Study Shows Biological Intelligence Surpasses Current A.I. Models
A groundbreaking study by Michael Timothy Bennett of the Australian National University challenges the supremacy of artificial intelligence.
Published in the Journal of The Royal Society Interface, the research reveals that biological systems are more efficient and adaptable than current AI models. Bennett’s study defines intelligence as the ability to adapt efficiently using minimal resources.
Biological systems excel at this, adapting with significantly less data and energy than AI systems. In a striking example, a small collection of cells grown in a lab outperformed modern reinforcement learning algorithms in playing the classic game Pong.
The key to biological systems’ superiority lies in their “multiscale competency architecture” (MCA). This structure allows adaptation at all levels, from cells to ecosystems. Each cell acts as an autonomous agent, cooperating with others to form larger, more complex systems.
Unlike centralized AI systems, biological systems decentralize decision-making. This approach enables faster and more flexible adaptation. The loss of a single cell or organ doesn’t necessarily compromise the entire organism’s survival.
Current AI systems face limitations due to static abstraction layers. Once designed, an AI’s architecture remains fixed, restricting its ability to respond to new situations. Biological systems, however, can dynamically adjust their interaction with the environment.
Key Takeaways from Bennett’s Study
The study also addresses imperfections in biological systems, using cancer as an example. Cancer occurs when cells lose connection with the organism’s collective identity. This phenomenon parallels potential failures in overly restricted artificial systems.
Bennett‘s research suggests that principles of delegation and adaptability could improve cybernetic system design. These insights might also inform AI regulation, warning against imposing excessive restrictions that could hinder adaptation.
The study concludes that biological intelligence is intrinsically linked to its ability to delegate control and adapt at various abstraction levels. This model offers valuable lessons for designing more robust and efficient artificial systems.
While primarily theoretical, Bennett’s work lays a foundation for future empirical research. It challenges current AI paradigms and opens new avenues for understanding and constructing intelligent systems, both biological and artificial.
Download study here.
This article was produced by The Rio Times’ automated newsroom system. How we use AI · Report an error · Editorial responsibility: Matthias Camenzind, Editor-in-Chief