5 Noteworthy AI and Deep Tech Articles: week of July 17, 2022
1) Doctors using AI catch breast cancer more often than either does alone (MIT Technology Review)
A doctor and AI working together could do better than either alone. AI software and a radiologist were 3.6% better at detecting breast cancer together than either screening for the cancer alone, according to a study published in The Lancet Digital Health. This is the first large-scale study to directly compare an AI’s performance in breast cancer screening according to whether it is used alone or to assist a human expert. The hope is that such AI systems could save lives by detecting cancers doctors miss, free up radiologists to see more patients, and ease the burden in places where there is a dire lack of specialists. The researchers suggest that the software could help detect more cancer cases and cut down on radiologists’ workloads.
Scientists led by Australia’s Curtin University used AI to help identify a Mars crater that ejected a 2.1-billion-year-old meteorite that crashed in the Sahara desert. The famous meteorite, nicknamed Black Beauty, was found in Africa in 2011. It’s unique in that it is the only found meteorite from Mars that is breccia, a type of sedimentary rock made of angular, broken rock fragments cemented together. Before the research, scientists did not know the meteorite’s origin site. To determine this, they developed a machine learning algorithm to analyze the millions of craters on the Martian surface visible in high-resolution photos. The scientists found and listed around 90 million craters using the AI software and a supercomputer. Further research enabled them to narrow down the list of candidates, eventually singling out one crater with the same properties as Black Beauty.
3) AI learns what an infant knows about the physical world (Scientific American)
To build AI systems capable of broadly generalizing across different use cases and experiences, researchers are looking to children for inspiration. DeepMind researchers have released an “intuitive physics” model to try to capture the knowledge a baby is born with in an AI system. We previously explored this topic in our Radical Talks podcast with renowned psychologist Alison Gopnik. She discussed how AI systems may benefit from a better understanding of the way children learn and play. Like scientists, children constantly test hypotheses to better understand the world. Gopnik argues that the causal inference demonstrated by a child offers clues into how to build more resilient AI systems.
4) AI Champions Awards winners revealed (AIMed)
The winners of this year’s prestigious AI Champions Awards have been revealed as part of AIMed’s Global Summit in San Francisco. The Temerty Centre for Artificial Intelligence Research and Education in Medicine (T-CAIREM) at the University of Toronto received the Hospital/Institution of the Year award for 2022. Muhammad Mamdani, T-CAIREM Director and Co-founder of Radical Ventures portfolio company Signal1, accepted the award. As the first centre of its kind in Canada, T-CAIREM champions the development and application of AI in medicine both nationally and internationally through world-class initiatives in education, research, and enabling data environments. They join previous winners, including Eric Topol, Jeremy Howard, and Stanford’s Center for Artificial Intelligence in Medicine and Imaging.
The Internet loves generating art with AI. Half of the fun is reading the wacky prompts created by the human users. The best results are often from individuals who understand how these models work with the prompts. Models such as DALL-E or Make-A-Scene, at first believed to be the end of human artists, are now applauded as tools to elevate human creativity by artists themselves. Some users, such as digital artist Noah Bradley, believe the impact of software such as DALL-E will be “similar to the effect of smartphones on photography—making visual creativity more accessible without replacing professionals.” Creating powerful, usable images still requires a lot of careful tweaking after something is first generated.