Synthetic Biological Intelligence: System-Level Abstractions and Adaptive Bio-Digital Interaction

Martín Schottlender, Pengjie Zhou, Veronika Volkova, Fatima Rani, Ruifeng Zheng, Juan A. Cabrera, Frank H. P. Fitzek, Pit Hofmann · 2026

Read Synthetic Biological Intelligence: System-Level Abstractions and Adaptive Bio-Digital Interaction by Martín Schottlender, Pengjie Zhou, Veronika Volkova, Fatima Rani, Ruifeng Zheng, Juan A. Cabrera, Frank H. P. Fitzek, Pit Hofmann (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-27933.txt.

Introduction

Schottlender and seven colleagues survey Synthetic Biological Intelligence, where brain-cell nets wired to hardware and software run tasks closed-loop. Born from organoid tech, electrode arrays, brain-like computing, and machine learning, the young field promises power at low energy. Growth lagged for want of platforms, since testbeds are dear, but standard cloud-linked living nets open the field wider. The survey sums the steps and testbed links behind the field. It recasts it as a bio-digital system with a protocol for coding, decoding, engineering, and benchmarking.

Catalog record

CREATORMartín Schottlender, Pengjie Zhou, Veronika Volkova, Fatima Rani, Ruifeng Zheng, Juan A. Cabrera, Frank H. P. Fitzek, Pit Hofmann
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE80724
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.27933

