#044 - Google’s $20 Million Fund for AI Experts, Stable Audio AI Tool To Create High-Quality Music, AI Hardware Startups Are Challenging the GPU Monopoly Of Nvidia.
Fresh & Hot curated AI happenings in one snack. Never miss a byte 🍔
This snack byte will take approx 7 minutes to consume.
AI BYTE # 1 📢 : Stable Audio: The First AI Tool To Create High-Quality Music For Commercial Use
⭐ Are you looking for a way to create original and engaging music for your projects, without spending hours in the studio or hiring expensive composers?
If so, you might want to check out Stable Audio, the latest AI tool from Stability AI, the London-based startup behind the open-source image-generating AI model Stable Diffusion.
Stable Audio is a web app that lets you generate music and sound effects in any style and genre, using only a text description and a desired duration.
You can type something like “EDM, driving, synthesizer, bass, upbeat, 120 BPM” and get a 90-second track that sounds like a professional production.
Or you can type “car passing by, city, traffic, honking” and get a realistic sound effect for your video game or film.
Stable Audio uses a technique called latent diffusion, which is similar to the one used by Stable Diffusion to generate images.
The underlying model, which has about 1.2 billion parameters, learns how to gradually remove noise from a starting song made almost entirely of noise, moving it closer to the text description.
This allows the model to create longer and more coherent audio than previous generative music tools, which often devolve into random noise after a few seconds.
Stable Audio was trained on a collection of around 800,000 songs from the commercial music library AudioSparx, which provided a diverse and rich dataset of mostly independent artists.
Stability AI partnered with AudioSparx to share revenue with the artists who contributed to the training data, as well as to offer them an opt-out mechanism if they didn’t want their work to be used.
Stable Audio is not only a fun and creative tool for music lovers and enthusiasts, but also a powerful and innovative tool for professionals and businesses who need high-quality music for commercial use.
It claims that Stable Audio is the first AI tool that can generate music at 44.1 kHz, which is the standard quality for CDs and digital platforms.
Users who pay $11.99 per month for the Pro tier can generate 500 commercializable tracks up to 90 seconds long monthly, while free tier users are limited to 20 non-commercialized tracks at 20 seconds long per month.
Stable Audio is part of Stability AI’s mission to unlock humanity’s potential by building foundational AI models across different content types or modalities.
It is also known for its research on generative image models, such as Stable Diffusion and Dance Diffusion, which can create realistic images and artworks from text or sketches.
With Stable Audio, Stability AI hopes to democratize music creation and inspire new forms of expression and creativity.
AI BYTE # 2 📢 : How AI Hardware Startups Are Challenging the GPU Monopoly Of Nvidia?
⭐ If you are working in the GenAI space, you know how important it is to have access to powerful and reliable hardware.
Generative AI models, such as OpenAI’s GPT-4, require a lot of compute resources to generate realistic text and images. But the supply of such hardware, especially GPUs, is not keeping up with the demand.
In fact, Microsoft recently warned its shareholders of potential Azure AI service disruptions if it can’t get enough AI chips for its data centers.
Nvidia, the leading supplier of GPUs, is facing a huge backlog of orders from tech giants like Baidu, ByteDance, Tencent and Alibaba. Nvidia’s best AI chips are sold out until 2024!
This situation is creating a gap between the “GPU rich” and the “GPU poor” in the AI industry. The former group includes incumbents like Google and OpenAI, who can afford to buy or build their own custom chips for AI inferencing.
The latter group consists of mostly European startups and government-backed supercomputers, which have to rely on scarce and expensive GPUs.
But there is hope for the GPU poor. A new wave of AI hardware startups, such as d-Matrix, Mythic, Groq and others, are developing innovative solutions for AI inferencing that promise to be cheaper, faster, and more efficient than GPUs.
These startups are using different technologies, such as analog computing, tensor processing, and neuromorphic engineering, to create new kinds of chips that can run AI models at a lower cost of ownership.
One of these startups, d-Matrix, recently raised $110 million to commercialize its inference compute platform, which it claims to be the first of its kind.
D-Matrix says that its platform enables inference at scale with minimal power consumption and latency. The startup also claims that its platform will make generative AI commercially viable.
If these claims are true, d-Matrix and its peers could be game-changers for the generative AI space. They could democratize access to high-performance hardware and level the playing field for startups and researchers who want to experiment with new AI techniques and applications.
They could also spur innovation and competition in the AI hardware market, which is currently dominated by Nvidia.
As a technology expert and a substack technology audience member, I am excited to see how these AI hardware startups will shape the future of generative AI.
I think they have the potential to disrupt the status quo and create new opportunities for growth and creativity. What do you think? Let me know in the comments below.
AI BYTE # 3 📢 - Google’s $20 Million Fund for AI Experts and Think Tanks
⭐ Google has announced a new initiative called the Digital Futures Project, which aims to support researchers and public policy solutions around AI.
The project will provide grants to think tanks and academic institutions developing AI expertise on topics such as global security, labor, economic growth, and responsible innovation.
As part of the project, Google’s charitable arm Google.org is establishing a $20 million fund that will support independent thinkers who are looking into the implications and opportunities of AI for society.
The fund will also foster collaboration and dialogue among different stakeholders, such as governments, industry, civil society, and academia.
The inaugural grantees of the Digital Futures Fund include the Aspen Institute, Brookings Institution, Carnegie Endowment for International Peace, the Center for a New American Security, the Center for Strategic and International Studies, the Institute for Security and Technology, Leadership Conference Education Fund, MIT Work of the Future, R Street Institute and SeedAI.
Google says that the fund will support organizations around the world, not just in the U.S., and it will have more to share on that front soon. The company also says that it hopes the project will advance independent research on AI that helps this transformational technology benefit everyone.
The announcement comes at a time when AI is becoming more powerful and potentially dangerous, raising questions about fairness, bias, misinformation, security, and the future of work.
Earlier this year, four of the top players in AI, including OpenAI, Microsoft, Anthropic as well as Google, formed the Frontier Model Forum to ensure the safe and responsible development of AI models.
The U.S. Congress will also turn its attention to AI in a closed-door meeting with all 100 senators who will hear from Elon Musk, along with Meta CEO Mark Zuckerberg, Microsoft co-founder Bill Gates, OpenAI CEO Sam Altman, Google CEO Sundar Pichai and other tech leaders.
Google has been one of the pioneers and leaders in AI research and innovation. The company has published its own AI principles and continues to invest in AI projects that have a positive social impact. However, the company has also faced some challenges and controversies around its AI ethics and practices.
With the Digital Futures Project, Google is showing its commitment to supporting diverse and independent voices on AI’s impact and governance. The project is also an opportunity for Google to engage with other stakeholders and learn from their perspectives and insights.
What do you think of Google’s new initiative?
Do you think it will help address some of the challenges and opportunities of AI? Share your thoughts in the comments below.
