Jan 19, 2022
Segna Newsletter — 20 January 2022
What we’ve been reading
AI that understands speech by looking as well as hearing
Meta AI
Meta AI is working on a new conversation AI system, AV-HuBERT (Audio-Visual Hidden Unit BERT), that, like humans, can understand speech better by using both ears and eyes. AI systems currently used in applications, such as smart speakers or tools for those with hearing impairments, often struggle to work well in ‘everyday’ situations where multiple people are speaking simultaneously, or when there’s lots of background noise. By incorporating data on both visual lip movement and spoken language, AV-HuBERT will bring applications closer to human-level speech perception.
FarmSense uses sensors and machine learning to bug-proof crops
TechCrunch
Bugs and plant diseases account for 40% of agricultural production loss worldwide. FarmSense is attempting to solve this by using optical sensors and a machine learning based classification system to identify and track insects, allowing for early detection of pests and thus the timely deployment of pest-management tools such as insecticide or biocontrols.
Study finds artificial intelligence accurately detects fractures on x-rays, alert human readers
Medical Dialogues
Researchers at the Boston University School of Medicine have developed an AI algorithm that helps radiologists interpret x-rays after an injury or suspected fracture, with tests resulting in a 29% reduction in missed fractures. Fracture interpretation errors represent up to 24% of emergency departments’ harmful diagnostic errors, with error rates spiking during the evenings and nights due to non-expert reading and fatigue.
John Deere thinks its self-driving tractor can help feed the world
Morning Brew
John Deere unveiled its first fully autonomous tractor, allowing farmers to leave the cab and remotely control the machine from an app. With a rising global population and growing food production needs, combined with difficulty finding skilled agricultural workers, autonomous machinery means farmers can “focus their attention on the jobs that require more expertise from them.” John Deere aims to make planting seeds, applying fertilizers, and harvesting crops autonomous in the next decade.
Seeing the plasma edge of fusion experiments in new ways with artificial intelligence
MIT News
A key challenge in making fusion power viable is confining superheated plasma long enough for the device to produce significant amounts of net energy. Researchers at MIT are looking to replace ‘first principle’ simulations of plasma behavior with machine learning power computer models which trade accuracy with much faster run times. It is hoped that these systems will result in actionable insights into the behavior of plasma and, by doing so, impact fusion reactor designs.
If you’d like to receive this newsletter in your inbox, you can subscribe here.
