Will AI use ALL of the energy?
Will AI use ALL of the energy?
Here’s a look at the most interesting AI+Energy content from the past week
Hi Folks, there was a lot of news last week regarding the energy consumption of AI, especially with Davos capturing so much of the media cycle. Though I hadn’t intended for this newsletter to carry a weekly theme, it made sense to focus exclusively on that topic for this edition of EnergyNews.ai.
Also, I’m going to start experimenting with some additional features/segments to the newsletter, with the goal of eventually establishing a subscriber-paid and/or sponsor-worthy version of the weekly post. I’d highly value your feedback as I stumble through figuring out how all of that works.
I’m calling this week’s new feature “Jake’s Takes”, where I will provide a bit of my own (ever-evolving) view on the main ideas presented by each link. Please let me know what you think!
-JM
OpenAI CEO Altman says future AI depends on energy breakthrough
While speaking at Davos, OpenAI's CEO Sam Altman said an energy breakthrough is necessary for future artificial intelligence, which will consume vastly more power than people have expected. Anticipating this demand, Altman has invested at least $375 million in nuclear fusion.
Video of Altman’s comments can be found below. Relevant comments begin at about 23:00:
published: Jan 16, 2024 Source: Reuters, Bloomberg
Jake’s take: There’s a lot to digest here. Fusion? Really? We’re still (and, seemingly, perpetually) 20-25 years away from any meaningful commercialization of fusion tech, and it almost certainly won’t happen soon enough to address climate change. Apparently, Altman doesn’t believe that existing/known energy tech (renewables, nukes, etc.) will keep up with AI’s demand. He does suggest “radically cheaper solar plus storage or something at massive scale, like a scale that no one is really planning for.” as an alternative approach.
I found this to be a powerful statement:
“My whole model of the world is that the two important currencies of the future are compute/intelligence and energy”
Sam has a massively influential voice here. If he’s right, we’ll have a significant energy hole to fill, and I think we’ll see a big uptick in interest around nuclear power again (remember Crypto?). I’ve got to wonder- might his $375M investment have a larger, faster impact applied towards solar+storage innovations?
Why does AI need so much energy?
Armand Ruiz, Director of AI at IBM, dives into the drivers of AI’s energy consumption. He also covers:
Environmental Impact of High Energy Consumption
Case Studies: AI's Energy Usage in the Real World
Sustainable Solutions and Innovations
Industry Initiatives and Corporate Responsibility
Looking Ahead: The Future of AI and the Environment
Published: Jan 13, 2024 Source: newsletter.nocode.ai
Jake’s take: Great info from a credible source. Armand is at the bleeding edge of AI development, and the fact that his article is (in my opinion) light on actionable sustainability solutions for AI developers suggests that there’s a big opportunity here. How do we develop the tools so that AI engineers can make intelligent decisions about how their models are run? Surely not every model training job has to be run immediately and at 100%. I’d love to see the dev tools provide energy-saving options to the engineers, e.g. “Run my model optimized for energy (and cost!) savings, even if it takes 50% longer”.
Also, (and related to the above post), the suggestion that nuclear power is “sustainable” might raise some eyebrows.
Deploying high-performance, energy-efficient AI
Investments into downsized infrastructure can help enterprises reap the benefits of AI while mitigating energy consumption, says corporate VP and GM of data center platform engineering and architecture at Intel, Zane Ball.
Published: Jan 10, 2024 Source: MIT Technology Review
Jake’s take: Ball is saying good things here, like “Enterprises need to be very aware of the energy consumption of their digital technologies, how big it is, and how their decisions are affecting it" and projecting “facilities that train AI models on a large scale while modulating energy consumption based on its availability” and the “continued growth of liquid cooling”.
I appreciate the focus on the Data Centers, and the suggestion that the DCs will need to offer more tools to their energy-aware AI customers is spot-on.
AI Is Ravenous for Energy. Can It Be Satisfied?
“As long as the competition between makers of AI continues to spur companies to use these ever more capable, ever more power-hungry models, there’s no end in sight to how much more electricity the global AI industry will demand. The only question, then, is at what rate its consumption of power will increase.”
Published: Dec 15, 2023 Source: WSJ
Jake’s take: There are some strong claims in this article, like: “global power usage for AI systems could ratchet up to 15 gigawatts… requiring about 15 average-size nuclear power plants”, and “the amount of electricity required to power the world’s data centers could jump by 50% by 2027, thanks to AI alone”.
The quoted researcher, Alex de Vries, goes on to say that those estimates don’t include the cooling load, which could be equal to the compute load… so, 30 new nukes?
I think the main takeaway here is that the estimates are still a bit “finger to the wind”, and likely on the low side.
If you’re looking for a more conservative view, check out David Mytton’s blog. David is highly credible on data center energy consumption and argues that most estimates depend upon inaccurate extrapolation techniques.
In case you missed it- here are our favorite links from previous editions of EnergyNews.ai:
How can AI help with climate change?
In this episode of Volts.wtf with David Roberts, Priya Donti, executive director of nonprofit Climate Change AI, speaks to how artificial intelligence and machine learning are affecting the fight against climate change.
source: volts.wtf
Why AI and energy are the new power couple
AI is increasingly vital in managing complex, data-rich power systems, especially with the rise of renewable energy sources. This IEA.org article highlights AI's role in improving the predictability and efficiency of power supply and demand, particularly in renewable energy. It also delves into AI's contribution to predictive maintenance, ensuring more robust and reliable energy infrastructures.
source: iea.org