🐉 NVIDIA's ML Software Moment
Weekly news digest curated by the industry insiders
NVIDIA is the undisputed leader of AI-first hardware. The GPU revolution catapulted NVIDIA's past incumbents such as Intel or Qualcomm both in market capitalization and AI-related technology innovation. NVIDIA has been driving innovation in AI chip design for different types of deep learning models, making it a central component of most modern ML infrastructures. However, NVIDIA AI influencing is rapidly expanding beyond hardware into providing end-to-end solutions for ML applications.
This week, NVIDIA hosted its annual GTC conference, where ML software was at the center of the agenda. NVIDIA dazzled data scientists with new ML software releases such as the RIVA speech SDK, the massively large Megatron language pretrained model, the Nsight Deep Learning Designer, and the new Data Science Workbench. Additionally, NVIDIA announced partnerships and optimizations for several data science platforms such as Domino. These announcements were accompanied by an overwhelming number of new AI-hardware offerings in areas such as computer vision, edge AI and even quantum computing. While NVIDIA, at its core, remains a hardware company, we shouldn’t underestimate its ambitions to become one of the most important ML platform stacks in the market.
🍂🍁 TheSequence Scope is our Sunday free digest. To receive high-quality educational content about the most relevant concepts, research papers and developments in the ML world every Tuesday and Thursday, please subscribe to TheSequence Edge 🍂🍁
🗓 Next week in TheSequence Edge:
Edge#141: we discuss Model Monitoring; we explore Google’s research paper about the building blocks of interpretability; we overview a few ML monitoring platforms.
Edge#142: we look into how Microsoft built Megatron-Turing NLG, one of the largest language models in history.
Now, let’s review the most important developments in the AI industry this week
🔎 ML Research
Fast and Cheap Ensembles
Google Research published a paper describing ensemble learning techniques that can help combine different ML models to arrive at a single output in a fast and efficient way →read more on Google Research blog
A Multilingual Model that Translates Better than Bilingual Models
Facebook AI Research (FAIR) unveiled a multilingual model that outperformed the best trained bilingual models in the prestigious WMT competition →read more on FAIR blog
About Private ML
Microsoft Research unveiled the Privacy-Preserving ML initiative (PPML) published a comprehensive analysis of some of the most recent research in this area →read more on Microsoft Research blog
Self-Supervised Learning for Video Analysis
Google Research published a paper detailing a self-supervised learning technique that can output predictions by watching unlabeled videos →read more on Google Research blog
🛠 Real World ML
The Walmart engineering team published some details about Mozrt, a personalized deep learning recommendation system used in their stores →read more on Walmart blog
Uber Database Migration
The Uber engineering team published a blog post detailing the architecture used to migrate financial data from DynamoDB to DocStore →read more in this blog post from the Uber engineering team
🤖 Cool AI Tech Releases
The new release of TensorFlow is here with new tools, improved stack traces and better usability →read more on TensorFlow blog
NVIDIA Data Science Workbench
NVIDIA unveiled Data Science Workbench, a complete stack for orchestrating software and hardware components of ML pipelines →read more on NVIDIA blog
RIVA and Megatron
NVIDIA also unveiled two major additions to its ML stack. RIVA is a GPU accelerated speech SDK, and Megatron is a 530B parameter language model. Both are available to NVIDIA enterprise customers →read more on NVIDIA blog
Nsight Deep Learning Designer
More about NVIDIA. The AI giant also announced the general availability of its Nsight Deep Learning Designer, a tool that streamlines the design and optimization of high-performance deep learning models →read more on NVIDIA blog
Continuing with NVIDIA, the AI giant unveiled Jetson AGX, a new small and energy-efficient supercomputer for robotics →read more on NVIDIA blog
💸 Money in AI
Insightful research from CBInsights
Specialized cloud services provider for GPU-based workloads CoreWeave partnered with GPU giant Nvidia and raised a $50 million round of investment led by Magnetar Capital. Read an interview with their CTO here.