Daily AI Newsletter: Nvidia CEO: ‘We need bigger GPUs’

In today’s email:

📣 Google introduced VLOGGER.
🎥 Stability AI introduced Stable Video 3D
📈 One Simple Prompt to Boost Claude 3 and Mistral’s Performance
🧰 10 new AI-powered tools and resources. Make sure to check the online version for the full list of tools.

Nvidia’s keynote at GTC held some surprises

Nvidia's President Jensen Huang captivated a large audience at the SAP Center in San Jose, not with a concert, but with an announcement at the company's GTC event, emphasizing Nvidia's focus on science, algorithms, computer architecture, and mathematics. He highlighted Nvidia's journey from its inception in 1993, focusing on accelerated computing and the advancements in graphics processing units (GPUs), which are crucial for gamers and the broader computing industry.

During the event, Huang introduced Nvidia's new Blackwell platform, a significant advancement over the previous Hopper GPU, boasting remarkable speed improvements and efficiency in power consumption. The Blackwell platform exemplifies Nvidia's commitment to pushing the boundaries of chip design, aiming to enhance computational power and efficiency.

Furthermore, Nvidia announced new tools for self-driving car developers and introduced the Nvidia NIM software platform, designed to streamline AI model deployment across various industries. NIM integrates with major platforms and supports models from leading AI research entities, showcasing Nvidia's strategy to facilitate and accelerate AI adoption and its potential impact on various digital domains, heralding a new era in generative AI.

NVIDIA also announces project GR00T. The foundation model for humanoid robots and major Isaac robotics platform update

AI-Powered Creativity

Instant, polished presentations powered by AI. Impress your audience effortlessly with Gamma. Engage users on any device. Measure engagement, get quick reactions, and collaborate seamlessly.

Try for free.

Google DeepMind researchers just developed VLOGGER.

VLOGGER is an innovative method developed by Enric Corona, Andrei Zanfir, Eduard Gabriel Bazavan, Nikos Kolotouros, Thiemo Alldieck, and Cristian Sminchisescu for generating talking human videos through text and audio input, using a single image of a person. This method employs a two-part process involving a stochastic human-to-3D-motion diffusion model and a novel diffusion architecture that combines text-to-image models with temporal and spatial controls. Unlike previous techniques, VLOGGER does not require individual training for each person, eschews the need for face detection and cropping, and produces entire images of the subject, accommodating a wide range of human scenarios.

VLOGGER's framework generates photorealistic, variable-length videos of a talking individual, incorporating head movements and gestures. The first stage of this two-stage pipeline converts an audio waveform into intermediate body motion controls, influencing gaze, facial expressions, and poses. Subsequently, a temporal image-to-image translation network uses these controls and a reference image to create the video frames, ensuring the subject's identity remains consistent throughout the video.

The method's ability to produce a diverse array of realistic videos of the subject is one of its standout features, with considerable variations in motion while maintaining a consistent background. This diversity, demonstrated in an analysis of 80 generated videos, indicates substantial pixel color variations, particularly in the subject's movements, showcasing VLOGGER's potential for creating dynamic and lifelike human video representations.

Sam Altman: OpenAI, GPT-5, Sora, Board Saga, Elon Musk, Ilya, Power & AGI

Here's a summary of the key points:

Compute is predicted to become a highly valuable commodity, potentially the most precious in the world.
Expectations are set for the development of remarkably capable systems by the end of this decade, leading towards AGI (Artificial General Intelligence).
The journey to AGI is anticipated to involve significant power struggles, with substantial power awarded to the first developer of AGI.
Sam Altman discusses his experiences with OpenAI's board, describing it as a painful, chaotic, and learning-filled episode.
OpenAI's resilience and the importance of structuring and incentivizing a board appropriately are highlighted.
The podcast touches on the rivalry and parting of ways between OpenAI and Elon Musk, with Musk doubting OpenAI's success.
OpenAI's progression from DALL·E 1 to 3 and GPT-3 to GPT-4 is discussed, showcasing advancements in AI capabilities.
The importance of balancing AI development with understanding potential risks and the impact on society is emphasized.
The conversation explores the challenges of AI in data usage, privacy, competition in the tech industry, and the potential for AI integration in search engines.
Altman speculates on the future of AGI, its potential timeline, and the importance of addressing existential risks associated with AI.
The discussion concludes with philosophical reflections on nature, evolution, and the human experience.

Other stuff

Using ChatGPT for Podcast Suggestions
Stability AI introduced Stable Video 3D
LLMs are trained to block harmful responses. Old-school images can override those rules.
AI versions of deceased loved ones could be a serious threat to mental health
One Simple Prompt Boosts Claude 3 and Mistral’s API Calling Performance Beyond GPT-4
YouTube creators, it’s time to start labeling AI-generated content in your videos

Win 1 year free Superpower ChatGPT Pro🤑