April 25, 2024

Valley Post

Read Latest News on Sports, Business, Entertainment, Blogs and Opinions from leading columnists.

NVIDIA has released an open source tool for building safer AI models – Nvidia

NVIDIA has released an open source tool for building safer AI models – Nvidia

Since March, NVIDIA Offers AI Foundations, a service that allows companies that don’t have a lot of money or time to develop their own large language models (LLMs) from scratch “to build, improve, and run custom language models and AI models that can be trained using their proprietary data.”

at recent days , The company announced NeMo Guardrails, a tool designed to help developers define a framework for protection, suitability, and security for the AI ​​applications they create. NeMo Guardrails allows software engineers to impose three different types of limits on the custom LLMs they create. More specifically, companies can put in place “native firewalls” that prevent apps from “handling” topics for which they are not trained. For example, NVIDIA has proposed a chatbot customer service that, with the help of software, refuses to answer a weather question. Companies can also set protection and security limits, which are designed to ensure that LLM pulls accurate information and connects to applications known to be safe.

According to NVIDIA, NeMo Guardrails works with all major language models, including ChatGPT. In addition, the company claims that almost any software developer can use its software because “it is not necessary to have specialized knowledge of machine learning or a data scientist.” Since NeMo Guardrails is open source, NVIDIA points out that it can work alongside all the tools enterprise application developers are already using.

NVIDIA has integrated NeMo Guardrails into the existing NeMo framework for building AI models. Its enterprise customers can access NeMo through the company’s AI Enterprise software platform, which also provides the framework, and through its AI foundations service. The rollout of NeMo Guardrails comes shortly after some high-profile AI solutions, including Microsoft Bing and Google Bard, came under scrutiny from regulators for their tendency to introduce “hallucinogensInfo (don’t forget that Google’s chatbot, Bard, made a huge mistake in the company’s first public demo).

See also  Pokémon's Diamond and Pearl Remake has been updated to version 1.2.0, here are the full patch notes

“The product of many years of research, NVIDIA has made NeMo Guardrails open source to contribute to the tremendous energy and work of the AI ​​security developer community,” said NVIDIA. “Together, our Guardrail efforts will help companies keep their smart services aligned with protection, privacy, and security requirements so that their innovation engines stay on track.” If you’d like to read more about how NeMo Guardrails work, you can visit the company’s website at Her official blog.



  • haha

    2