ChatGPT ko: what happened, why we felt lost and what's energy got to do with it

Artificial intelligence has not only changed the Internet, but also the consumption behind the great Internet: every small question we ask an AI-based chatbot (such as ChatGPT) consumes on average 15 to 20 times more energy than the usual search made on an engine such as Google Search. And one of the most recent estimates, made by the Vrije Universiteit in Amsterdam, claims that by 2027 the entire AI industry could consume between 85 and 134 Terawatt-hours per year. While a study published in Medium found, among other things, that OpenAI's GPT-4 training used up to 62,000 megawatt hours, equivalent to the energy needs of 1,000 US households over 5-6 years.

This is the state of affairs. Whether it was excessive consumption that knocked out the world's most widely used chatbot will probably be clarified by the company in the coming hours. Or maybe not.

The blackout did, however, give a real taste of how dependence on chatbots is beginning to be a factor. And come to think of it, in a futuristic film that tries to tell the story of a new apocalypse, in the dimension of a world increasingly run by artificial intelligence, perhaps the AI blackout will be the real plot. Perhaps a global, synchronised, irreversible blackout. Of unknown origin.

OpenAI, Google DeepMind, Anthropic, Baidu, all LLM models - from the most sophisticated to embedded in home devices - crashing. And the AIs that have become an integral part of our daily lives (answering messages, planning agendas, writing code, managing sensitive data and guiding diagnostics in hospitals) are no longer able to help us.

Without them, billions of people feel lost. Social media become bubbles of panic, fake news and paranoia. Scenes from an American film, in short. Or at least one hopes.