Technology
Generative AI brings robots one step closer to being able to do many things
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.
Artificial Intelligence
Google DeepMind Shows Off A Robot That Plays Table Tennis At A Fun “Solidly Amateur” Level
Have you ever wanted to play table tennis but didn’t have anyone to play with? We have a big scientific discovery for you! Google DeepMind just showed off a robot that could give you a run for your money in a game. But don’t think you’d be beaten badly—the engineers say their robot plays at a “solidly amateur” level.
From scary faces to robo-snails that work together to Atlas, who is now retired and happy, it seems like we’re always just one step away from another amazing robotics achievement. But people can still do a lot of things that robots haven’t come close to.
In terms of speed and performance in physical tasks, engineers are still trying to make machines that can be like humans. With the creation of their table-tennis-playing robot, a team at DeepMind has taken a step toward that goal.
What the team says in their new preprint, which hasn’t been published yet in a peer-reviewed journal, is that competitive matches are often incredibly dynamic, with complicated movements, quick eye-hand coordination, and high-level strategies that change based on the opponent’s strengths and weaknesses. Pure strategy games like chess, which robots are already good at (though with… mixed results), don’t have these features. Games like table tennis do.
People who play games spend years practicing to get better. The DeepMind team wanted to make a robot that could really compete with a human opponent and make the game fun for both of them. They say that their robot is the first to reach these goals.
They came up with a library of “low-level skills” and a “high-level controller” that picks the best skill for each situation. As the team explained in their announcement of their new idea, the skill library has a number of different table tennis techniques, such as forehand and backhand serves. The controller uses descriptions of these skills along with information about how the game is going and its opponent’s skill level to choose the best skill that it can physically do.
The robot began with some information about people. It was then taught through simulations that helped it learn new skills through reinforcement learning. It continued to learn and change by playing against people. Watch the video below to see for yourself what happened.
“It’s really cool to see the robot play against players of all skill levels and styles.” Our goal was for the robot to be at an intermediate level when we started. “It really did that, all of our hard work paid off,” said Barney J. Reed, a professional table tennis coach who helped with the project. “I think the robot was even better than I thought it would be.”
The team held competitions where the robot competed against 29 people whose skills ranged from beginner to advanced+. The matches were played according to normal rules, with one important exception: the robot could not physically serve the ball.
The robot won every game it played against beginners, but it lost every game it played against advanced and advanced+ players. It won 55% of the time against opponents at an intermediate level, which led the team to believe it had reached an intermediate level of human skill.
The important thing is that all of the opponents, no matter how good they were, thought the matches were “fun” and “engaging.” They even had fun taking advantage of the robot’s flaws. The more skilled players thought that this kind of system could be better than a ball thrower as a way to train.
There probably won’t be a robot team in the Olympics any time soon, but it could be used as a training tool. Who knows what will happen in the future?
The preprint has been put on arXiv.
Astronomy
Witness the rare celestial event of Mars and Jupiter reaching their closest proximity in the sky this week, a phenomenon that will not occur again until 2033.
Mars and Jupiter will be only 0.3 degrees apart in the sky on August 14. From our point of view, this passage is very close. If you miss it, you won’t be able to see another one until 2033.
When two objects pass each other in the sky from our point of view, this is called a conjunction. Every time two planets came together, the closer one would block out the other because they would all be moving in a perfectly flat plane. The orbits of the planets are slightly different from those of the other planets, though, so they move slightly to the north and south of each other. Every time, that gap is a different size.
When two things happen close together, the results are especially stunning. Jupiter and Saturn were close enough to each other in 2020 that they could be seen in the same field of view through a telescope. This is a treat for people who like to observe the sky.
Being 0.5 degrees wide, the full moon will fit in any view that can hold the whole moon. This pair will also look good before and after the full moon.
