Tags:
A research collaboration between Walton Institute at South East Technological University (SETU); University of Nebraska-Lincoln; VistaMilk SFI Research Centre; and University of Cambridge recently published work that shows that a single bacterial cell embodies vast computational potential with incredible energy efficiency. This prompted the researchers to propose a ground-breaking concept: wet-tinyML (wet-neuromorphic) computing.
It can be argued that Artificial intelligence (AI) has the potential to reshape almost all aspects of our lives. Machine learning (ML) as the core of AI has evolved tremendously in the past few decades, targeting more powerful and energy-efficient approaches, leading to the development of increasingly intelligent and smaller devices.
This insight propelled the researchers to view bacteria through a computational lens, as they inherently possess computing capabilities with unparalleled energy efficiency. This led them to introduce Gene Regulatory Neural Networks (GRNNs), showcasing gene regulation-based components resembling neural networks in their previous studies.
Building on this, the researchers’ latest collaboration with Walton Institute at South East Technological University (SETU); University of Nebraska-Lincoln; VistaMilk SFI Research Centre; and University of Cambridge revealed that a single bacterial cell embodies vast computational potential with incredible energy efficiency.
The research, titled, “Wet TinyML: Chemical Neural Network Using Gene Regulation and Cell Plasticity,” found that 20% of E. coli’s gene regulatory network (its primary computational system) contains a staggering 5.9 x 10297 neural network-like components. Focusing on this immense computing diversity, it proposes a ground-breaking concept: wet-tinyML (wet-neuromorphic) computing. Here, bacterial cells function akin to neuromorphic chips, that morph over time and in response to various inputs. The research team termed these properties as temporal and input-dependent plasticity that offer a near-infinite array of computational possibilities within a cell barely 2 micrometres wide—a fraction of a human hair’s diameter.
We spoke to research team member Samitha Sulakshana Somathilaka, who says: “Remarkably, these potent computing capabilities require only picowatts (one trillionth of a watt) of energy. Therefore, our research heralds a new era of non-silicon-based, energy-efficient, sustainable, and biocompatible computing devices, potentially transforming the landscape of AI into a domain where living organisms play a crucial role. This exploration not only paves the way for innovative computing solutions but also enriches the burgeoning field of living AI.”
The research, accepted as a full paper by the tinyML Research Symposium 2024, can be found here.
This research is supported by VistaMilk SFI Research Centre and the National Science Foundation (NSF) and is a collaborative effort by Samitha Somathilaka; Walton Institute at SETU and University of Nebraska-Lincoln; Adrian Merle Ratwatte, University of Nebraska-Lincoln; Sasitharan Balasubramaniam, Walton Institute at SETU Adjunct Professor, University of Nebraska-Lincoln; Mehmet C. Vuran, University of Nebraska-Lincoln; Witawas Srisa-an, University of Nebraska-Lincoln; and Pietro Lio, University of Cambridge.