Search for: All records

Generative AI, exemplified in ChatGPT, Dall-E 2, and Stable Diffusion, are exciting new applications consuming growing quantities of computing. We study the compute, energy, and carbon impacts of generative AI inference. Using ChatGPT as an exemplar, we create a workload model and compare request direction approaches (Local, Balance, CarbonMin), assessing their power use and carbon impacts. Our workload model shows that for ChatGPT-like services, in- ference dominates emissions, in one year producing 25x the carbon-emissions of training GPT-3. The workload model characterizes user experience, and experiments show that carbon emissions-aware algorithms (CarbonMin) can both maintain user experience and reduce carbon emissions dramatically (35%). We also consider a future scenario (2035 workload and power grids), and show that CarbonMin can reduce emissions by 56%. In both cases, the key is intelligent direction of requests to locations with low-carbon power. Combined with hardware technology advances, CarbonMin can keep emissions increase to only 20% compared to 2022 levels for 55x greater workload. Finally we consider datacenter headroom to increase effectiveness of shifting. With headroom, CarbonMin reduces 2035 emissions by 71%. 

Hydrogels are extensively used as tunable, biomimetic three-dimensional cell culture matrices, but optically deep, high-resolution images are often difficult to obtain, limiting nanoscale quantification of cell–matrix interactions and outside-in signalling. Here we present photopolymerized hydrogels for expansion microscopy that enable optical clearance and tunable ×4.6–6.7 homogeneous expansion of not only monolayer cell cultures and tissue sections, but cells embedded within hydrogels. The photopolymerized hydrogels for expansion microscopy formulation relies on a rapid photoinitiated thiol/acrylate mixed-mode polymerization that is not inhibited by oxygen and decouples monomer diffusion from polymerization, which is particularly beneficial when expanding cells embedded within hydrogels. Using this technology, we visualize human mesenchymal stem cells and their interactions with nascently deposited proteins at <120 nm resolution when cultured in proteolytically degradable synthetic polyethylene glycol hydrogels. Results support the notion that focal adhesion maturation requires cellular fibronectin deposition; nuclear deformation precedes cellular spreading; and human mesenchymal stem cells display cell-surface metalloproteinases for matrix remodelling.