Andrej Karpathy, a founding member of OpenAI, has officially left the company. It marks his second departure after he spent several years at Tesla leading the development of the Autopilot software.
Neural networks RSS
With today's highlighted deal, upgrade your programming and web development skills with 38 hours of content on Python, Artificial Intelligence, Keras, PyTorch, and more via Neowin Deals!
In a white paper, researchers at Bloomberg modeled supply chain data as a graph and used GNNs to create a long-short portfolio. The results demonstrate an edge over traditional approaches.
Researchers at Google have developed BLUERT, an automatic metric that gauges the performance of natural language generation models and delivers SOTA performance on two academic benchmarks.
Using a trained deep reinforcement learning model, a team at Google calculated the optimal chip placement on an IC—a problem that would have taken several weeks for human experts to solve.
_medium.jpg)
DLSS 2.0 uses TensorCores to run the deep learning network at 2X the speed of the original. Other key features include training for multiple games, increased upscaling, and better rendering.
Researchers at UC Berkeley have created the Berkeley Autonomous Driving Ground Robot (BADGR) that uses live input to an artificial neural network to plan out obstacle-free paths to a goal.
Taking inspiration from the olfactory bulb in the human brain, researchers at Intel have implemented a neural algorithm on the Intel Loihi neuromorphic system to distinguish between smells.
Researchers have trained an artificial neural network to identify atypical and ephemeral experimental signatures resulting from proton-proton collisions inside the Large Hadron Collider at CERN.
With today's highlighted deal, take a deep dive into Machine Learning and Data Science with included certification across 8 courses and 48 hours. Save hundreds of dollars off with Neowin Deals!
In an industry first, deep transfer learning has now been used to train an artificial neural network to classify galaxies as spiral or elliptical with 99.8% accuracy based on their pictures.
Google's Meena, a chatbot, sporting a 2.6 billion parameter neural conversational model trained on 341GB of public social media data surpasses all other chatbots on the new SSA metric.
By working on only de-identified mammograms, Google's machine learning model surpassed individual experts in accurately identifying breast cancers despite working with lesser information.
With today's highlighted deal, take a deep dive into Machine Learning and Data Science with included certification across 8 courses and 48 hours. Save hundreds of dollars off with Neowin Deals!
_medium.jpg)
With a novel algorithm—Automatic Domain Randomization—the team at OpenAI was able to train a robotic hand to solve the venerable Rubik's Cube in a variety of unexpected situations.
With today's highlighted deal, take a deep dive into Machine Learning and Data Science with included certification across 8 courses and 48 hours. Plus when you use a promo code get a further 20% off!
In an industry-first, a deep learning-based convolutional neural network model has been used to provide more accurate values pertaining to dark matter when compared to traditional methods.
The 4D tracking techniques would use 4D dynamic solid modeling of a cluttered space by collecting depth data from 3D cameras. These modeling concepts can then be applied in a number of use cases.
Nvidia is calling out Intel, claiming that the giant chip maker manipulated and skewed benchmarks to favor its Xeon Phi processor, by comparing it to years-old GPUs using outdated software.
