A call for embodied AI

Giuseppe Paolo,Jonas Gonzalez-Billandon,Balázs Kégl

2024-09-13

Abstract:We propose Embodied AI as the next fundamental step in the pursuit of Artificial General Intelligence, juxtaposing it against current AI advancements, particularly Large Language Models. We traverse the evolution of the embodiment concept across diverse fields - philosophy, psychology, neuroscience, and robotics - to highlight how EAI distinguishes itself from the classical paradigm of static learning. By broadening the scope of Embodied AI, we introduce a theoretical framework based on cognitive architectures, emphasizing perception, action, memory, and learning as essential components of an embodied agent. This framework is aligned with Friston's active inference principle, offering a comprehensive approach to EAI development. Despite the progress made in the field of AI, substantial challenges, such as the formulation of a novel AI learning theory and the innovation of advanced hardware, persist. Our discussion lays down a foundational guideline for future Embodied AI research. Highlighting the importance of creating Embodied AI agents capable of seamless communication, collaboration, and coexistence with humans and other intelligent entities within real-world environments, we aim to steer the AI community towards addressing the multifaceted challenges and seizing the opportunities that lie ahead in the quest for AGI.

Artificial Intelligence

What problem does this paper attempt to address?

### Problems Addressed by the Paper

The paper primarily explores the development of **Embodied Artificial Intelligence (E-AI)** and views it as a crucial step towards achieving **Artificial General Intelligence (AGI)**. The core points of the paper include:

1. **Limitations of Current AI Technology**:
- Current large language models (LLMs), although performing well on certain tasks, are essentially static and cannot evolve over time and experience.
- These models lack the ability to dynamically adjust knowledge and cannot actively search for valuable new information.
- There are issues with alignment difficulties and generating inaccurate information (i.e., fabrication).

2. **Importance of E-AI**:
- Embodied intelligence emphasizes core capabilities such as perception, action, memory, and learning, which enable AI to interact continuously and dynamically with the real world.
- The goal of E-AI is to design AI agents that can adapt to environmental changes and evolve without human intervention.
- Through interaction with the real world, E-AI agents can better understand causal relationships, leading to more reasonable decision-making.

3. **Theoretical Framework**:
- A theoretical framework based on cognitive architecture is proposed, emphasizing the four core components of perception, action, memory, and learning.
- This framework aligns with Friston's active inference principle, providing a comprehensive approach to the development of E-AI.

4. **Challenges and Opportunities**:
- Despite the immense potential of E-AI, it still faces many challenges, such as the establishment of new AI learning theories and innovations in advanced hardware.
- By addressing these issues, E-AI not only aids in achieving AGI but also deepens our understanding of general cognition.

In summary, the paper aims to encourage the AI research community to focus on the development of E-AI, considering it a necessary step towards achieving true intelligence and general AI.