything (like the rotator) and the two drivers aren’t putting equal amounts of force on it.

black and white microscope gif of spherical algae cells "driving" a micromachine that consists of a box attached to two baskets forming a rough triangle shape

The two biohybrid machines made from algae are opening the door to new ways to make powerful vehicles. In a press release, senior author of the study and professor Shoji Takeuchi from the Graduate School of Information Science and Technology at the University of Tokyo said, “These methods could develop into a technology that can be used for environmental monitoring in aquatic environments and for substance transport using microorganisms, such as moving pollutants or nutrients in water.”

The study was written up in the journal Small.

Continue Reading

Physics

Light is the fastest thing that can “move” across a surface

blank

Published

on

blank

Einstein’s theory of special relativity says that it is impossible to move faster than light in a vacuum.

Things that don’t have mass have to move at the speed of light. But things that do have mass can’t get close to 299,792,458 meters per second (983,571,056 feet per second) without using up all of their energy. Physicists and sci-fi authors have tried to get around this by using concepts like the warp drive. But it’s likely that these will be illegal because of those pesky physics laws. Traveling faster than light can cause paradoxes that break the rules of the universe.

You are not in a dark room, though, because there is something in this room right now that can slow down or stop light. It is possible for shadows to go faster than light, and they can even smash through it.

You might ask things like, “What the hell are you talking about?” Imagine that you have a flashlight that is strong enough to light up some of the moon. If you quickly move your finger across the front of the flashlight, the shadow it casts can move across the moon’s surface at speeds much faster than light.

If you wave a laser across the night sky, you can get the same kind of effect. Think of a huge dome that is, say, 100 light-years across and surrounds you. When this laser hits that dome 100 years from now, the points will fly across it at speeds much faster than light.

But these two examples are just tricks.

Astrophysicist Michio Kaku told Big Think, “There is no message, no net information, and no physical object that actually moves along this image. There is only the image of the beam as it races across the night sky.”

No, the laser point isn’t really moving. What you’re seeing are photons hitting the dome and then different photons hitting a different part of the dome 100 years later after you moved your laser.

blank

 

 

 

 

 

 

 

 

The universe and physics stayed the same because nothing really moved faster than light, and no information was sent.

Continue Reading

Space Exploration

Someone in high school builds a model rocket that can land vertically, like a Falcon 9 Booster

blank

Published

on

blank

After three years of hard work, a high school student has finished a big project: making a model rocket that lands vertically.

It’s really rocket science to say that landing a rocket vertically is not easy. And SpaceX will tell you that they have blown up many a rocket stage while trying to land rocket boosters. But sometimes they do land before they blow up.

A student named Aryan Kapoor started building his own vertical lander in August 2021. At the end of May 2024, he finally hit the ground.

In a video for his YouTube channel JRD Propulsion, Kapoor said, “This rocket works differently than other model rockets of its kind.” “My rockets don’t have fins to keep them stable; instead, they use thrust vector control.” Thrust vector control lets the rocket’s engine move like a gimbal, giving the pilot control over the rocket’s path in space.

Even more impressive is the fact that software controls the rocket’s flight on its own.

“To guide the rocket, a flight computer makes all inflight decisions, such as steering the rocket and deciding when to ignite the landing motor.”

The onboard barometer gave the wrong reading of the rocket’s altitude during its first test flight in 2023. Kapoor wrote on his JRD Propulsion website, “The rocket did well in all other ways and collected useful data.” “Future flights will use only the accelerometer to measure altitude, providing much higher accuracy and precision.”

On his fifth attempt, Kapoor has landed successfully once more.

Continue Reading

Trending