Udio: AI music creation stuns the industry

AI tool transforms concepts into realistic musical compositions

Source: Nvidia

Good morning!

In this edition, we highlight the launch of Udio, an innovative artificial intelligence platform that has impressed the music creation industry. Udio allows users to generate realistic music from simple ideas, making access to music production easier. We will explore this and other updates below.

In today's Mail:

  • Udio: Revolution in Music Creation with Artificial Intelligence

  • OpenAI launches new GPT-4 Turbo

  • Mistral releases new Mistral 8x22B model

  • JPMorgan Chase reports intensive use of AI

  • Tech giants compete for data to train their models

  • USA makes chip deal with TSMC

  • Intel launches Gaudi 3 for business AI

  • Apple plans to update Mac line with M4 chips

  • OpenAI fires researchers for alleged leaks

  • Invisibility cloak tricks AI cameras

  • And more: TikTok tests avatar creation, updates on Google's Gemini 1.5 Pro, Sanctuary AI robots in the automotive industry, Adobe buys videos to train model, and other news.

  • This week's video: Google robots learn to1 play football

  • Nanotutorial: edit images generated by DALL-E.

Reading time: 10 minutes

NEWS OF THE WEEK

UDIO

The new artificial intelligence tool called Udio promises to revolutionize music creation, allowing the generation of realistic musical compositions on demand. Udio can synthesize tracks that closely mimic music produced by humans, presenting impressive quality that has received praise from both users and experts in the field of digital music. This breakthrough represents a significant change in the music production landscape, facilitating access to music creation.

In addition to its technical capability, Udio offers an intuitive interface that makes it easy for musicians and music enthusiasts to explore their creativity without the need for expensive equipment or advanced musical instrument skills. The tool has the potential to be used in a variety of contexts, from industry professionals seeking inspiration to amateurs wanting to experiment with composition. The impact of Udio on the music industry promises to be broad, as it not only reduces barriers to music creation but may also influence new forms of music consumption and production in the future. Listen to some examples here.

OPENAI

OpenAI has announced a significant update to the GPT-4 Turbo model, now available to subscribers of the Plus, Team, or Enterprise plans of ChatGPT. This enhanced version of GPT-4 Turbo features notable improvements in writing, math, logical reasoning, and coding skills.

GPT-4 Turbo is a multimodal model (accepting text or image inputs and generating text) that can solve challenging problems with greater accuracy than any of the previous models, thanks to its broader general knowledge and advanced reasoning capabilities (its knowledge base goes up to Dec/2023). Additionally, GPT-4 Turbo has a context window of 128k, equivalent to 300 pages of text in a single prompt, and is optimized for chat, but also performs well for traditional completion tasks using the Chat Completions API. With this update, OpenAI reclaims its leadership in developing advanced language models, returning to the top spot in the Arena Leaderboard rankings.

MISTRAL

The AI startup Mistral recently unveiled its newest language model, the Mistral 8x22B, with 281 GB, positioning itself as a direct competitor to tech giants like OpenAI, Meta, and Google. This model is a significant evolution in terms of capacity and processing potential, offering an advanced artificial intelligence tool that promises to revolutionize the way we interact with language-based technologies. The Mistral 8x22B is designed to understand and generate natural language more effectively, supporting a wide range of applications, from virtual assistants to advanced text analysis systems.

Mistral highlights that the 8x22B not only improves performance in traditional language processing tasks but also introduces innovative capabilities that can handle linguistic and cultural complexities more nuancedly. This is achieved through an optimized architecture that allows for deeper and adaptive learning. With this, Mistral aims not only to keep up with its established competitors but also to set new standards of excellence and accessibility in the development of AI language technologies, promoting significant advancement in the field and expanding the possibilities for these technologies in everyday and corporate environments.

JPMORGAN CHASE

The CEO of JPMorgan Chase highlighted in his annual letter to shareholders the growing importance of artificial intelligence in the bank's operations. The 2023 report points out that AI is already being used in various fronts, from improving internal processes to offering new, more efficient services to customers. The technology has been a key tool in analyzing large volumes of data to detect fraud and in personalizing financial services, providing a more targeted and secure experience for users.

In addition to improving operational efficiency, AI is also involved in the development of innovative solutions that can transform key areas of the banking sector. JPMorgan Chase plans to expand the use of this technology, investing significantly in research and development. The strategy includes partnerships with technology startups and the integration of advanced AI systems into its platforms, underscoring the bank's vision of being at the forefront of technological innovation in the financial sector. The initiative is seen as a crucial step in maintaining competitiveness and meeting the increasingly high expectations of customers for personalized and secure financial services.

DATA

A recent Reuters report reveals an intense competition among large technology companies to acquire valuable training data for artificial intelligence systems. These data are crucial for the development of more efficient and accurate algorithms, and companies like Google, Apple, and Microsoft are at the center of this dispute. The search for large volumes of high-quality data has led these companies to explore emerging markets and unconventional data sources, from texts and images to complex behavioral data.

