Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Yesterday, Nvidia Research published a blog post highlighting the many developments in AI research throughout 2024. Some of these developments are typical fare for generative AI. However, some of them are more interesting than just creating content using other people’s copyrighted materials – for example, “StormCast” and “MaskedMimic”, which are compatible with more advanced weather forecasting and motion graphics (full motion reconstruction from visual motion partly) for robots, respectively.
There’s also “GluFormer,” which uses artificial intelligence to predict blood sugar levels for up to four years, though it requires previous glucose monitoring data. GluFormer also makes it easier to determine how dietary changes affect blood sugar over the long term, with studies showing high accuracy for people with conditions up to and including diabetes.
Some of the other improvements touted are a little less impressive and more ethically questionable than the rest of the field of generative AI. However, they still show the evolution of technology. ConsiStory allows for multiple prompts for AI images with a consistent theme, improving utility for those trying to make something narratively consistent with these tools.
Unified whole-body control of humanoids using NVIDIA Project GR00T’s MaskedMimic – YouTube
Meanwhile, “Edify3D” and “LATTE3D” are generative AI tools for creating easy 3D models. Current 3D modelers are not huge fans of this. However, they point out that AI remapping and UV mapping would be very useful for existing 3D art workloads without removing all the fun and/or billable work of making models yourself. There’s also “Fugatto”, a generative AI model that aims to create new audio files (including music) or modify existing files (such as removing background music).
Finally, at the end of the featured blog post, Nvidia Research summarizes several improvements and benchmark wins. These wins include Hydra-MDP, an autonomous driving framework that won the Autonomous Grand Challenge at CVPR 2024, and Nvidia Blackwell’s industry-leading performance in MLPerf Standardsand “FoundationPose,” which took first place on the BOP leaderboard for model-based pose estimation of unseen objects.
In its original blog post, Nvidia links to related research papers More detailed overviews of these developments.
