In a new paper published this month, Apple researchers reveal that they have developed new methods for training large language models using both text and visual information. According to Apple’s researchers, this represents a way to obtain state-of-the-art results.
Surgically removing tumors from sensitive areas, such as the head and neck, poses significant challenges. The goal during surgery is to take out the cancerous tissue while saving as much healthy tissue as possible. This balance is crucial because leaving behind too much cancerous tissue can lead to the cancer’s return or spread. Doing a resection that has precise margins—specifically, a 5mm margin of healthy tissue—is essential but difficult. This margin, roughly the size of a pencil eraser, ensures that all cancerous cells are removed while minimizing damage. Tumors often have clear horizontal edges but unclear vertical boundaries, making depth assessment challenging despite careful pre-surgical planning. Surgeons can mark the horizontal borders but have limited ability to determine the appropriate depth for removal due to the inability to see beyond the surface. Additionally, surgeons face obstacles like fatigue and visual limitations, which can affect their performance. Now, a new robotic system has been designed to perform tumor removal from the tongue with precision levels that could match or surpass those of human surgeons.
The Autonomous System for Tumor Resection (ASTR) designed by researchers at Johns Hopkins (Baltimore, MD, USA) translates human guidance into robotic precision. This system builds upon the technology from their Smart Tissue Autonomous Robot (STAR), which previously conducted the first fully autonomous laparoscopic surgery to connect the two intestinal ends. ASTR, an advanced dual-arm, vision-guided robotic system, is specifically designed for tissue removal in contrast to STAR’s focus on tissue connection. In tests using pig tongue tissue, the team demonstrated ASTR’s ability to accurately remove a tumor and the required 5mm of surrounding healthy tissue. After focusing on tongue tumors due to their accessibility and relevance to experimental surgery, the team now plans to extend ASTR’s application to internal organs like the kidney, which are more challenging to access.
Nvidia on Monday announced a new generation of artificial intelligence chips and software for running AI models.
Google Research releases the Skin Condition Image Network (SCIN) dataset in collaboration with physicians at Stanford Med.
Designed to reflect the broad range of conditions searched for online, it’s freely available as a resource for researchers, educators, & devs → https://goo.gle/4amfMwW
#AI #medicine
Continue reading “SCIN: A new resource for representative dermatology images” »
NASA is developing autonomous space robots to build shelters, solar arrays, and more on the moon and Mars.
A new research paper has discovered the usefulness of DRAM cache for GPUs which can help enable higher performance at low power.
Researchers Propose The Use Of Dedicated DRAM Caches Onto Newly-Built SCMs For GPUs, Replacing Conventional HBM Configuration
The GPU industry, which involves consumer, workstation, and AI GPUs, is proceeding in a way that we are seeing advancements in memory capacities and bandwidth, but it isn’t sustainable, and ultimately, we could hit the limits if an innovative approach isn’t taken.
The advanced artificial intelligence model powering Copilot’s Pro tier wasn’t free to all until now.
A small research group from the University of Michigan has developed a three-legged skating/shuffling robot called SKOOTR that rolls as it walks, can move along in any direction and can even rise up to overcome obstacles.
The idea for the SKOOTR – or SKating, Omni-Oriented, Tripedal Robot – project came from assistant professor Talia Y. Moore at the University of Michigan’s Evolution and Motion of Biology and Robotics (EMBiR) Lab.
Continue reading “Omnidirectional tripedal robot scoots, shuffles and climbs” »
MIT CSAIL researchers introduce FeatUp, a model-agnostic framework designed to significantly enhance the spatial resolution of deep learning features for improved performance in computer vision tasks such as semantic segmentation, depth prediction, and object detection.
According to Nvidia’s roadmap, it’ll unveil its next-gen Blackwell architecture soon. The company always launches a new architecture with data center products first and then reveals the cut-down GeForce versions many months later, so that’s what’s expected this time as well. On that note, the company’s semi-annual GTC technology conference starts in two weeks, so we expect a lot to be revealed at the show. As proof that Nvidia is close to pulling the wraps off its new data center GPUs, a Dell executive has already shared some juicy info about next-gen Nvidia hardware, saying in a recent earnings call the company has a 1,000W data center GPU in the pipeline.
The executive who has probably already received an angry call from Jensen is Jeff Clarke, a COO at Dell. On a Feb. 29 earnings call (PDF), the executive discussed Dell’s engineering superiority and how upcoming hardware from Nvidia will give the company a chance to show it off. “We’re excited about what happens at the B100 and the B200,” he said, which are the die names for Nvidia’s next-generation data center GPU and its apparent successor. For context, Nvidia currently has the H100 as its flagship data center GPU and is just now launching the second iteration with faster HBM3e memory, dubbed H200. We all know the B100 is the Blackwell successor to this chip, so it appears the B200 will be that GPU’s second iteration, though it does not currently appear on Nvidia’s roadmap (below).
