Ad
Short

OpenAI is providing the US AI Safety Institute at the National Institute of Standards and Technology (NIST) early access to its next frontier AI model. The collaboration aims to improve AI evaluation methods, says OpenAI CEO Sam Altman. He also reaffirmed the company's pledge to dedicate at least 20% of its computing resources to safety measures. This comes after former OpenAI safety researcher Jan Leike claimed the company had not kept this commitment. Altman also said that in May, OpenAI removed clauses that allowed the company to revoke previously granted stock options and required employees to remain silent. The change aims to create an environment where employees can "raise concerns and feel comfortable doing so," Altman writes.

Ad
Ad
Ad
Ad
Short

The Brazilian government has unveiled an artificial intelligence investment plan worth 23 billion reais ($4.07 billion USD) for 2024-2028. The plan aims to develop sustainable and socially responsible technologies while achieving technological independence and competitiveness in the AI sector. Nearly 14 billion reais will fund company innovation projects, while over 5 billion will be invested in AI infrastructure and development. The remaining funds will support training programs, public service improvements, AI regulation measures, and initiatives directly impacting areas like public health, agriculture, the environment, the economy, and education. President Lula da Silva stresses that AI must generate income and create jobs.

Short

Shutterstock and Getty Images have expanded their generative AI services for designers and artists using technology licensed from NVIDIA. Shutterstock's new Generative 3D allows users to create 3D objects and 360-degree backdrops for virtual scenes with text or image prompts. Getty Images has also improved its AI image generation service, which can now produce images twice as fast and give users more control over the style and composition. Both services use NVIDIA's Edify, a multimodal generative AI architecture, and NVIDIA NIM for performance optimization. Edify lets companies train generative models with their own licensed data.

Short

Runway now supports image-to-video with its latest Gen-3 Alpha model. Users can use any image as the starting point for creating a video. The image can be used alone or in combination with a text instruction to further control the video generation. RunwayML introduced the new Gen-3 model at the beginning of July, which is a significant improvement over the previous version.

Video: RunwayML

Ad
Ad
Google News