NVIDA has surpassed the 2 terabyte-per-second memory bandwidth mark with its new GPU, the Santa Clara graphics giant announced Monday.

The top-of-the-line A100 80GB GPU is expected to be integrated in multiple GPU configurations in systems during the first half of 2021.

Earlier this year, NVIDIA unveiled the A100 featuring Ampere architecture, asserting that the GPU provided “the largest leap in performance” ever in its lineup of graphics hardware. It said AI training on the GPU could see performance boosts of 20 times the speed of its earlier generation units.

To perform tasks that involve moving or handling objects, robots should swiftly adapt their grasp and manipulation strategies based on the properties of these objects and the environment surrounding them. Most robotic hands developed so far, however, have a fixed and limiting structure; thus, they can perform a limited number of movements and can only grasp specific types of objects.

Researchers at Hong Kong University of Science and Technology have recently developed a robotic fingertip that can change its shape and switch across three different configurations, which could allow it to grasp a broader variety of objects. This fingertip’s unique design, outlined in a paper presented at this year’s IEEE International Conference on Automation Science and Engineering (CASE), is inspired by origami, the renowned Japanese art of paper folding.

“Our study was inspired by two common observations in current research and industrial applications,” Zicheng Kan and Yazhan Zhang, two of the researchers who carried out the study, told TechXplore via email. “The first relates to parallel grippers developed in past research studies, which could help to achieve industrial automation. These grippers require well-selected grasping points, otherwise static equilibrium might not be achieved.”

This article is part of our reviews of AI research papers, a series of posts that explore the latest findings in artificial intelligence.

Creating machines that have the general problem–solving capabilities of human brains has been the holy grain of artificial intelligence scientists for decades. And despite tremendous advances in various fields of computer science, artificial general intelligence still eludes researchers.

Our current AI methods either require a huge amount of data, or a very large number of hand-coded rules, and they’re only suitable for very narrow domains. AGI, on the other hand, should be able to perform multiple tasks with little data and specific instructions.

A survey of more than 800 AI systems identified four distinct types of tools.