• The Mahazine
  • Posts
  • Meta brings GenAI to metaverse, French AI lab released Moshi, Meta AI image label backlash, TIME partners with Elevenlabs

Meta brings GenAI to metaverse, French AI lab released Moshi, Meta AI image label backlash, TIME partners with Elevenlabs

Artlist funny commercials, Motorola AI Fashion show, AI Music video clip, Magnific Photoshop plugin

Welcome to the mahazine newsletter #12

Trending AI stories, ads & marketing campaigns 👇

📣 In the news

Meta is planning to integrate generative AI technology into its VR, AR, and mixed reality games, with a primary focus on its Horizon platform. The company aims to create games that change dynamically with each playthrough and follow "non-deterministic" paths. Additionally, Meta is looking to develop or partner with third parties to create AI-powered tools that could streamline game development processes and improve time-to-market.

This initiative is part of Meta's broader strategy to reinvigorate its metaverse efforts, which have faced challenges including significant financial losses in its Reality Labs division. The company is actively seeking talent to research and prototype these new consumer experiences, viewing generative AI as a potential game-changer for content creation in the metaverse.

Kyutai, a French AI lab, has introduced Moshi, an AI assistant capable of natural conversations with users. Developed by a team of eight in just six months, Moshi stands out for its real-time speech capabilities and low latency (200-240 milliseconds). The assistant uses a novel "Audio Language Model" approach, compressing audio data and treating it like pseudo-words to work directly with audio.

Despite limitations due to its 7 billion parameter model, Moshi shows impressive language capabilities and speed. Kyutai plans to release Moshi as open-source technology in the coming months, allowing developers and researchers to examine and extend it. This aligns with Kyutai's commitment to open science and transparency in AI development.

Meta is updating its labeling system for AI-generated images on Facebook and Instagram, changing from "Made with AI" to "AI info" in response to widespread complaints from photographers. The issue arose when the platform began flagging images that had undergone minor modifications using AI tools, such as retouching in Photoshop, as fully AI-generated.

The new "AI info" label aims to provide more context, although it won't offer specific details about the AI tools used. Meta acknowledges that the previous labels didn't always align with user expectations and is working with industry partners to improve the labeling process. The contextual menu that appears when users tap on the badge will remain unchanged, offering a generic description of generative AI.

TIME has partnered with ElevenLabs to introduce automated voiceovers on TIME.com, integrating the Audio Native technology to create AI-generated audio versions of select articles. This collaboration, which began in March 2023, aligns with TIME's mission to make trusted information more accessible and engaging.

The implementation of Audio Native allows readers to listen to articles voiced by AI, opening a new avenue for content consumption. This move follows TIME's decision to remove its paywall in 2023, further demonstrating the publication's commitment to providing free access to quality journalism. TIME's CTO, Burhan Hamid, expressed enthusiasm for the partnership, highlighting its potential to enhance content accessibility and engagement.

👀 Creative picks

This isn't your regular AI voiceover ad

Artlist, the royalty free music company, created a series of commercials to showcase their new AI voiceover feature. The campaign consists of short, disruptive 30-second videos that play with classic cinematic tropes and flip them on their heads. For example, one ad transforms a typical basketball shoe commercial into an ad for shoes made out of basketballs, while another creates a nature documentary about imaginary pink penguins.

The campaign was born from the creative team's experimentation with the AI voiceover product itself. By feeding various prompts to the voice generator and listening to the catalog of voices, they discovered opportunities to subvert familiar advertising tropes. The resulting videos are designed to make viewers stop, watch, and enjoy a brief moment of creative disruption.

Credits: Artlist / Widepix

Subscribe to keep reading

This content is free, but you must be subscribed to The Mahazine to continue reading.

Already a subscriber?Sign In.Not now

Reply

or to participate.