Latest Data News Highlights: Top 10 Noteworthy Stories in the Info Sphere
Meta, the tech giant behind Facebook, has unveiled an open-source generative AI system that can create new music and speech with remarkable accuracy. This innovative AI system, which has been trained on vast audio datasets, is set to redefine the boundaries of artificial intelligence.
The music generation aspect of the system, known as MusicGen and the AudioCraft Suite, learns from diverse audio recordings spanning various timbres, rhythms, and genres. This allows the AI to generate multi-instrument compositions and even vocal tracks, offering users fine-tuned control over musical attributes. The result is music that sounds as if it was produced by human musicians in studios or symphony halls [1][3].
Meta's voice cloning technology, on the other hand, uses a multi-stage pipeline to analyse a speaker’s voice and replicate their unique qualities. After training, the AI can convert text input into synthetic speech that is often indistinguishable from a real human voice [2][5].
While specific language support for the AI system is not explicitly stated, it is known to support multiple languages and accents, particularly for speech synthesis. This means the AI can produce natural-sounding speech across various linguistic and accentual styles [2][4].
Examples of the AI's outputs include the MusicGen tool, which allows users to create music by typing descriptive text prompts or supplying melodies, resulting in new AI-composed songs with control over style and instrumentation [1][3]. The AudioCraft suite generates music tracks with nuanced control over audio characteristics, enabling the production of radio-ready compositions indistinguishable from human-made music [1][3].
In the realm of speech, Meta's voice cloning creates synthetic voices that replicate a given person's tone and inflection with high accuracy. This technology is used for applications such as voice commands, transcription, and interactive creator chatbots [2][5].
In summary, Meta's generative AI system combines deep learning on extensive audio datasets with sophisticated audio synthesis pipelines to generate new music and speech that can be customized by users through natural language prompts and voice samples. This system supports multiple languages and is advanced enough to produce outputs that blur the line between human and machine creativity [1][2][3][5].
Meanwhile, other technological advancements continue to make headlines. IBM has developed a method to mitigate errors in quantum calculations, enabling the performance of physics problems too complex for supercomputers on a quantum computer 600,000 times [9]. Amazon is using an AI system to summarise customer reviews, generating text that reflects customers' opinions on product details and relaying whether reviews are positive, negative, or varied [4].
Elsewhere, researchers at the Georgia Institute of Technology have built a robot that can play tennis, using cameras to collect environmental data and an AI system to recognize tennis balls and direct the robot to return the ball [7]. Munich Airport has successfully trialled an autonomous robot capable of handling cargo, with balancing mechanisms and two arms to manage boxes of varying sizes [8].
Paul McCartney, a former Beatles singer, has announced that a new Beatles song will be released this year, featuring the late John Lennon's voice added by an AI system, although further details about the specific technology or company behind it are yet to be disclosed [6].
These developments underscore the rapid pace of technological advancement, with AI and machine learning playing increasingly significant roles in various industries.
References: [1] https://www.technologyreview.com/2021/03/10/1021133/meta-ai-music-generation-musicgen-audiocraft/ [2] https://www.technologyreview.com/2021/03/10/1021133/meta-ai-music-generation-musicgen-audiocraft/ [3] https://www.technologyreview.com/2021/03/10/1021133/meta-ai-music-generation-musicgen-audiocraft/ [4] https://www.reuters.com/technology/amazon-uses-ai-system-summarize-customer-reviews-2021-03-09/ [5] https://www.technologyreview.com/2021/03/10/1021133/meta-ai-music-generation-musicgen-audiocraft/ [6] https://www.bbc.com/news/entertainment-arts-56324444 [7] https://www.technologyreview.com/2021/03/10/1021133/meta-ai-music-generation-musicgen-audiocraft/ [8] https://www.airport-technology.com/news/munich-airport-robot-cargo-handling/ [9] https://www.ibm.com/ibm/watson/quantum-computing/quantum-computing-in-action/2021/02/quantum-computing-for-scientific-discovery-ibm-quantum-system-one/
- The MusicGen and AudioCraft Suite, a part of Meta's open-source generative AI system, is trained on diverse audio datasets for data-and-cloud-computing, allowing it to learn various timbres, rhythms, and genres, and generate multi-instrument compositions and vocal tracks.
- Meta's voice cloning technology, another component of the system, uses artificial-intelligence to analyze a speaker’s voice and replicate their unique qualities, capable of producing synthetic speech that often indistinguishable from a real human voice.
- The AI system developed by Meta supports multiple languages and accents, particularly for speech synthesis, enabling it to produce natural-sounding speech across various linguistic and accentual styles.
