Green IT : A 360-Degree Scan of Current Research , Projects and Initiatives
Technology Machine-to-Machine ( M2M )
Wireless Sensor Network ( WSN )
RFID
Short Description Direct communication capability between wired and wirelessly connected IoT devices without human intervention . Network of sensor nodes . Uses scheduling , interference reduction , resource allocation and routing to be energy efficient and acts as a gateway to a local area network .
Electronic tag that can answer an identification request from a reader . Microcontroller Unit Low cost processor unit . Energy Harvesting
Energy generation for batteries of e . g . IoT network sensors by mechanical , thermal , radiative or biochemical sources . High-frequency signals have advantages over low-frequency signals , as they can produce energy and process information at the same time .
Table 3-3 : Green IoT technologies .
According to Figure 3-2 , the most important keywords in the field of Green AI are " machine learning ," " cloud computing " and " energy efficiency ," but technical aspects such as " approximate computing " also appear . In addition to data centers , the energy consumption of training and using artificial intelligence should not be underestimated . According to forecasts , deep learning processes will consume almost 300,000 times more computing power in 2019 than in 2013 8 .
In general , the energy consumption of AI should not be neglected . A single large speech transformer model can generate nearly 300 tons of CO2 during training from scratch . To put this in perspective , an average human is responsible for just under 5 tons of CO2 over the course of a year 9 . However , as methods for greening AI itself currently need more research , greening other processes with the help of AI , Green by AI , can have a direct impact on the combat against climate change .
The potential for savings through artificial intelligence , i . e . " green by AI ," exists and relates to energy consumption , resource conservation and CO2 emissions . According to a study by the
8
R . Schwartz , J . Dodge , N . A . Smith und O . Etzioni , “ Green AI ,” in Communications of the ACM , pp . 54-63 , December 2020 .
9
E . Strubell , A . Ganesh , A . McCallum , “ Energy and Policy Considerations for Deep Learning in NLP ,” 2019 . 14
April 2023