Summary
Transcript
In a breakthrough of robotics speed and power, a new contender from China is turning heads with its blistering performance and precision while mimicking human movements. Meet the S one, a cutting edge household robot developed by Astrobot, a subsidiary of Stardust Intelligence. As demonstrated in real time video, the S one robot executes household tasks at a pace that seems to defy the boundaries of conventional robotics. With movements so fluid and fast that they almost blur the line between machine and human, the robots unassuming design is matched with next generation capability.
And when it comes to learning, the robot has already conquered a wide array of tasks, wielding a form of artificial intelligence called imitation, learning to do things like open a bottle of wine and pour it, or even writing chinese smoothly, all at a speed that is potentially far faster than humans. But the s one’s capabilities don’t end there. From shaving a cucumber with a kitchen knife in one seamless, unbroken motion, to flipping a sandwich in a sizzling frying pan with the dexterity of a seasoned chef, this robot seems to have mastered the art of domestic multitasking.
And perhaps most impressive is the video showcasing the s one’s remarkable calligraphy skills as a clear demonstration of the next tier of robot dexterity. This requires the use of an exact amount of speed and pressure to match the output of a master calligrapher, which the robot flawlessly recreated the stroke of a human hand. And as for the secret behind the s one robots next level agility and precision, Astrobot says the key lies in their cutting edge training technique known as imitation learning, which works by having the robot closely study and emulate the movements and actions of humans it observes.
But where the s one really has superhuman ability is in its accuracy, which maintains repeatability within just 0. 03 mm, whereas a human is over 30 times less accurate at 1 mm. More importantly, though, is the team’s ability to execute, with members having previously worked with tech giants like Tencent, Baidu and the Hong Kong Polytechnic University, as well as having also helped to develop other robots like Tencent’s Ollibot, which seems to have provided ample experience on their quest to create truly human like robot assistants.
And while Astrobot has remained tight lipped about further technical details behind the s one, the company has said this robot is the closest product to human operating performance in the market today. As for now, the s one is undergoing rigorous testing with other AI models, with the company aiming for a commercial release later in 2024 to bring this general purpose feat of engineering into homes and workplaces around the world.
Strategically their unveiling of the s one comes at a time when the robotics industry is just starting to have its chat GPT moment. For example, OpenAI just stunned the world with not one but two groundbreaking robotic creations. First, there was a laundry folding robot showcasing exceptional soft touch capabilities that could revolutionize household chores. Next, in a separate collaboration with figure, OpenAI also unveiled its advanced robot that exhibits unprecedented natural language reasoning skills, marking a significant leap forward in human machine communication.
One thing is the era of truly human like robots is dawning, and marvels like the s one are just the beginning of what promises to be an exhilarating journey into the future of artificial intelligence and robotics. Meanwhile, Sanctuary AI has once again raised the bar of humanoids. With its unveiling of the 7th generation Phoenix robot and the powerful carbon AI control system, this latest iteration represents a substantial leap forward, bringing us even closer to the tantalizing prospect of machines that can match, and perhaps even surpass human intelligence.
For years, the innovative minds at Sanctuary have been tirelessly refining their vision of a humanoid robot capable of performing a vast array of tasks, from the mundane to the complex, with the same versatility and cognitive aptitude as their fleshy creators. And with each successive generation, theyve inched ever closer to realizing that dream. And while previous generations of Phoenix focused primarily on training the robot’s upper torso to tackle duties like labeling boxes, bagging groceries, and soldering, the 7th gen model takes things to dizzying new heights.
Boasting significant advancements in both hardware and software. This cutting edge automation promises unparalleled visual perception, tactile sensing, and operational uptime, which are all critical components in the quest to develop a truly general purpose intelligence. But what truly sets the 7th gen Phoenix apart is the relentless pace of its development. Less than a year after the 6th generation’s debut and a mere 16 months since the fifth gen model’s historic commercial deployment in January 2023, Sanctuary AI has once again leapfrogged the competition with a robot that seems to defy the boundaries of what was previously thought possible.
As for the future, glimpses from Sanctuary AI suggest that the 7th gen Phoenix will boast a wider range of motion in its wrists, hands, and elbows, opening up a whole new world of potential applications. Perhaps even more exciting is the development team’s reported breakthrough in reducing the time it takes for the robot to learn new tasks and perform them autonomously, which is a feat that could revolutionize the very concept of task automation.
Equally impressive are the strides made in cost effective manufacturing, with the company claiming to have slashed the expenses associated with building materials, and by using smaller, more efficient hydraulic systems that promise lower weight and power consumption, as well as a streamlined commissioning process that allows for rapid deployment of new units, the 7th gen Phoenix seems poised to shatter preconceived notions of what’s possible in the realm of intelligent robotics.
Finally, OpenAI’s Sora faces a serious contender, as chinese tech giant Shengshuai and Tsinghua University presented their latest artificial intelligence breakthrough, named Vidoo, which is a high definition text to video model. Video can swiftly generate 16 2nd long videos at 1080p resolution with a single click, but its foundation lies in its universal vision transformer, known as Uvit for short, which is a novel architecture that integrates the capabilities of the diffusion and transformer AI models.
This unveiling follows closely on the heels of Sora, developed by us based OpenAI. Garnering global attention. Both research teams at Tsinghua University’s Institute for Artificial Intelligence and Shengshu AI emphasized the alignment of Vidoo, with their technical roadmap driving their continued research efforts. Importantly, the teams claim that their universal vision transformer technology actually predates Sora’s diffusion based architecture, which was pioneered by the team in September 2022. Moreover, experts are hailing VDoo’s architecture to be the world’s first ever visual transformer model, combining the strengths of diffusion and transformer.
During a live demonstration, Vadu showcased its ability to simulate intricate real world scenes. Adhering to physical laws with precision, it can generate dynamic shots offering flexibility beyond static imagery. Notably, vadus development in China imbues it with a deep understanding of local elements, enabling it to depict iconic chinese symbols like pandas and dragons. In unveiling vadou, Shengshu AI and Tsinghua University have underscored China’s prowess in AI innovation, positioning vadoo as a significant contender in the global AI landscape.
With its capacity to create compelling videos effortlessly, vadoo is poised to revolutionize various industries, from entertainment to marketing, offering new avenues for creative expression and communication. .