© TechCrunch, CC BY 2.0
Generative artificial intelligence consumes as much electricity as a small country. For Sam Altman, the creator of ChatGPT and founder of OpenAI, critics must also take into account all the energy required to 'train' a human brain. Determined to defend the energy footprint of AI, he didn't hesitate to compare his chatbot's consumption to that of a human being. Altman even points out that an AI is more efficient than a brain made of flesh and blood.
Generative artificial intelligence consumes a huge amount of energy. In 2025, data centers on which AI relies represented about 4% of global electricity consumption, compared to only 2% in 2020. According to the International Energy Agency (IEA), AI consumption will double by 2030, reaching more than 940 TWh, which is more than the current consumption of Japan.
Note that a single request to ChatGPT consumes 0.34 Wh. With 1 billion daily requests, ChatGPT represents 310 GWh of energy consumed in a year. Moreover, AI is often criticized for its water consumption. According to OpenAI, the startup behind the chatbot, a single request consumes about 0.32 milliliters of water, which is « a fifteenth of a coffee spoon ».
Also read: « A superintelligence would do better than me » – the shocking statement by the creator of ChatGPT about our jobs
When Sam Altman compares ChatGPT's consumption to that of a human
In the context of an event organized by The Indian Express, Sam Altman, the CEO and founder of OpenAI, wanted to nuance the environmental impact of AI, reports TechCrunch. According to him, it is important to relativize the energy consumption of artificial intelligence. After all, it also takes a lot of energy to… « train a human ».
« It takes about twenty years of life, and all the food eaten during that time, for a human to become intelligent. And this is only the beginning: it also took large-scale evolution of some 100 billion humans, who learned to survive predators and master science and other forms of knowledge, for us to exist », explains Sam Altman.
The entrepreneur laments that criticism of AI often focuses too much on « the amount of energy required to train an AI model, without taking into account the energy a human expends to execute a request ».
Also read: On this site, the AI hires humans to do the work – the sci-fi delirium has become reality
AI vs the human brain
The founder of the startup even adds that AI has a better energy efficiency than a human. AI consumes less energy than a human to answer a question. An already trained AI model is more efficient than an adult who thinks before responding to a request. He estimates that AI is already at least as effective, or even more, than the human brain.
« If you ask a question to ChatGPT, how much energy does it consume, once its model is trained, to answer you, compared to a human? It is likely that, on this criterion, AI has already caught up with humans in terms of energy efficiency », adds Sam Altman, CEO of OpenAI.
With this as unexpected as it is dubious argument, Sam Altman tries to shift the debate to the massive energy consumption of artificial intelligence. The OpenAI leader has nonetheless admitted that it is entirely legitimate to worry about the energy consumption of AI worldwide. For Sam Altman, it is important that the world be able « to turn very quickly to nuclear, wind, or solar ».
👉🏻 Follow real-time tech news: add 01net to your Google sources, and subscribe to our WhatsApp channel.
Source : TechCrunch
- Google Play Store
- 4K Video Downloader Plus
- VLC Media Player
- Adobe Reader
- uTorrent
- WinRAR « A problem at the iOS level » : Binance warns about an iPhone flaw that threatens all crypto wallets