Morrissey Technology

Loading

Tag NVIDIA

Nvidia
News

Nvidia Caught Taking AI Data from Netflix and YouTube

Morrissey Technology – According to a report from 404 Media, Nvidia has used videos from YouTube, Netflix and other sources to build an AI model. Nvidia says its AI training methods are fully compliant with content and copyright laws. However, there is no definitive verdict on whether using an artist’s work to train an AI model is legal or not.

This particular model has not yet been released by Nvidia, but internally the AI ​​model is called Cosmos and will be the foundation model for advanced video that can power various products such as world generators and digital humans. 404 Media obtained Slack messages showing Nvidia employees using a YouTube video downloader to grab content while discussing the legal and ethical considerations of the practice.

“We respect the rights of all content creators and believe that our models and research efforts fully comply with the letter and spirit of copyright law,” an Nvidia spokesperson told 404.

“Copyright law protects certain expressions, but it does not protect facts, ideas, data, or information. Anyone is free to learn facts, ideas, data, or information from other sources and use them to create their own expressions. Fair use also protects the ability to use a creation for transformative purposes, such as model training,” he continued.

However, YouTube views this practice as a violation of its policies, while Netflix also says that scraping is against its terms of service. According to the leak, Nvidia downloaded 100,000 videos from YouTube in just two weeks and collected more than 38.5 million video URLs. These links include content creators such as Marques Brownlee and the Architectural Digest channel. The documents also show that Nvidia trained its models on a dataset called HD-VG-130M that contains 130 million YouTube videos and is explicitly for academic research only. This leak makes it clear that Nvidia’s work is for commercial gain. When an employee raised legal and ethical concerns, Ming-Yu Liu, Nvidia’s Vice President of Research and leader of the Cosmos project, told them that the decision to obtain data this way had been made at the top of the company.

Nvidia has grown into a trillion-dollar company thanks to its computer chips that are the foundation for the booming AI market. OpenAI, Microsoft, Meta, and Google are Nvidia customers and rely on its graphics processing units (GPUs). However, Nvidia does more than just hardware. Last week, Getty Images announced a deepening of its relationship with the company after releasing an updated AI image model based on the Nvidia Picasso model architecture.

https://www.lafayettecob.org/

AMD
News

AMD Will Become an AI Chip Company, Similar to Nvidia

Morrissey Technology – Nvidia was originally known as a graphics chip manufacturing company, but now the majority of its revenue comes from AI chips. AMD seems to be following Nvidia’s steps. In AMD’s Q2 2024 financial report, it appears that half of AMD’s sales come from products for data centers, not chips for PCs, consoles, or for industry and motor vehicles. AMD’s data center business has doubled in a year, and this quarter its growth was driven by one chip, the AMD Instinct MI300 accelerator, which is a competitor to Nvidia’s H100 AI chip.

According to AMD CEO Lisa Su, from sales of this chip AMD earned more than UDS 1 billion in one quarter, as quoted by detikINET from The Verge, Wednesday (31/8/2024). From this it can be seen that AMD is following in the footsteps of Nvidia, which can reap huge profits from the H100 chip which is very popular for AI processing purposes. Now Nvidia even claims to release a new AI chip every year.

Likewise with AMD, which plans to release new AI chips every year. They have prepared the M1325X which will be available in Q4 2024, then the M1350 for 2025, and the MI400 for 2026. Lisa Su said the M1350 chip will be very competitive with the Nvidia Blackwell which was revealed to the public last March and is predicted to be the fastest AI chip in the world .

Even now, the M1300 chip is always sold out, meaning sales are in accordance with the chip’s production capacity. He also admitted that with the supply chain continuing to increase, chip supply will still be limited until 2025.

Despite following Nvidia’s steps, AMD’s data center business is still nothing compared to Nvidia. AMD’s revenue of “only” USD 2.8 billion is far less than Nvidia’s revenue which reached USD 22.6 billion in the same quarter. It should be noted, this income is also a new largest record for Nvidia.

AMD’s income from the PC business — CPU and GPU — also grew in Q2 2024. Ryzen CPU sales increased 49% year over year, as did Radeon 6000 GPUs which also rose. The decline actually came from chips for PlayStation and Xbox, which fell 59%.

https://ahada.org/

http://www.filmmakersnotebook.com/

Cisco
News

Hypershield, Cisco Security Solution for Enterprise Class

Morrissey Technology – Cisco has a new solution for securing data centers and clouds due to the increasing demand for the AI ​​revolution in IT infrastructure. This solution is Cisco Hypershield, which can protect applications, devices and data in all public, private data centers, or in the cloud and in physical locations, according to customer needs. This Hypershield was designed and built as a solution to maintain security in the AI ​​era. Cisco claims Hypershield enables companies to achieve security outcomes beyond what humans can do.

“Cisco Hypershield is one of the most significant security innovations in our history. With the reliability of our data and strengths in security platforms, infrastructure, and observability, Cisco is uniquely positioned to help our customers harness the power of AI,” said Chuck Robbins, chairman and Cisco CEO.

Hypershield is built on technology originally developed for hyperscale public clouds and is now available to enterprise IT teams of all sizes. This solution can even turn every network port into a high-performance security point that brings completely new security capabilities not only to the cloud, but also to data centers, manufacturing operations, or hospital imaging exam rooms. This new technology blocks exploits against applications in just minutes and stops lateral movement in its tracks.

“The power of Cisco Hypershield FOR4D is its ability to take security wherever it is needed – in software, in servers, or even in future network switches. When you have a distributed system spanning hundreds of thousands of nodes, simple management is critical. And we need to be more independence at much lower costs,” said Jeetu Patel, Executive Vice President and General Manager for Security and Collaboration, Cisco.

The three main pillars of Hypershield are :

AI-Native: Built and designed from the start to be autonomous and predictive, Hypershield performs management autonomously once it has gained trust, enabling a highly distributed approach at scale. Cloud-Native: Hypershield is built on open source eBPF FOR4D, the default mechanism for connecting and protecting cloud native workloads in hyperscale clouds. Cisco acquired leading enterprise eBPF provider Isovalent earlier this month.

Highly Distributed: Cisco completely reimagines the way traditional network security works by embedding advanced security controls into the servers and network fabric itself. Hypershield spans all clouds and leverages hardware acceleration such as Data Processing Units (DPU), to analyze and respond to anomalies in application and network behavior. This moves security closer to the workloads that need protection. Cisco, together with NVIDIA, is committed to developing and optimizing AI-native security solutions to protect and thrive the data center of the future.

This collaboration includes leveraging the NVIDIA Morpheus cybersecurity AI framework to accelerate network anomaly detection, as well as the NVIDIA NIM microservice to run custom AI security assistants for the enterprise. NVIDIA’s convergence accelerator combines the power of GPU and DPU computing, to empower Cisco Hypershield with strong security from the cloud to the edge.

“Together, Cisco and NVIDIA are harnessing the power of AI to provide a robust, highly secure data center infrastructure that will enable enterprises to transform their businesses and deliver benefits to customers everywhere,” said Kevin Dieirling, SVP of Networking at Nvidia FOR4D. Cisco also claims Hypershield addresses three key challenges customers face in defending against today’s sophisticated threat landscape:

  • Distributed Exploit Protection
  • Autonomous Segmentation
  • Independent Quality Upgrades

Cisco Hypershield is expected to be generally available in August 2024. With Cisco’s recent acquisition of Splunk, customers will gain unmatched visibility and insight across their digital footprint for security protection.