How much energy does AI consume? Google only shared Gemini numbers

explosion Artificial intelligence tools All over the world IncreasedHowever The companies that make this Tools often do not express their environmental effect in detail.

Google only has She released a technical paper Detail of energy, emissions and water use of it twin The development of artificial intelligence. The effect of one wave is, he says, minimal. According to her methodology To measure the effect of artificial intelligenceThe energy consumption to demand one about what is equivalent to watching the TV for less than 9 seconds.

This is in one meal, except when you think about a variety of used Chatbots, with billions of claims that are easily sent every day.

On the most positive aspect of progress, technology behind these claims has become more efficient. During the past 12 months, one energy twin The text router has been reduced by 33X, and the total carbon fingerprint has been reduced by 44X, says Google. According to the technology giant, this is not baseless, and it is a momentum that will need to keep moving forward.

Google did not immediately respond to CNET’s request for more comments.

Google calculates is much more than that

The typical calculation of the energy cost to demand AI in the active device that has been operated, which shows a much smaller imprint for everything. But Google’s method of measuring a waved effect claims to extend over a wide range of factors that draws a clearer image, including the dynamic energy of the full system, inactivity machines, general data expenses, water consumption and more.

For comparison, it is estimated that only active TPU and GPU, one Gemini uses 0.10 watts of power, 0.12 milliliters of water and emit 0.02 grams of carbon dioxide equivalent. This is a promising number, but the broader Google methodology tells a different story. With more considerations in place, Gemini text is used 0.24Wh of energy, 0.26 ml of water and emit 0.03 GCO2E – about twice the plate.

Will new competencies continue to use artificial intelligence?

Through a multi -layer series of efficiency, Google constantly works on ways to make the artificial intelligence effect less burden, than a more efficient model structure and data centers to custom devices.

With the most intelligent models and emerging cases and tools, these competencies will be very important because we immerse ourselves deeper in AI.

For more, you must Stop using ChatGPT for these things.