But even with the naked eye, a close conjunction can make the sky look even more amazing. The contrast between the red of Mars and the white of Jupiter will be especially striking. However, Mars’ brightness changes a lot. When it’s at its brightest, it’s about the same brightness as Jupiter. Right now, it’s 16 times less bright. They are so bright that, unless there are clouds, you should be able to see them from all but the dirtiest cities.
Most people in the world will miss this sight, though, because they can’t see the pair of planets in the evening from anywhere on Earth. The exact time they rise depends on where you live, but it’s usually between midnight and 3 am. To see this, you will mostly need to get up before astronomical twilight starts so that you have time to get through the thickest part of the atmosphere.
For people in Europe, Africa, west Asia, and the Americas, the closest time will be 14:53 UTC, which is during the day. The mornings before and after, though, will look almost as close.
Mars and Jupiter meet about every two and a half years, but the most recent one was almost twice as far away and could only be seen in the morning. In 2029, the gaps will be just under two degrees. The next one will be even wider, at more than a degree.
When planets are close to each other, that doesn’t always mean that their distance from each other is very small. Mars has been around the Sun for 687 days, but it is now less than 100 days past its perihelion, which means it is closer than usual. Even though Jupiter is a little closer than usual, it’s not really that close. To be as close as possible to each other, Mars has to be at its farthest point, and Jupiter has to be at its closest point. So this one is not unusual.
But if you want to see something beautiful, you will have to wait more than nine years to see it again.
Engineering
New concrete that doesn’t need cement could cut carbon emissions in the construction industry
Even though concrete is a very common building material, it is not at all the most environmentally friendly choice. Because of this, scientists and engineers have been looking for alternatives that are better for the environment. They may have found one: concrete that doesn’t need cement.
Cement production, which is a crucial ingredient in concrete, ranks as the third most significant contributor to human-caused carbon emissions globally. Nevertheless, in recent years, a multitude of alternative techniques for producing more environmentally friendly concrete have surfaced. One proposed method involves utilizing industrial waste and steel slag as CO2-reducing additives in the concrete mixture. Another suggestion is to utilize spent coffee grounds to enhance the strength of the concrete while reducing the amount of sand required.
However, a certain company has devised a technique to produce cement-free concrete suitable for commercial enterprises.
The concrete has the potential to have a net reduction in carbon dioxide and has the ability to prevent approximately 1 metric ton of carbon emissions for every metric ton used. If this statement is accurate, the cement-free binder will serve as a noteworthy substitute for Portland cement. According to BGR, the new concrete also complies with all the industry standards of traditional cement concrete, ensuring that there is no compromise in terms of strength and durability.
While it is still in the early stages, the situation seems encouraging. C-Crete Technologies, a company specializing in materials science and holding the patents for a novel form of concrete, has utilized approximately 140 tons of this new cast-in-place (pourable) concrete in recent construction endeavors.
In September 2023, the company was granted an initial sum of almost $1 million, promptly succeeded by an additional $2 million, by the US Department of Energy to advance the progress of its technology. In addition, it has garnered numerous accolades that are facilitating its growth in operations.
The widespread adoption of cement-free concrete in future construction projects has the potential to significantly alter the environmental impact of the industry. Although C-Crete seems to be one of the few companies currently exploring these new alternatives on a large scale, it is likely that others will also start embracing them in the near future.
- Gadgets9 years ago
Why the Nexus 7 is still a good tablet in 2015
- Mobile Devices9 years ago
Samsung Galaxy Note 4 vs Galaxy Note 5: is there room for improvement?
- Editorials9 years ago
Samsung Galaxy Note 4 – How bad updates prevent people from enjoying their phones
- Mobile Devices9 years ago
Nexus 5 2015 and Android M born to be together
- Gaming9 years ago
New Teaser For Five Nights At Freddy’s 4
- Mobile Devices9 years ago
Google not releasing Android M to Nexus 7
- Gadgets10 years ago
Moto G Android 5.0.2 Lollipop still has a memory leak bug
- Mobile Devices9 years ago
Nexus 7 2015: Huawei and Google changing the game