Nvidia listed the products it developed in the sphere of artificial intelligence (AI) in 2024 and praised itself as saying that nobody has created more groundbreaking innovation than them. That is an example of the corporate’s recent ambition to directly take part in the sphere of AI solutions beyond GPUs.
NVIDIA announced on its official blog on the thirtieth (local time), ‘Research Wealthy in 2024: Summary of AI Advances in 3D Simulation, Climate Science, and Audio Engineering‘ was posted.
Prior to the performance disclosure, he said, “The pace of technological innovation accelerated in 2024, especially dramatically in AI,” and added, “And nobody has created more groundbreaking innovation than NVIDIA Research.”
He continued, “NVIDIA Research is comprised of lots of of outstanding talents working to push the boundaries of data not only in AI but across a wide range of technology fields.”
The key research projects released last 12 months were ▲AI ‘ConsiStory’, which enables consistent image creation ▲AI ‘Edify 3D’, which creates 3D objects ▲AI ‘Fugatto’, which creates music and sounds based on text and audio prompts )’ ▲’GluFormer’, which predicts blood sugar levels 4 years in the longer term with AI ▲Text ‘LATTE3D’, one other 3D generator that converts prompts into 3D inside 1 second ▲’MaskedMimic’, a learning dataset creation tool for humanoid robots ▲’StormCast’, a model for weather prediction and climate simulation ‘, etc. were introduced.
As well as, the AI microservice ‘cuOpt’ used to enhance logistics achieved world records in 23 benchmarks, and the ‘Blackwell’ platform demonstrated world-class performance within the ML Puff benchmark.
In the sphere of autonomous vehicles, the end-to-end autonomous driving framework ‘Hydra-MDP’ took first place within the challenge of CVPR 2024, and in robotics, the model ‘Foundation Pose ( FoundationPose), which ranked first on the BOP leaderboard within the pose estimation category.
In this fashion, the research results presented by Nvidia are far beyond the spectrum of Open AI or Google, which focuses on large language models (LLM). Particularly, it has focused on ‘Physic-AI’ and has expanded its scope to incorporate robots and autonomous driving.
This move is consistent with Nvidia’s recent intention to expand beyond hardware into all areas of generative AI. Since last 12 months, the corporate has launched various generative AI models to maximise GPU utilization, and within the second half of the 12 months, it also began developing small language models (sLM) and achieved remarkable results.
Among the many AI models introduced that day, models equivalent to Latte 3D and Pukato became a hot topic with performance that surpassed other firms’ models. As well as, major events equivalent to ‘NVIDIA AI Summit India’ in October, ‘CoRL’ conference and ‘AI Summit Japan’ held in Germany in November, and ‘Siggraph Asia’ and ‘NeurIPS’ in December. We’re engaging in energetic marketing by hosting or participating in and showing off our research results.
As Microsoft, Amazon, and Google begin producing their very own AI chips, predictions proceed to emerge that NVIDIA may also move beyond GPU production to directly provide cloud services or AI model services.
And the incontrovertible fact that the last post in 2024 was concerning the results of generative AI research shows what direction NVIDIA will absorb the longer term.
Actually, NVIDIA said at the top of the post, “NVIDIA Research has lots of of scientists and engineers worldwide,” and introduced that it was recruiting talent in AI, computer graphics, computer vision, autonomous driving, and robotics.
Reporter Lim Da-jun ydj@aitimes.com
