6 AI-related evolutions in Q1 2022

datics_Article_Cover_for_Website (1)

The world we see today is constantly improving with new adaptable technology – Artificial Intelligence and Deep Learning. Various sectors are implementing fresh research that has been used using A.I. Now, we will unveil the top six evolutions in the first four months of 2022. 

Deep learning is being used to predict consumers’ cursory impressions of human faces.

The work, which was published in the Proceedings of the National Academy of Sciences (PNAS), introduced a machine learning model that can anticipate arbitrary judgments that people make about individual photographs of faces with high accuracy.  

Joshua Peterson, one of the researchers, and his colleagues found that, of the past studies on face judgments, only a handful employed cutting-edge machine learning technologies to investigate the issue. The team’s efforts have resulted in the compilation of a large and thorough dataset of face-related prejudices and stereotypes. 

Artificial intelligence is assisting in the preservation of biodiversity. 

Hydrologic investigations have revealed that it also consumes a significant amount of water. It has power over the earth and modifies how the watershed works by not allowing moisture or precipitation to seep into the ground.  

As a result, erosion, and water flow onto nearshore reef habitats occur. Because it can be trained to recognize weeds in photos and expedite workflow, artificial intelligence can help discover the weed much faster by speeding up the analysis. AI has the potential to be a powerful instrument in assisting the “endangered species capital of the globe” in preserving the health of mauka to makai and conserving what is left. 

Brain scans and machine learning are used in a prognostic model to predict outcomes in TBI patients. 

The prognostic model, developed by data scientists and neurotrauma surgeons at the University of Pittsburgh School of Medicine, is the world’s first to use machine learning and automated brain scans to guide outcomes in patients with traumatic brain injuries (TBI).  

The researchers published their findings in the journal Radiology, revealing an advanced machine learning system capable of assessing brain scans and pertinent clinical data from TBI patients to predict both survival and recovery six months after the severe injury.  

The customized artificial intelligence model analyzes numerous brain scans from each patient, integrating them with a coma severity assessment and data from blood tests, vital signs, and heart function.  

By filtering out city noise, a deep-learning program might identify earthquakes. 

Stanford researchers claimed in a report published in Science Advances that they may enhance earthquake detection and monitoring in cities and other built-up regions. Algorithms designed to filter out background noise might be useful in particularly crowded and earthquake-prone places such as Tokyo.  

The team’s deep learning system, UrbanDenoiser, was trained on datasets containing 80,000 samples of urban seismic noise and 33,752 samples indicating earthquake activity gathered in Long Beach and San Jacinto, California, respectively.  

When the algorithms were applied to datasets from the Long Beach region, they detected many more earthquakes and made it easier to establish where they began. Datasets from a 2014 earthquake in La Habra, California, resulted in 4x more seismic detections in the denoised data, compared to the raw data. 

Machine learning aids in seeing into the depths of a volcano. 

Hierarchical clustering enabled the finding of crystal composition populations in multidimensional space that were previously invisible using traditional imaging approaches. Researchers may then determine plausible conditions for the development of each cluster of mineral compositions by integrating thermodynamic models of magmatic fractionation with the data.  

Because certain volcanoes erupt regularly while transitioning from explosive to effusive, this approach can aid in mapping out the volcano’s plumbing system over time and identifying what the products of both explosive and non-explosive eruptions are. 

When background noise inhibits hearing, machine learning predicts it. 

The researchers employed a deep machine learning-based voice recognition model with numerous layers to extract higher-level characteristics from raw input data. When paired with traditional amplitude-enhancing methods, the model could extract phonemes or sound units that serve as the building blocks of speech.  

The system was trained using random basic sentence recordings from ten female and ten male speakers. The speech was then masked with eight different noise signals, resulting in a simple steady noise and another person speaking over the speaker.  

The recordings were also degraded to different levels in order to simulate how they might sound to persons with varying degrees of hearing loss. The researchers then polled subjects with varying degrees of hearing. 

Contact us!

Looking to explore the solution?

Leave your information in the box so that our consultants will call you shortly or call our hotline:


You’ll also like