Nvidia tackles agentic AI safety and security with new NeMo Guardrails NIMs

Pivot 5: 5 stories. 5 minutes a day. 5 days a week.

1. Nvidia tackles agentic AI safety and security with new NeMo Guardrails NIMs

Nvidia has announced updates to its NeMo Guardrails technology, designed to address the needs of agentic AI. The guardrails concept provides policy and control for large language models (LLMs) to prevent unauthorized outputs. The new NeMo Guardrails NIMs are optimized for Nvidia's GPUs and offer three specific services for content safety, topic control, and jailbreak detection.

These guardrails aim to make deployment easier and provide more granular controls, addressing security, data privacy, and governance requirements. The new Guardrails are designed to address latency concerns and improve agentic AI performance.

Read the full story here

2. MatterGen: A new paradigm of materials design with generative AI

Materials innovation is crucial for technological advancements, such as lithium-ion battery technology, solar cells, and adsorbents. A new generative AI tool, MatterGen, aims to generate novel materials based on design requirements, allowing for efficient exploration beyond known materials.

MatterGen uses a diffusion model to adjust positions, elements, and periodic lattice from random structures, generating proposed structures with desired properties. The base model achieves state-of-the-art performance in generating novel, stable, diverse materials, trained on 608,000 stable materials from the Materials Project and Alexandria databases.

Read the full story here

3. Transforming learning in Nigeria, one prompt at a time

A pilot program using generative AI in Nigeria's Edo Boys High School has shown positive effects on learning outcomes. Students who participated in the program significantly outperformed their peers in English language, AI knowledge, and digital skills.

The program also benefited all students, including girls, and showed that deeper engagement led to bigger gains. The learning gains were striking, equivalent to nearly two years of typical learning in just six weeks. This is the first study to assess the impact of generative AI as a virtual tutor in such settings.

Read the full story here

4. iGenius expects to complete $1 billion data centre project with Nvidia in the summer

Italian AI startup iGenius plans to complete a $1 billion data centre project in southern Italy using Nvidia technology, requiring a five-year investment spread over five years. The project will house 80 of Nvidia's most powerful servers, each with 72 Blackwell chips.

The new supercomputer will be able to execute "115 billion of billions" of calculations per second, with the latest generation of chips allowing for 35 times more computing power but 25 times less energy. iGenius, founded in 2016, is one of the few AI startups in Europe valued at more than $1 billion and competes with France's Mistral and Germany's DeepL.

Read the full story here

5. Joint speech and text machine translation for up to 100 languages

The Babel Fish is a tool for translating speech between languages, requiring advanced technology and linguistic expertise. SEAMLESSM4T, a unified system, supports speech-to-speech, speech-to-text, text-to-speech, text-to-text, and automatic speech recognition.

Built using a new corpus of automatically aligned speech translations, it outperforms existing systems, achieving 8% and 23% higher BLEU scores in speech-to-text and speech-to-speech tasks, respectively. The system is also 50% more resilient against background noise and speaker variations.

Read the full story here