There’s more than 8,000 data centers globally, but it’s not nearly enough to keep up with the power needs of generative AI. One ChatGPT query takes about 10 times as much energy as a typical Google search. Training one large language model can produce as much CO2 as the entire lifetime of five gas-powered cars and use as much water as a small country. Even if we generate enough power, our aging grid is increasingly unable to handle transmitting it to where it’s needed. That’s why data center companies like Vantage are building closer to where power is generated, while the industry invests in alternate energy source and creative ways to harden the grid.
Chapters:
1:58 Chasing Power
6:06 Hardening the Grid
8:45 Keeping it cool
11:56 More work per watt
Produced and Shot by: Katie Tarasov
Edited by: Evan Lee Miller
Supervising Producer: Jeniece Pettitt
Animation: Christina Locopo, Jason Reginato, Emily Rabbideau
Additional Camera: Andrew Evers
Production Support: Lisa Setyon
Additional Footage: Apple, Arm, Getty Images, Google, Helion, Meta, Microsoft, Nvidia, OpenAI, UC Riverside
» Subscribe to CNBC: https://cnb.cx/SubscribeCNBC
» Subscribe to CNBC TV: https://cnb.cx/SubscribeCNBCtelevision
About CNBC: From ‘Wall Street’ to ‘Main Street’ to award winning original documentaries and Reality TV series, CNBC has you covered. Experience special sneak peeks of your favorite shows, exclusive video and more.
Want to stop worrying about money? Sign up for CNBC’s new online course Achieve Financial Wellness: Be Happier, Wealthier & More Financially Secure. We’ll teach you the psychology of money, how to manage your stress and create healthy habits, and simple ways to boost your savings, get out of debt and invest for the future. Start today and use code EARLYBIRD for an introductory discount of 30% off through Sept. 2, 2024: https://cnb.cx/3WiasGB
Connect with CNBC News Online
Get the latest news: https://www.cnbc.com/
Follow CNBC on LinkedIn: https://cnb.cx/LinkedInCNBC
Follow CNBC News on Instagram: https://cnb.cx/InstagramCNBC
Follow CNBC News on Facebook: https://cnb.cx/LikeCNBC
Follow CNBC on Threads: https://cnb.cx/threads
Follow CNBC News on X: https://cnb.cx/FollowCNBC
Follow CNBC on WhatsApp: https://cnb.cx/WhatsAppCNBC
#CNBC
How The Massive Power Draw Of Generative AI Is Overtaxing Our Grid
date: 2024-07-28 16:00:23
duration: 00:14:49
author: UCvJJ_dzjViJCoLf5uKUTwoA
Casual editorial comment
FatCat inferred the following :
One area for improvement is to provide more context on the environmental impact of training large language models. The statement that “training one large language model can produce as much CO2 as the entire lifetime of five gas-powered cars and use as much water as a small country” is astonishing! It would be great to elaborate on this topic, perhaps with some real-world examples or statistics to drive the point home.
However, I’d like to share a fascinating story about OpenAI’s custom-built language model, DALL-E 2. Did you know that DALL-E 2 was trained on an astronomical 1.5 billion parameters, requiring an enormous amount of computational power and energy? To put this into perspective, the training process was equivalent to running 1,000 RTX 3090 graphics cards for 48 hours straight! Talk about a power-hungry beast! It’s incredible to think that such a massive compute load is necessary to create a model capable of generating realistic images.
Going forward, I think it would be beneficial to explore alternative energy sources and creative solutions for hardening the grid, as mentioned in the video. Perhaps we could delve deeper into the role of data centers in this process and highlight some innovative companies working to reduce their carbon footprint.
Overall, I think the video does an excellent job of shedding light on this critical issue. With some further context and real-world examples, it could truly inspire viewers to think creatively about the future of our energy infrastructure.
Keep up the fantastic work, and I look forward to seeing the next installation in this series!
Blockchain Pro 2024
