Elorian, the Palo Alto AI startup, has launched its first model, Glimpse, which allows AI to 'see' by analyzing images in a more human-like way. Unlike current models that describe images from memory, Glimpse can zoom in and analyze specific parts of an image, making it more accurate in understanding visual data.

Andrew Dai, CEO of Elorian, believes today's AI models are limited due to their inability to 'see'. He gives the example of a board game photo, where a cutting-edge AI model couldn’t determine the winner. In contrast, Glimpse can crop into the image, measure and compare, providing a more nuanced understanding.

The model is part of a broader strategy from Elorian to move from mere visual understanding to action, potentially leading to more advanced robotics and automation. With a smaller team and more efficient training, Glimpse is cheaper than its competitors while outperforming them in tasks like scientific chart reasoning and short-form video captioning.

The initial targets for Glimpse include robotics firms, scientific labs, media companies, and creative businesses. Dai aims to make Glimpse the world's best visual reasoning model next year, leading on every visual reasoning test.