Lifeboat Foundation

Training efficiency has become a significant factor for deep learning as the neural network models, and training data size grows. GPT-3 is an excellent example to show how critical training efficiency factor could be as it takes weeks of training with thousands of GPUs to demonstrate remarkable capabilities in few-shot learning.

To address this problem, the Google AI team introduce two families of neural networks for image recognition. First is EfficientNetV2, consisting of CNN (Convolutional neural networks) with a small-scale dataset for faster training efficiency such as ImageNet1k (with 1.28 million images). Second is a hybrid model called CoAtNet, which combines convolution and self-attention to achieve higher accuracy on large-scale datasets such as ImageNet21 (with 13 million images) and JFT (with billions of images). As per the research report by Google, EfficientNetV2 and CoAtNet both are 4 to 10 times faster while achieving state-of-the-art and 90.88% top-1 accuracy on the well-established ImageNet dataset.

In recently published research using the Atacama Large Millimeter/submillimeter Array (ALMA), astronomers announced that they had found the first clear detection of a circumplanetary, moon-forming disk surrounding the exoplanet PDS 70c — a first in astrophysics.

The circumplanetary disk, or CPD, has been seen in past research from other groups. However, due to the inability to tell the disk apart from its surrounding environment, the presence of the disk around PDS 70c could not be confirmed.

Until now.

Is Elon right Eric Klien?

It’s been nearly five years since Musk first announced the full self driving feature and very little progress has been made on it.

The chips at the heart of our digital devices are manufactured by a few large companies, but an open-source approach to design could end their dominance — with implications for everyone.

❤️ Check out the Gradient Dissent podcast by Weights & Biases: http://wandb.me/gd.

📝 The paper “The Relightables: Volumetric Performance Capture of Humans with Realistic Relighting” is available here:
https://augmentedperception.github.io/therelightables/

Continue reading “This AI Makes Digital Copies of Humans! 👤” »

Data collected from 18 states between March and August suggest the Pfizer-BioNTech vaccine reduces the risk of being hospitalized with COVID-19 by 91% in the first four months after receiving the second dose. Beyond 120 days, however, that vaccine efficacy drops to 77%.

Meanwhile, Moderna’s vaccine was 93% effective at reducing the short-term risk of COVID-19 hospitalization and remained 92% effective after 120 days.

Moderna’s COVID-19 vaccine does a significantly better job of preventing COVID-19 hospitalizations compared with Pfizer’s shot.

A newly discovered manipulation mechanism used by parasitic bacteria to slow down plant aging, may offer new ways to protect disease-threatened food crops.

Parasites manipulate the organisms they live off to suit their needs, sometimes in drastic ways. When under the spell of a parasite, some plants undergo such extensive changes that they are described as “zombies”. They stop reproducing and serve only as a habitat and host for the parasitic pathogens.

Until now, there’s been little understanding of how this happens on a molecular and mechanistic level.

“You may hit the tipping point when you’re 50; it may happen when you’re 80; it may never happen,” Schindler said. “But once you pass the tipping point, you’re going to accumulate high levels of amyloid that are likely to cause dementia. If we know how much amyloid someone has right now, we can calculate how long ago they hit the tipping point and estimate how much longer it will be until they are likely to develop symptoms.”

Summary: A new algorithm uses neuroimaging data of amyloid levels in the brain and takes into account a person’s age to determine when a person with genetic Alzheimer’s risk factors, and with no signs of cognitive decline, will develop the disease.

Source; WUSTL

Continue reading “Time Until Dementia Symptoms Appear Can Be Estimated via Brain Scan” »

Intel, Qualcomm try to wean auto companies from dependence on obsolete technology.