Modern day LLMs can only model reality through.. well language. They have the ability to predict text, but thats not enough. The future is in models which are REALITY NATIVE. A model which models reality as accurately as possible, through senses similar to a human, or better. A model which has sound, light, touch, the change of temperature, and many more things that aren't sensed by humans all represented by A SINGLE TENSOR at each moment. That is the future. From this point on, I will commit my entire life to making that the reality. And along the way I will DEMOCRATIZE IT ALL. Knowledge deserves to be accessible to every human on the earth. Not just the rich and powerful. Train a model like that with new levels of compute and data. Then give it an immense context window, and let it experience. Then, we'll have a wise robot.