• AI KATANA
  • Posts
  • Nvidia Launches AI Factories in DGX Cloud Amid the GPU Squeeze

Nvidia Launches AI Factories in DGX Cloud Amid the GPU Squeeze

Also: Spotify CEO teases potential AI-powered capabilities surrounding personalization

Author

AI KATANA
July 26, 2023

Welcome!

Today's newsletter is teeming with notable advancements and critical discussions in the realm of AI. Nvidia is tackling the GPU squeeze with the launch of its DGX Cloud service, aimed at simplifying AI adoption and offering users worldwide the opportunity to tap into the power of AI supercomputing. The health sector is discussing the role of AI in easing the burden of administrative tasks for physicians and creating a health renaissance, with Tom Lawry of Second Century Tech and Todd Unger of the American Medical Association leading the conversation. Meanwhile, Spotify CEO, Daniel Ek, has offered insights into the potential of AI for improving user engagement and reducing advertising costs. On a national level, South Korea is forming an AI alliance involving 20 key companies to boost AI integration across a range of industries. We will also cover the latest AI tools updates and Venture Capital news.

Sliced:

  • 🏭 Nvidia Launches AI Factories in DGX Cloud Amid the GPU Squeeze

  • 🏥 The future of AI in medicine and what it means for physicians and practices with Tom Lawry

  • 🎵 Spotify CEO teases potential AI-powered capabilities surrounding personalization

  • 🇰🇷 Korea launches industry AI alliance joined by 20 major companies

Nvidia has launched the DGX Cloud service, enabling users worldwide to rent its AI supercomputers that run on its newest GPUs - H100 and A100, which are currently scarce. The DGX Cloud can be accessed via Nvidia's or Oracle's cloud infrastructure and includes a software interface for managing and monitoring AI workloads. Nvidia's goal is to operate its AI infrastructure like a factory, converting data into actionable information, and offering technical expertise for its deployment and operation. Despite the high premium for Nvidia's hardware and software access, the company believes investments in AI will yield long-term operational savings for customers. The launch comes amid increased interest in AI hardware due to the potential of generative AI models in various industries.

In a conversation between Tom Lawry, the managing director of Second Century Tech and a leading AI transformation advisor, and Todd Unger, the American Medical Association’s Chief Experience Officer, the pair discussed the future of AI in medicine. Lawry emphasized that AI is not meant to replace but to mimic human brain functions, with its key role being to automate low-value, repetitive tasks, thus freeing physicians to focus more on patient care and less on administrative burdens. He also highlighted the potential for AI to alleviate the cognitive load on physicians, who currently struggle to keep up with the exponentially growing volume of medical data. Lawry urged leaders to understand AI, think differently, and digitally upskill their teams. Looking ahead to 2033, he envisioned a "health renaissance," with physicians liberated from many administrative burdens, and patients more engaged in managing their own health.

Spotify's CEO, Daniel Ek, discussed potential future applications of AI during the company's Q2 earnings call. Ek highlighted how AI could enhance user experience by personalizing content, summarizing podcasts, and generating ads. Citing the success of their AI-powered DJ feature, he suggested the company could further exploit AI's potential to personalize and contextualize content. Generative AI could be deployed to succinctly summarise podcast content, improving user engagement and aiding creators. Another suggested use was in creating AI-generated audio ads, reducing costs for advertisers. The discussion comes amid Spotify's attempts to patent a "text-to-speech synthesis" system that converts text into emotionally expressive speech, suggesting a move beyond the current DJ feature. Spotify's user base grew by 17% YoY, reaching 220 million paying subscribers and 551 million monthly active users, while also reporting an operating loss of $274 million. The company also announced a price increase for its premium plans.

South Korea has formed an industrial artificial intelligence (AI) alliance comprised of 20 major companies across diverse sectors including automotive, shipbuilding, and steel. Notable participants include Hyundai Motor Co., LG Electronics Inc., and Samsung SDI Co. This alliance, an expansion of the existing Industrial Digital Transformation Alliance, is set to expedite the incorporation of AI technology across various industries through three technology divisions and two policy divisions. By year's end, the technology division plans to identify 40 projects for commercializing industry AI solutions, while the policy division will develop tasks to use and promote industrial data and recommend areas for legal and regulatory enhancements. The initiative aims to drive fundamental industry innovation and increase competitiveness across the industry value chain amidst global uncertainties.

🛠️ AI tools updates

Workspace Labs lets select organizations experiment with new generative AI features in Workspace and provide feedback to Google so we can improve the features before they’re broadly available. Organizations in Workspace Labs will have access to the latest AI-assisted writing capabilities in Gmail and Google Docs and we’ll be bringing exciting new features to the program in the coming months.

This research illustrates that large-scale training of reinforcement learning (RL) agents results in a general in-context learning algorithm capable of adapting swiftly to novel, open-ended embodied 3D tasks, similar to human capacity. The Adaptive Agent (AdA) developed displays hypothesis-driven exploration, efficient exploitation of obtained knowledge, and successfully learns from first-person demonstrations. This adaptability arises from meta-reinforcement learning across a large, varied task distribution, a policy parameterised with an extensive attention-based memory architecture, and an effective automated curriculum that prioritises tasks at the frontier of an agent's capabilities. The study also provides evidence of scaling laws related to network size, memory length, and the diversity of the training task distribution, setting the stage for increasingly versatile RL agents that excel in expanding open-ended domains.

💵 Venture Capital updates

Hippocratic AI, a generative AI company specializing in safety-focused language models for healthcare, has raised an additional $15 million in funding, taking its total to $65 million since its launch. The company also introduced its Founding Partner Program, partnering with various healthcare systems and digital health firms to refine its technology, set up use cases, and ensure patient safety. In other funding news, medical generative AI startup GenHealth AI raised $13 million in a round co-led by Obvious Ventures and Craft Venture, and senior and disability tech startup K4Connect secured $8.9 million in funding, co-led by Bryce Catalyst and AVP, taking its total funding to $39 million. GenHealth AI develops a large medical model trained on medical data, while K4Connect aims to enhance and scale up its technology offerings.

🫡 Meme of the day

⭐️ Generative AI image of the day