LLM field will also be relevant to the field of robotics?

[rifat28dddd]

This is System . System , on the other hand, is slow and deliberate, more like reasoning and planning, using our conscious thought. I think GPT- will appear on System One at some point. My favorite example is the verb “open.” Imagine the complexity of the word “open.” Opening a door is not the same as opening a window, nor is opening a bottle or a cell phone. But there is no problem for people to understand that “open” means different actions when interacting with different objects.



But But so far we have not seen a robot model that can cameroon phone numbers generalize these verbs to low-level motor control. I would like to see a model that understands the abstract meaning of these verbs and can generalize to different scenarios that are meaningful to humans. We have not seen that yet, but I am optimistic that this moment will come in the next two to three years. Sonia Huang What about “System” thinking? How do you think we can achieve that? Do you think the work of reasoning in the Jim Fan Yes, absolutely relevant.



I think we've seen some very powerful models on the System Two side that can reason, plan, and even program.These are the LLM and the most advanced models we see today. But integrating the “System” model with the “System” is a research challenge in itself. So the question is, should we have one big model for the basic robot model, or should we have some kind of cascade approach, keeping the “System One” and “System Two” models independent and having them communicate in some way? This is still an open question and each has its own advantages and disadvantages.