BitEnergy AI has developed artificial intelligence that radically saves electricity consumption The boom of nuclear power plants in the United States may have to wait.
In short, it was possible to replace floating point multiplication (FPM) with integer addition. The new method, «Linear Complexity Multiplication» (L-Mul), approximates FPM results using a simpler algorithm. It still maintains high accuracy comparable to FPM results. This method can reduce the power consumption of AI systems by up to 95%, making it extremely interesting for commercial use.
Because it is a new process, existing accelerators from NVIDIA and others do not work with it — new hardware is needed for practical applications.It may require significant investment, but a 95% reduction in power consumption is likely to make the largest tech companies think twice.
Google has put its climate goals on the back burner due to AI’s consumption requirements: the company’s greenhouse gas emissions have increased by 48% since 2019 instead of decreasing as expected Google pays attention to nuclear energy and other alternatives.
If AI processing becomes energy-efficient, it will be possible to reap the benefits of AI technologies without sacrificing the environment. In addition, a 95% reduction in energy consumption will reduce the burden on national power generation systems and eliminate the need to build many new power plants.
Sources: Tom`s Hardware
Spelling error report
The following text will be sent to our editors: