Our modern world is powered by computer chips, from the ones in cars and smartphones to those that help track animals and optimize various industries. As technology continues to advance, there is a constant push to develop faster, more efficient, and innovative chips. To achieve this, some researchers are turning to artificial intelligence (AI) to assist in the chip design process, sometimes taking a step back from human control.

A team of scientists recently shared their process of allowing AI technology to design and test more efficient computer chips. Led by Kaushik Sengupta, an electrical engineer at Princeton University, the research explores the potential of using AI to enhance chip development. Sengupta, who was recently awarded an IEEE fellowship for his work on wireless chip technology, emphasizes that the goal is not to replace human engineers, but rather to augment their productivity. By publishing their findings in the open-access, peer-reviewed journal Nature Communications, Sengupta’s team is making their AI-driven research available to the broader scientific community.

Radiologists are increasingly relying on AI-based computer vision models to assist with the time-consuming task of interpreting medical scans. However, these AI models require vast amounts of accurately labeled data to function effectively, meaning radiologists must still invest significant time annotating medical images. To address this challenge, an international team led by Johns Hopkins Bloomberg Distinguished Professor Alan Yuille has developed AbdomenAtlas, the largest abdominal CT dataset to date. With over 45,000 3D CT scans and 142 annotated anatomical structures from 145 hospitals worldwide, AbdomenAtlas is more than 36 times larger than its nearest competitor, TotalSegmentator V2. This remarkable dataset and its findings were published in Medical Image Analysis.

Historically, abdominal organ datasets were created through the labor-intensive process of having radiologists manually label each individual organ in CT scans. This process required thousands of hours of expert labor. “Annotating 45,000 CT scans with 6 million anatomical shapes would require an expert radiologist to have started working around 420 BCE—the era of Hippocrates—to complete the task by 2025,” explains lead author Zongwei Zhou, an assistant research scientist at Johns Hopkins University.

Synthesizing new proteins, which are fundamental to all biological life, is an area of immense scientific promise. A groundbreaking development from researchers in the U.S. has taken a major step forward in this field with the use of an advanced AI model called EvolutionaryScale Model 3 (ESM3). This model has been used to create a new protein, called esmGFP (green fluorescent protein), which shares just 58 percent of its material with its closest natural relative, tagRFP.

The research team estimates that this breakthrough represents the equivalent of processing 500 million years of evolution via AI, opening new doors to creating custom-made proteins designed for specific applications, or enhancing the functions of existing proteins. According to the researchers, led by Thomas Hayes, founder of EvolutionaryScale in New York, “More than three billion years of evolution have produced an image of biology encoded into the space of natural proteins.” Their study demonstrates how language models trained on evolutionary data can generate functional proteins that are significantly distant from any known natural proteins.