Ani Kembhavi
Ani Kembhavi is a venture partner at AI House and an expert in computer vision and embodied AI. He is director of science strategy at Wayve and was previously a senior director at the Allen Institute for AI and a professor at the University of Washington.
Allen Institute for AI launches Satlas: AI super-resolution map enhancing Sentinel-2 images to reveal turbines, platforms, solar farms, and tree cover
The Allen Institute for AI’s new Satlas uses deep-learning super-resolution on Sentinel-2 satellite imagery to generate sharper, downloadable monthly maps highlighting offshore and onshore wind turbines, platforms, solar farms, and tree cover for most of the world.
Allen Institute for AI launches Satlas: AI super-resolution map of global tree cover and renewable energy projects
The Allen Institute for AI launched Satlas, a free global mapping tool that uses satellite imagery and generative AI super-resolution to produce high-resolution maps of tree cover and renewable energy projects, updated monthly to help policymakers track canopy change and solar and wind deployment.
OpenAI reveals GPT-2 language model that generates translations, summaries, trivia answers and human-like essays
OpenAI’s GPT-2 demonstrates strong language-generation abilities—able to translate, summarize, answer trivia, and produce convincing essays—highlighting advances and potential risks in AI-written text.
Allen Institute’s Iconary uses Pictionary-style play to train AllenAI on common-sense scene understanding
Researchers at the Allen Institute for Artificial Intelligence released Iconary, an online Pictionary-style game where a system called AllenAI tries to guess drawings of complex scenes it hasn’t seen before to teach and evaluate common-sense reasoning. The game forces AI to identify visual elements, infer their relationships, and translate scenes into simple language—skills standard game-based benchmarks don’t capture.
Allen Institute launches Iconary: Pictionary-style game where AllenAI and humans draw and guess to test common-sense image-language understanding
Allen Institute for AI created Iconary, a Pictionary-style game that lets humans and its AllenAI algorithms communicate by drawing and guessing with a menu of icons, testing machine image recognition, language understanding, and rudimentary common sense. The web-based game lets players and the system alternate roles as drawer and guesser using symbol palettes rather than freehand sketches.
Organizations
-
AI House Venture Partner · current
