– Researchers from Carnegie Mellon University and AI startup Hugging Face collaborated to measure the carbon footprint of AI-generated images.
– Creating an AI image consumes the same amount of energy as charging a smartphone.
– Generating text requires significantly less energy compared to creating images.
– The researchers examined 13 tasks, such as summarization and text classification, to measure carbon dioxide emissions.
– The researchers emphasize the importance of transparency in understanding the environmental impacts of AI models.
– OpenAI’s chatbot, ChatGPT, has millions of daily and monthly active users, resulting in a significant energy usage.
– The article also mentions other news stories, including the confusion caused by computational photography and the debate over the definition of “indie” in the gaming industry.
– The Analogue Pocket handheld console is restocking, but cart adapters are delayed.
– Google has reportedly postponed the launch of its Gemini AI to 2024, citing difficulties with non-English queries.
Researchers from Carnegie Mellon University and AI startup Hugging Face collaborated to see the carbon footprint of all those whimsical AI images we’ve generated over the last few years. They discovered that creating an image using artificial intelligence uses the same energy as charging a smartphone. Generating text, whether a conversation with a chatbot or cleaning up an essay, predictably requires much less energy. The researchers examined 13 tasks, ranging from summarization to text classification, and measured the carbon dioxide produced per every 1,000 grams.
The researchers urge machine learning scientists and practitioners to “practice transparency regarding the nature and impacts of their models, to enable better understanding of their environmental impacts.” Take ChatGPT, OpenAI’s chatbot. It has upward of 10 million users per day and 100 million monthly active users. That’s a lot of energy.
— Mat Smith
You can get these reports delivered daily direct to your inbox. Subscribe right here!
The biggest stories you might have missed
You can replicate it quite easily. But I’m not sure why you would.
One woman, writer-comedian Tessa Coates, stood in front of a mirror, but with three different arm positions. How? Well, it’s likely she was moving at the instant the photo was taken, so the algorithm stitched the photo together from multiple images — exactly what the iPhone and other phones do to capture more information and synthesize better images. It appears the algo considered each figure a separate woman, capturing her at different points at different places in the image. But it’s no blue-or-gold-dress conundrum, is it?
Even if millions are spent.
One title nominated for Best Independent at The Game Awards this year, Dave the Diver, was produced by Nexon, one of the largest video game studios in South Korea. Fans quickly pointed out the error and reignited the debate over what “indie” means. Engadget’s Jessica Conditt lays out her thoughts on the matter.
But cart adapters are delayed again.
The Analogue Pocket multi-system portable handheld console is an indie hit and has been sold out for weeks. Analogue just announced a major restock, with consoles available to buy on December 4 at 11AM ET. So… now. The company promises these orders will arrive in time for the holidays. However, this only applies to the original black and white designs — not those .
Its GPT-4 rival was announced at I/O 2023.
Google has canceled its Gemini launch events scheduled for next week and now plans to launch its GPT-4 competitor in January, according to The Information. The company teased Gemini at I/O 2023, touting it as a foundational model with “impressive multimodal capabilities”
The new AI is intended to handle various applications, combining data types, like images and text, for more advanced tasks. However, sources told The Information that Gemini was struggling with non-English queries, prompting CEO Sundar Pichai to delay its release.
This article originally appeared on Engadget at https://www.engadget.com/the-morning-after-the-cost-of-generating-ai-images-121549213.html?src=rss
AI Eclipse TLDR:
Researchers from Carnegie Mellon University and AI startup Hugging Face conducted a study to determine the carbon footprint of generating AI images. They found that creating an image using artificial intelligence consumes the same amount of energy as charging a smartphone. In contrast, generating text requires significantly less energy. The researchers examined 13 tasks and measured the carbon dioxide produced per every 1,000 grams. They encourage machine learning scientists and practitioners to be transparent about the nature and impacts of their models to better understand their environmental effects. The study highlights the energy consumption associated with popular AI applications, such as OpenAI’s chatbot ChatGPT, which has millions of daily users.