Digital Transformation (DX) with AI

Green AI + Intumit

Green AI

Intumit continues to tackle Green AI to promote low-carbon business practices for brighter business sustainability by enabling cloud computing, reducing deep learning computational time while increasing output, and reducing inference cost and large carbon footprints for NLP models in efforts for environmental sustainability.

Taking Carbon Challenges

Intumit is working together with top academic researchers to seek methods to reduce training data, training time, and training models by improving the computational efficiency with many pretrained subnetworks of different sizes. 

To train a deep-learning model at a large scale requires about 626,000 pounds of carbon dioxide, which is equivalent to lifetime emissions of five cars. NLP aims to decipher and analyze human language, with applications like predictive text generation that are working under complex systems of neural networks. Studies show in some cases, part of that network can be repurposed for another, a subnetwork picked for one task could be repurposed for another, allowing powerful modules to use less computing power.

As we seek forward on energy efficient technologies to control various amounts of data, parameters, and model factors; Intumit will continue to invest in innovation and technologies that will decarbonize future operations to meet carbon neutrality.

Reduce Carbon Footprint

Taking carbon challenges, Intumit is here to support our partners and enterprises to minimize environmental impact while maximizing benefits of sustainability by reducing carbon footprints. Such initiatives will take place to accelerate carbon reduction and removal opportunities, globally and domestically, especially for our partners in Japan and ASEAN territories.

By 2030, we believe cloud-enabled technologies and sustainable practices used in enterprises will grow in an aim towards carbon negative.

Intumit plans to publish a SmartBERT Green Edition in the second half of 2021 by reducing the complexity of neural networks’ computation time for carbon footprint reduction.