One of the first technologies unveiled is the latest artificial intelligence chip, which can allegedly perform some tasks 30 times faster than the previous one.

In addition to the B200 “Blackwell” chip, Jensen Huang detailed a new set of software tools.

The CEO also introduced the Blackwell platform. It is a GPU architecture for working with generative artificial intelligence. According to the company, the platform can perform “real-time generative AI on large language models with trillions of parameters at up to 25 times lower cost and power consumption compared to its predecessor Hopper.”

“Generative AI is the defining technology of our time. Blackwell is the engine that will drive this new industrial revolution. Working with the world’s most dynamic companies, we will realize the promise of AI for every industry,” commented Jensen Huang.

Blackwell consists of six technologies: a 208 billion transistor chip, a second-generation transformer, a fifth-generation NLink, a RAS engine, a decompression engine, and support for new encryption protocols on local interfaces.

According to NVIDIA, Blackwell will enable breakthroughs in areas such as data processing, engineering simulation, electronic design automation, computer-aided drug design, quantum computing and generative AI.

A number of companies are expected to adopt Blackwell, including AWS, Dell, Google, Meta, Microsoft, OpenAI, Oracle, Tesla and xAI.

NVIDIA also introduced a new supercomputer, the DGX SuperPOD, which is powered by Blackwell chips. According to the company, each of these Superchips provides a 30x performance increase for LLM inputs compared to NVIDIA H100 Tensor Core GPUs.

“The new DGX SuperPOD combines NVIDIA’s latest advancements in accelerated computing, networking and software to enable every company, industry and country to refine and generate their own AI,” says the CEO.

Also, SuperPOD offers predictive management capabilities that help in reducing downtime and inefficiencies while identifying problem areas. This allows it to suggest maintenance steps, adjust computing resource usage, and optimize and resume tasks.

Jensen Huang also shared a new series of chips to create humanoid robots, inviting several of the robots to join him on stage.

The company has an 80% market share and hopes to consolidate its dominant position. It is the third largest company in the US, behind Microsoft and Apple. Its shares have risen 240% in the past year and its market value reached $2 billion last month.

Three writers – Brian Keene, Abdi Nazemian and Stewart O’Nan share that parts of their works have been implemented in 196,640 books. All the data from these books was used to train NeMo to recreate written language.

After some time, the database of NeMo training books was removed due to reports that it was affected and violated copyrights.

The authors filed a class-action lawsuit against the organization in federal court in San Francisco on Friday. They report that NVIDIA’s removal of the book database makes it clear to them that the company acknowledges that using these resources infringes their copyright.

The writers are looking to see if there are other authors in the US whose copyrights have been infringed and affected in training NeMo to accurately recreate language models over the past 3 years.

In their class-action lawsuit, the authors state that the company uses the following works: Keene’s 2008 novel “Ghost Walk”, Nazemian’s 2019 novel “Like a Love Story”, and O’Nan’s 2007 novella – “Last Night at the Lobster”.

NVIDIA refused to comment on the case on Sunday. This case puts her in the middle of litigation brought by various writers who want their works protected from being used by an AI that creates new content based on their work.

The model demonstrates that improved learning methods can enhance the reasoning abilities of smaller language models to make them equivalent to larger models.

“Orca 2’s success lies in its application of diverse reasoning techniques and the identification of optimal solutions for various tasks. While it has several limitations, including limitations inherited from its base models and common to other language models, Orca 2’s potential for future advancements is evident, especially in improved reasoning, specialization, control, and safety of smaller models. The use of carefully filtered synthetic data for post-training emerges as a key strategy in these improvements,” the Microsoft team wrote in the previously mentioned blog post.

A Microsoft blog post says that compared to models of similar size, including the original Orca, Orca 2 significantly outperforms them and achieves performance levels similar to or better than those of models that are 5-10 times larger.

“Our findings underscore the value of smaller models in scenarios where efficiency and capability need to be balanced. As larger models continue to excel, our work with Orca 2 marks a significant step in diversifying the applications and deployment options of language models”, the Microsoft team added in a blog post.

Available in two sizes (7 billion and 13 billion parameters), both sizes of Orca 2 have undergone fine-tuning using specialized, high-quality synthetic data sourced from LLAMA 2 base models. Microsoft has made the weights of Orca 2 publicly accessible, aiming to promote additional research in the advancement, assessment, and alignment of smaller language models.

Elaborate instructions and numerous queries were employed to elicit responses from the teacher model, enabling the student model to grasp the underlying strategies and reasoning capabilities even in the absence of explicit task instructions. The objective is to enhance the performance of smaller models by customizing solution strategies according to the specific task at hand.

Following the news that Microsoft is making a multi-billion dollar investment in OpenAI, it is integrating the company’s technology into more of its products and services. Microsoft announced that Teams Premium is now widely available. The service includes major language models powered by OpenAI’s GPT-3.5, as well as other technologies aimed at making meetings “smarter, more personalized and secure,” Microsoft says. This Engadget wrote on the subject.

With more than 400 new features and enhancements added to Microsoft Teams last year and many more to come in 2023, all Teams users can count on continued innovation to deepen connections and foster collaboration. Some customers are looking for Teams to do more – to offer more advanced meeting capabilities to boost productivity and help consolidate their software investments to reduce costs. Today, many organizations not only pay for meeting solutions, but also purchase expensive add-on products for webinars, virtual meetings, meeting intelligence, and more.

With intelligent summarization, time spent reviewing meeting recordings can now be saved. The AI-generated parts divide the meeting into sections, so it is easy to select a specific piece of content. Meeting recordings are stored in PowerPoint Live. Intelligent summarization will automatically generate meeting chapters and based on the meeting transcript.

Soon, these custom meeting highlights will expand to include speaker timeline markers that will show who spoke during the meeting, when they spoke, and allow you to jump to that point. Speaker timeline markers are intelligently organized by who the user talks to most often, so it will never again be possible to miss a call coming from your colleagues.

Earlier, the company introduced the Azure OpenAI service for developers, as well as a tool that helps beginners create their own apps, and a graphic design app that is powered by OpenAI technology. Microsoft is rumored to be embedding ChatGPT, OpenAI’s astoundingly popular chatbot, into Bing (Google is also reportedly working on an AI-powered search chatbot).