This race for data could lead to significant advancements in the capabilities of AIs but raises important concerns about privacy and ethical use of information. The acquisition and use of these data are becoming an increasingly regulatory battleground, with governments around the world starting to draft legislation to protect citizens' privacy while allowing technological development. The outcome of this dynamic will be crucial in defining the future of AI technologies and their integration into society and the global economy.

TSMC

President Joe Biden announced a significant preliminary agreement with Taiwan Semiconductor Manufacturing Company (TSMC) in the context of the Chips and Science Act. This agreement is a crucial step in strengthening the semiconductor supply chain in the United States and reducing dependence on foreign sources. The collaboration between the US government and TSMC aims to increase the production of advanced chips, essential for various critical technologies, including defense, telecommunications, and infrastructure.

Beyond economic and strategic benefits, this agreement aligns with the broader national security objectives of the USA. The partnership will boost the semiconductor sector, fostering innovation and global competitiveness. The implementation of this agreement is seen as a vital component in ensuring US leadership in the technology sector, particularly at a time when global demand for semiconductors is growing rapidly.

INTEL

Intel announced the launch of the AI accelerator Gaudi 3, part of a comprehensive open systems strategy for business AI. The Intel Vision 2024 event, held in Phoenix, Arizona, was the stage for the launch. Gaudi 3 promises to offer an average of 50% more efficiency in inference and 40% more energy efficiency compared to the Nvidia H100, and this for a fraction of the cost. Intel also announced the availability of Gaudi 3 to original equipment manufacturers (OEMs), including Dell Technologies, HPE, Lenovo, and Supermicro, expanding the market offerings of AI data centers for businesses.

Additionally, Intel revealed its intention to create an open platform for business AI in collaboration with industry leaders such as SAP, RedHat, VMware, among others, to accelerate the deployment of secure generative AI systems, enabled by recovery-augmented generation (RAG). Through the Ultra Ethernet Consortium (UEC), Intel is leading the open Ethernet network for AI fabric, introducing a variety of Ethernet solutions optimized for AI, including the AI network interface card (NIC) and AI connectivity chiplets.

APPLE

Apple is nearing the production of M4 computer processors, which will have artificial intelligence processing capabilities, and plans to update all Mac models with them. The company intends to launch the updated computers later this year and early next year, including new iMacs, an entry-level 14-inch MacBook Pro, high-end 14 and 16-inch MacBook Pros, and Mac minis. After a decline in sales since the end of the pandemic boom, the PC industry places its hopes of revival on a new generation of laptops and desktops with more powerful chips capable of handling AI tasks, such as summarizing documents without the need to send data to the cloud. Intel is preparing such chips, as are competitors, including Qualcomm. Nvidia also plans to use its strength in AI chips to enter the PC market with a new chip by 2025.

Apple plans to highlight the AI processing capabilities of the new chips and how they will integrate into the next version of macOS. The news comes ahead of Apple's annual developer conference in June, where the iPhone maker is expected to announce new AI partnerships and reveal significant changes in iOS. Mac sales fell 27% in Apple's last fiscal year, which ended in September.

OPENAI

OpenAI recently fired two of its AI security researchers, Leopold Aschenbrenner and Pavel Izmailov, accused of leaking confidential information. Aschenbrenner was involved in the Superalignment project, which aims to understand and govern advanced artificial intelligence, while Izmailov worked specifically with AI reasoning. These incidents highlight growing concerns about internal security and the management of sensitive information within organizations that develop critical technologies.

The dismissal of these researchers evidences the tension between the need for accelerated AI development and the imperative security of its development processes. OpenAI, as a leader in the field of artificial intelligence, faces significant challenges in balancing innovation and confidentiality, reinforcing measures to prevent future leak incidents that could compromise both its integrity and global security.

RESEARCH

Scientists from the University of Maryland have developed an “invisibility cloak” that tricks artificial intelligence cameras and prevents them from recognizing people. The innovation comes in the form of a sweater that uses “adversarial patterns” to break human recognition AI systems, making the person “invisible” to AI cameras.

The sweater, which features a waterproof microfleece lining, a modern cut, and anti-AI patterns, has proven to be an effective way to hide from object detectors. However, it is important to note that the sweaters achieved a success rate of about 50% in usability tests. The research represents a significant advance in the area of adversarial attacks on AI and opens new possibilities for personal privacy and security in a world increasingly monitored by recognition technologies.

SHORTS

  • 🎵 Spotify launches "AI Playlist," allowing Premium users to create personalized playlists from simple ideas. The tool, still in beta, uses artificial intelligence to transform simple concepts into music lists, aiming to improve the user experience and make the service more interactive and personalized. Learn more.

  • 🚨 NTT and Yomiuri Shimbun warn of the risk of social order collapse due to generative artificial intelligence. The Japanese institutions propose urgent legislation to contain the risks associated with these technologies, highlighting the need to protect elections and national security from AI abuse. Learn more.