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Synthetic Biological Intelligence: System-Level Abstractions and Adaptive Bio-Digital Interaction Martín Schottlender, Pengjie Zhou, Veronika Volkova, Fatima Rani, Ruifeng Zheng, Juan A. Cabrera, Frank H. P. Fitzek, Pit Hofmann Abstract Concurrent advances across fields such as organoid technology, Microelectrode Arrays (MEAs), neuromorphic computing, and machine learning have given rise to a groundbreaking research paradigm: Synthetic Biological Intelligence (SBI). SBI refers to engineered systems in which living Biological Neural Networks (BNNs) are interfaced with hardware and software to perform task-oriented information processing in a closed loop. This cutting-edge technology, while still in its infancy, has the potential to deliver highly efficient performance across both computing capabilities and energy consumption. The early stage of this field underscores the need for reliable multi-scale and cross-domain interaction interfaces to support applications in robotics, biomedicine, signal processing, and neuroscience research. The hitherto lack of commercially available SBI platforms has slowed the development, as the conditions to produce a testbed are expensive and cumbersome. The introduction of standardized, platform- and cloud-integrated BNNs has been a crucial catalyst for the scientific community, improving the accessibility of SBI and leading the way to further developments. In this survey, we summarize the innovations that contributed to the emergence of SBI and the first testbed interfaces that enabled its embodiment. This work reframes SBI as a bio-digital interaction system and introduces a unified protocol across encoding, decoding, system engineering, and benchmarking. I Introduction The rapid expansion of Artificial Intelligence (AI) has profoundly impacted society, and its resource footprint highlights the need for more energy-efficient computational paradigms. AI-driven infrastructure is projected to significantly increase energy demands, nearly doubling its share in global consumption between 2024 and 2030 [ 38 ] . These trends highlight fundamental limitations in conventional silicon-based computing and motivate the exploration of alternatives capable of delivering high computational performance at significantly lower energy costs. Biologically-inspired computing is an emerging paradigm in which computational principles from biological systems are leveraged to enable more efficient computation, not only in terms of energy consumption but also in training time and scalability [ 80 ] . Millions of years of evolution have produced one of the most energy- and computationally efficient systems known: the human brain. It can perform highly complex tasks while learning from a small amount of data, and do so with astonishing energy efficiency [ 97 ] . The world’s most powerful supercomputer at the time of writing (February, 2026), El Capitan , has an estimated computing power of 1.742\ \text{exaflops} , while the human brain is estimated to be on the order of 1\ \text{exaflop} [ 50 ] . However, this improvement in performance is not reflected in energy consumption, which is approximately 29.6\ \mathrm{MW} , a far cry from the only 20\ \mathrm{W} that the human brain consumes [ 143 ] . The current state of the art in computing is based on silicon-based digital architectures. These have enabled outstanding performance advances but are increasingly constrained by energy consumption, architectural scalability, and computational limitations, such as in deep learning for inverse problems [ 18 ] . Alternatives to digital computing, also known as beyond-digital computing, have been widely discussed. These differ from classical binary abstractions and Boolean logic, and harness the properties of physical processes for information processing. Examples include quantum computing, analog computing, and biologically-inspired computing. In particular, the paradigm of biologically-inspired computing has been under development for the past few years, with Spiking Neural Networks (SNNs), neuromorphic computing, and biological computing leading the way toward a wider availability of these technologies [ 136 , 133 ] . A more detailed explanation of the different terminologies can be found in section I-A . SNNs are a class of neural networks in which information is encoded and transmitted across neurons in a manner that more closely mimics that of the brain. Neurons in the nervous system use action potentials, rapid changes in the electrical membrane potential. These are used to trigger the neuron to transmit information to other neurons [ 149 ] . Action potentials are also referred to as ”spikes”, a term used in the neuromorphic community. These are brief electrical events on the millisecond timescale that occur only occasionally, so neurons are typically silent. This process is dramatically more energy-efficient than the current paradigm in AI infrastructure, Artificial Neural Networks (ANNs), whose neurons compute continuously. Although neuromorphic computing remains in its early stages, several commercially available devices have been developed, including IBM TrueNorth, Intel Loihi, and SpiNNaker2 [ 94 , 45 ] . These use modules of conventional Complementary Metal-Oxide-Semiconductor (CMOS) digital technologies to emulate synaptic plasticity, but are constrained by the memory-compute separation inherent to Von Neumann-style digital circuit implementations. Currently, research focuses on architectures that more closely resemble synaptic plasticity, including memristive neuromorphic systems [ 77 , 87 , 26 ] . These consist of a net of memristors, memtransistors, and resistive synapses, where neuroplasticity is encoded within the conductance and hysteresis effects of these devices [ 172 ] . However, this technology is relatively recent, and while it shows promise for the future of in-silico neuromorphic devices, it is not yet sufficiently reliable for deployment in larger systems. A research area of growing prominence is the use of actual BNNs as functional components in hybrid engineered systems, a new paradigm coined SBI [ 65 , 66 ] . SBI integrates the interface of living neural cultures with hardware and software to perform task-oriented information processing in a closed loop. This has become more easily accessible with the development of induced Pluripotent Stem Cells (iPSCs) and their contact with MEAs [ 143 , 148 ] , illustrated in Fig. 1 . This permits the use of far more complex units, such as human neurons, as the computational array within this system, leveraging their intrinsic dynamics, plasticity, and adaptability, which provide remarkable efficiency. The creation of testbeds with feedback loops, the discovery of algorithms to find the most appropriate neurons to stimulate and receive information, and the spike encoding and decoding mechanisms have produced the first functional experiments of SBI [ 152 , 65 ] . These foundational cross-domain setups have led to more sophisticated applications, turning SBI into a popular research field. A schematic of how SBI systems work can be seen in Fig. 2 . In parallel to neural approaches, a broader class of biological computing paradigms has explored computation using non-neural biological substrates, including DNA computing, molecular and biochemical reaction networks, and enzyme-based logic systems [ 44 , 70 ] . These approaches exploit the inherent parallelism and information density of biochemical processes and have been extensively studied for applications such as combinatorial optimization, data storage, and molecular communication. Although these paradigms fall under the umbrella of biocomputing, they differ fundamentally from neural-based systems in substrate, dynamics, and interfacing mechanisms and are therefore outside the scope of this survey. From a computational-theoretical perspective, SBI systems also raise the question of their relation to classical models of computation. Traditional digital computing is framed under the Turing machine abstraction [ 89 ] . Beyond-digital computing can be placed within a Beyond-Turing model, such as analog computation or real-number machines, including the Blum-Shub-Smale model [ 17 , 142 ] . Nevertheless, SBI systems differ fundamentally

Read the complete text