Meta Strikes A New Note 🎸

Plus Search is getting smarter

Thumbnail showing the Logo and a Screenshot of Meta Strikes A New Note 🎸

Meta is a beat-maker now.

Meta is making some noise about, well, making noise. The social media behemoth has announced AudioCraft, a set of AI tools that the company promises will be able to produce “high-quality, realistic audio and music.” The tool is not aimed at the Ed Sheerans and Bad Bunnys of this world, but creators who want something even fresher than the latest Taylor Swift or Travis Scott track for their Instagram reel.

The first two AudioCraft tools are MusicGen, which can be used to generate music via text prompts. AudioGen is about letting creators create sound effects using text inputs. Meta says MusicGen was trained on proprietary and licensed content, while AudioGen was trained on public sounds (think sirens, barking dogs, and so on). The last part of the AudioCraft lineup is a fresh build of its EnCodec compression tool, which utilizes AI to compress and rebuild shared audio tracks. As the company explains on its AI blog, this newest EnCode model should help to reduce audible artifacts. Meta is making AudioCraft open source, noting researchers and creators will be free to train the models using their own datasets.

This latest move by Meta follows the company’s recent rollout of its LLaMA 2 large-language model, which it also said was open source. However, many in the open source community have taken umbrage with that claim, noting LLaMA 2 is not covered by recognized open source licenses. As Wired notes, Meta requires companies with over 700 million users to agree to a _separate _license to use LLaMA 2, which suggests billionaire Zuck’s generous nature definitely has its limits.

Why it matters:

While there are many interesting things happening in the AI text-to-speech space, Meta’s push into music and ambient noise creation are certainly interesting - especially now those tools are being made available to AI experts and enthusiasts.

Google Search wants to show you things.

Google introduced its Search Generative Experience (SGE) as a beta tool a few months back, and now the company is adding even more tools to its arsenal.

For the uninitiated, SGE is an AI feature Google is testing within its Search product. Essentially, it delivers an AI-generated response ahead of that familiar cascade of blue links. Now Google is adding even more to SGE, rolling in videos and photos - and perhaps giving users a reason to not click away.

Google CEO Sundar Pichai has been vocal about how he sees AI playing a key role in Search going forward, noting on the company’s most recent earnings call he believes SGE “will just be how Search works” eventually. But considering the Google Cemetery is a thing, don’t be too surprised if Pichai’s prediction doesn’t fully pan out.

Why it matters:

Roughly 60% of Google’s income comes via its Search page, so any changes to the product will always garner a lot of attention. So while AI-powered results could be great for users who just want quick answers, those who rely on Search to drive traffic to their websites may wonder if there’s any upside for them.

Profile Picture of Tom Wilton

Written By: Tom Wilton

Lead Newsletter Writer

Published Date: Aug 04, 2023

Discovering Meta's AudioCraft, an AI-powered tool for creating unique music as it aims to revolutionize audio content creation.

By clicking “Accept”, you agree AllThingsAI can store cookies on your device and disclose information in accordance with our Cookie Policy.