  • 🎥 Adobe is buying videos for US$3 per minute to build an AI model for video generation, following the example of OpenAI. The company is seeking videos that show people performing everyday actions and expressing emotions, to train its artificial intelligence. Learn more.

  • 🌐 Google announced the Gemini 1.5 Pro, now available in more than 180 countries. This enhanced version introduces the ability to understand native audio and features such as system instructions and JSON mode, expanding the possibilities for developers. Learn more.

  • 🛍️ eBay introduces "Shop the Look," a new functionality that uses artificial intelligence to curate personalized looks. The tool analyzes users' style preferences and suggests combinations of clothes and accessories available on the platform, facilitating the shopping experience and making it more personalized. Learn more.

  • 🤖 Sanctuary AI expands its reach in the automotive industry with general-purpose robots, thanks to a strategic partnership and investment from Magna. This agreement aims to integrate Sanctuary robots into automotive manufacturing processes, promoting innovation in challenging industrial environments. Learn more.

  • 🧠 Archetype AI introduced the foundational model Newton to understand the physical world, employing AI to solve real-world problems. The Newton analyzes multimodal temporal data and natural language, improving interactions with physical environments. Learn more.

  • 🌍 Sam Altman of OpenAI proposes a global AI coalition during his trip to the Middle East. During meetings with investors and government officials in the United Arab Emirates, Altman discussed support for the development of large-scale AI infrastructures, including chip supply, energy, and data center capacity. Learn more.

  • 🤑 CEOs of Geometric Intelligence, Gary Marcus, and ingk.com, Damion Hankejh, bet up to US$10 million against Elon Musk's prediction that AI will surpass human intelligence by 2026. They challenge Musk's vision, arguing that current language models still face significant problems and that reliable technology may be decades away. Learn more.

  • 📸 Google Photos expands its AI editing tools to all users. The features, including Magic Eraser, Photo Unblur, and Portrait Light, will be available without a subscription from May 15. In addition, the Magic Editor, which allows complex edits using AI, will be accessible on all Pixel devices and will offer 10 free uses per month for Android and iOS users. Learn more.

  • 🤖 TikTok is developing a tool that creates AI-generated profile images, integrated into its own app. The tool allows users to upload photos and choose styles to generate personalized avatars, which can be used as profile images or shared in TikTok stories. Learn more.

  • 💻 Meta announced details about the next generation of the Meta Training and Inference Accelerator (MTIA), a family of custom chips designed for Meta's AI workloads. This latest version displays significant performance improvements over MTIA v1 and helps power the models for ad classification and recommendation. Learn more.

VIDEO OF THE WEEK

Researchers from Google DeepMind used deep reinforcement learning to train a low-cost humanoid robot to play a simplified one-on-one football game. The resulting agent exhibits robust and dynamic movement skills, such as quick fall recovery, walking, turning, and kicking, and smooth transitions between them. Click here or on the image above to watch.

TOOLS

  • 🏡 mnml.ai: AI rendering platform for architecture and interior design. It offers tools that transform sketches into realistic renderings in seconds, helping architects and designers optimize workflows. With more than 40 rendering styles, it includes options for exteriors, interiors, and landscaping, as well as a concept statement generator. Link.

  • 🧠 Aboard: data management tool that uses artificial intelligence to simplify the organization of daily work. Aboard allows you to organize links, notes, and other data visually and intuitively, offering a browser extension that facilitates the capture of important information. The platform transforms tabs, spreadsheets, and favorites with a visual, searchable, and shareable design, enriched by AI recommendations. Link.

  • 🧬 Sequel: personal longevity assistant with artificial intelligence. Sequel provides a comprehensive view of health by integrating data from lab tests, MRI scans, DEXA scans, supplements, and pharmaceuticals, offering personalized insights. Data processing is done locally on the user's device, ensuring privacy. Users can choose between a completely local model or an advanced experience powered by OpenAI, without compromising data privacy. Link.

  •  Shopify Magic: new tool integrated into the Shopify platform, designed to automate and personalize the e-commerce experience. Using artificial intelligence, Shopify Magic offers personalized product recommendations, optimizes page layouts according to user behavior, and improves digital marketing strategies. Link.

NANOTUTORIAL

🎨 Edit images in ChatGPT

In this nanotutorial, you will learn how to use the new editing interface of ChatGPT to edit your generated images precisely. Step by step:

  • Visit ChatGPT and generate an image with DALL-E 3, typing "Create an image of [desired image]." You will need a ChatGPT+ subscription.

  • Click on the image to open the editor interface.

  • Highlight the area you want to modify using the selection tool.

  • Describe the desired changes in the conversation panel (for example, add, remove, or modify elements).

  • Done! Now you can download for future use by pressing the Save button.

Tip: You can also access the editor in the ChatGPT mobile app.