Kris Glowacki’s Post

I have an ordinary marble in an ordinary glass. I turn the glass upside down as I set it on the table. I then move the glass to the microwave oven. Where is the marble? - this simple task that Matt Berman presents to LLM models in his YouTube videos is enough to break them down. Some models can come close to solving it using Chain of Thought, but for a human, it's so simple - you just need to imagine the described situation. Fei Fei Li, one of the leading AI researchers, has just launched a project aimed at equipping machines with Spatial Intelligence. LWM, or Large World Models, will be trained to understand and generate scenarios in three dimensions, grasping the laws of physics and the relationships between objects. Intuitively, I agree with her and Yann LeCun - language alone is not enough to achieve true AGI. AI must learn through the observation of images and understand them. https://lnkd.in/gdtG3My6

To view or add a comment, sign in

Explore topics