Generative AI’s Energy Problem Today is Foundational
Daily Feed: Quite a debatable topic. While we have seen infrastructure grow and adapt by itself as the technology advances, it seems to turn out in a different trajectory this time
https://spectrum.ieee.org/ai-energy-consumption
IEEE Spectrum
Generative AI’s Energy Problem Today is Foundational
Before AI can take over, it will need to find a new approach to energy
4 Replies
Relatable to what might happen if all the vehicles on road go electric suddenly! The power generation and grid will cease to support the massive load demand...
Maybe the technology is moving at a faster than ever pace?
That's a super interesting read. There's also not a clear (to me) way you can recoup the expended energy outside of some sort of tax.
What we need to consider is something like what's being worked on with EVs - EV discharge. Basically the grid would ask that all EVs plugged into the infrastructure allow the grid to draw power from their batteries when not in use in the event of a demand surge: https://embeddedcomputing.com/application/automotive/electric-vehicles-powertrain/success-story-how-electric-vehicles-can-give-back-to-the-grid
Embedded Computing Design
Success Story: How Electric Vehicles Can Give Back to the Grid - Em...
I used to think electric vehicles (EVs) were invented recently in response to the global outcry to reduce carbon emissions. So, you can imagine my surprise when I learned EVs have been in and out of popularity since the late 19th century but didn’t stick until the 21st century mostly because fossil fuels were just so available.
For LLMs, I'm not sure there's an analog. Can you think of one?
Probably powering them from the cars 😉
EVs giving back to the grid can be of use in managing the load demand throughout the day more so in a decentralized power network. Like, someone else charging their car from the energy given back is only an efficiency drop in the transmission.