• AI KATANA
  • Posts
  • SIGGRAPH Special Address: NVIDIA CEO Brings Generative AI to LA Show

SIGGRAPH Special Address: NVIDIA CEO Brings Generative AI to LA Show

Also: Google and Universal Music negotiate deal over AI ‘deepfakes’

Author

AI KATANA
August 09, 2023

Hi!

Welcome to your AI KATANA! Dive into today's top stories as NVIDIA's CEO reveals transformative updates on generative AI at the SIGGRAPH conference, promising a revolution in graphics and computational capabilities. On the music front, negotiations between tech giant Google and Universal Music seek to reshape the landscape of AI-generated tunes. And as you dig deeper, explore the inherent biases of AI language models, marking a compelling conversation around the blend of tech and sociology. Also, don't miss out on the latest tool and venture capital updates in the AI domain.

Sliced:

  • 🦾 SIGGRAPH Special Address: NVIDIA CEO Brings Generative AI to LA Show

  • 🤖 Nvidia CEO: We bet the farm on AI and no one knew it

  • 🎧 Google and Universal Music negotiate deal over AI ‘deepfakes’

  • 🤓 Why it’s impossible to build an unbiased AI language model

At the prestigious SIGGRAPH conference, NVIDIA's CEO, Jensen Huang, presented a series of significant advancements in the realm of generative AI. Among the highlights, he introduced the GH200 Grace Hopper Superchip, a next-gen platform that merges a 72-core Grace CPU with a Hopper GPU, designed to handle the world's most intricate generative tasks. To simplify and expedite the adoption of generative AI, Huang launched the NVIDIA AI Workbench, an inclusive toolkit that offers developers an effortless process for creating and scaling AI models across platforms. Moreover, NVIDIA has revealed major updates to its Omniverse platform, integrating generative AI and OpenUSD, thereby offering tools for constructing, simulating, and collaborating across diverse virtual worlds. Furthermore, NVIDIA announced partnerships with industry leaders and introduced cutting-edge hardware systems to support the demands of generative AI and digitalization, demonstrating its commitment to pushing the frontiers of AI-infused graphics.

In 2018, Nvidia made a pivotal business decision to shift focus towards AI-powered image processing, specifically in the realms of ray tracing and intelligent upscaling (RTX and DLSS). As Nvidia's CEO Jensen Huang reflected during his keynote at SIGGRAPH in Los Angeles, this move was in response to the perceived limits of rasterization, a traditional method of rendering 3D scenes. The hardware and software architectures developed to support RTX and DLSS proved beneficial to the burgeoning machine learning community. With Nvidia's unique hardware, like the H100 and GH200, the company catered to the massive computational requirements of AI model training, claiming a significant position in the market. Huang envisions a future dominated by natural language interfaces across industries and highlights Nvidia's role in it. He sees the shift as a boon for Nvidia, likening the company's position to selling tools during a gold rush.

Google and Universal Music are negotiating a licensing agreement for AI-generated songs that replicate artists' voices and melodies. The surge of generative AI technology has led to an increase in "deepfake" songs that can convincingly imitate famous artists, often without permission. For example, AI has been used to produce tracks mimicking Frank Sinatra, Johnny Cash, and even rappers Tupac and Notorious B.I.G. Universal Music's general counsel has emphasized the importance of an artist's voice to their identity and livelihood, suggesting unauthorized use is wrong. The goal of the Google-Universal partnership would be to allow fans to create AI-generated songs legitimately, compensating copyright holders in the process. Warner Music has also been in discussions with Google on similar lines. While some artists, like Drake and Ice Cube, have criticized the misuse of their voices, others, like Grimes, see potential in the technology. This move could also bolster Google's position against competitors like Microsoft, which has invested heavily in AI with OpenAI's GPT-4. Universal Music has pushed streaming platforms to prevent unauthorized AI scraping, and Google has previously explored AI-powered music software but had indicated no commercial release intentions until recently.

AI language models, such as OpenAI's ChatGPT, are often seen as objective tools, but in reality, they harbor various biases stemming from their training data and the developers who create and train them. Despite calls for unbiased, fact-based AI systems, creating an entirely neutral AI remains technically unfeasible. Recent research on large language models has highlighted the spectrum of political biases in these AI systems, emphasizing that no language model can be entirely free from political leanings. The data used in training AI is a significant source of these biases, and even the procedures of developing these models can introduce further predispositions. Addressing this issue isn't straightforward because biases are complex social challenges without easy technological solutions. Acknowledging these biases and aiming for transparency can help users approach AI outputs more critically.

🛠️ AI tools updates

Spotify is globally rolling out its AI-driven "DJ" feature, which offers personalized music lineups complemented by AI-generated spoken-word commentary that mimics a radio DJ experience. First introduced in North America, this feature moved to the U.K. and Ireland and is now expanding to approximately 50 global markets, including Sweden, Australia, and South Africa. Despite its broader availability, the "DJ" remains in beta, accessible only to premium users and is exclusively available in English, with most of the European Union yet to receive access.

💵 Venture Capital updates

SoftBank Group reported its first investment gains in 18 months, marking $1 billion in gains at its Vision Fund unit during the April-June quarter. Encouraged by this performance, and despite an overall quarterly net loss of approximately $3.3 billion influenced by factors such as a slowdown in the semiconductor industry, the company is set to further invest in AI-related sectors. There's a noticeable uptick in market conditions as the number of companies in SoftBank's funds that saw increased valuations nearly doubled from the previous quarter. CFO Yoshimitsu Goto expressed optimism about the tech startup sector and highlighted a stronger focus on genuine AI technology in future investments. The impending listing of chip designer Arm in the U.S. is also underway.

🫡 Meme of the day

⭐️ Generative AI image of the day