We all need and want a world where robots do all the work for us and create the goods and most of the necessary goods and services, yet this progress has been jeopardized because robotics is being done the wrong way.
Here's what's wrong and what we need to do:
What differentiates animals from humans are higher faculties, something other beasts don't have, this higher faculties allow us to plan, engineer, architecture, have feelings, understand compassion, and other abstract concepts that other beasts don't have.
The thing is: Robotics in the status quo are still being done,by training robots from scratch to overfit the understanding of the world, how the engineers are doing is they would train a robot to perform a basic tasks like colleting fruits is by teaching what each thing is and teaching it to perform that action, this doesn't work well and doesn't generalize well, sure there are new methods like modelling physical world and having the robot train their action in the digital model of the physical world, this is awesome and will certainly push the industry forward, yet there is a better way.
Robots should have a physical intelligence model and higher faculties model, in this way the robots can have extraordinary intelligence.
The physical intelligence model is the model responsible for moving, gestures, sensing, basic functions, and understanding basic commands from hugger faculties and translating this commands into physical movements. The higher faculty models are responsible for everything else.
They communicate by a protocol or even natural language: the higher faculties models can send commands to the physical intelligence like “move your hands and grab the apple in front of you” and the physical layer would perform the movement to grab the apple.
Models like Gemini 2.0 and ChatGPT, can already received a video feed image and understand with great complexity what is going on around, a model such as this could be the higher faculties model. The adjustment needed to deploy this specialized robotics model would be to make it send commands to the physical intelligence layer and adjusting its model to learn and retain memory.
For the physical intelligence model, there are many robotics models that could fulfill this role, robotics models are small and that's one of the reasons why it doesn't perform as well. The adjustments needed is just for it to be adjusted to work well with the higher faculties model and perform commands seamlessly.
This is already possible with current technology, we just need to put these technologies together and build excellent robots that work, and advance the human endeavor. This is essential to create a prosperity society.