Nvidia debuts AI tool that can create music and edit voices

Nvidia has unveiled Fugatto, a cutting-edge AI tool designed to create and modify audio. Announced on Monday, this model is targeted at music producers, filmmakers, and video game developers.

Fugatto can generate sound effects and music from text prompts while transforming existing audio in highly innovative ways. For example, it can take a piano melody and turn it into a vocal line or make a trumpet sound like a barking dog.

Nvidia hasn’t announced public access to the tool, citing ethical concerns and the risk of misuse. “We need to be careful about that,” said Bryan Catanzaro, Nvidia’s vice president of applied deep learning research. The name Fugatto is derived from “Foundational Generative Audio Transformer Opus 1 (Fugatto for short).”

Generative AI, however, is not new to controversy. Companies like OpenAI are facing backlash over the ethical dilemmas these tools bring. Scarlett Johansson recently accused OpenAI of mimicking her voice without permission. Nvidia is treading carefully.

Fugatto was trained on open-source data to avoid intellectual property conflicts. Still, the risks of misuse are real. AI-generated misinformation, copyright infringements, and other potential abuses make companies like Nvidia hesitant to rush these tools to market.

While other AI developers, including Meta and OpenAI, remain vague about when their generative audio and video tools will launch, Nvidia’s cautious approach highlights the tightrope companies walk. As Catanzaro put it, “Any generative technology always carries some risks.”

Nvidia’s CEO calls for better AI answers

Nvidia CEO Jensen Huang isn’t convinced AI is ready for prime time. Speaking at the Hong Kong University of Science and Technology, Huang admitted that today’s AI models don’t deliver consistently reliable answers. He described a trust gap that needs to be closed before people fully embrace the technology.

“We’re several years away from being able to do that,” he said. Huang highlighted the issue of “hallucination,” where AI produces false or misleading results. He argued that users shouldn’t have to second-guess whether AI-generated answers are accurate or nonsense.

Huang also stressed that merely collecting data and pre-training models isn’t enough anymore. AI must evolve beyond the “just-grab-all-the-data” phase. He compared pre-training to graduating college—an important milestone but far from the finish line.

Nvidia’s solution? More computational power and advanced training techniques. The company is doubling down on innovation, which is crucial as large language models like ChatGPT, while impressive, continue to face issues like inaccuracy and ethical concerns.

OpenAI knows this too well. It was sued last year after its chatbot falsely created a legal complaint about a radio host.

China keeps Nvidia busy amid US restrictions

On the same day Fugatto was announced, Nvidia’s executive Jay Puri met with Wang Shouwen, China’s vice commerce minister, in Beijing. China rolled out the red carpet, pledging to improve its business environment for foreign firms like Nvidia.

This comes at a time when U.S.-China relations are tense, with the Biden administration preparing new restrictions on semiconductor exports to China.

China’s commerce ministry expressed a desire to repair trade relations with the U.S., describing Nvidia as a key partner in their tech strategy. The ministry’s statement also hinted at strengthening communication between the two countries.

Back in the U.S., Nvidia is watching as the Biden administration sharpens its focus on limiting China’s access to cutting-edge semiconductors. This raises the stakes for companies like Nvidia that have deep business ties in both countries.

Land a High-Paying Web3 Job in 90 Days: The Ultimate Roadmap