Neural Tissues at the Computational Core of AGI

chevron-icon
RFP Proposals
Top
chevron-icon
project-presentation-img
user-profile-img
DanielRG
Project Owner

Neural Tissues at the Computational Core of AGI

Expert Rating

n/a

Overview

Artificial Intelligence is limited by the energy demands and rigidity of silicon processors. The brain, in contrast, is the most efficient and flexible processor known. Intactis.bio is creating a new computing paradigm that integrates 3D-bioprinted neural tissues, brain-machine interfaces, and silicon hardware into AI systems. We will enhance our bio-platform by evaluating and integrating novel silicon components. By thoroughly vetting computation paradigms, mapping SingularityNET’s Hyperon to our pipeline, and testing the system in-vitro, we will advance biologically grounded AGI for all. Our team’s deep experience in neurotech, hardware systems, and AI positions us to lead this frontier.

RFP Guidelines

Explore novel hardware architectures and computing paradigms for AGI

Complete & Awarded
  • Type SingularityNET RFP
  • Total RFP Funding $80,000 USD
  • Proposals 9
  • Awarded Projects 1
author-img
SingularityNET
Apr. 14, 2025

The purpose of this RFP is to identify, assess, and experiment with novel computing paradigms that could enhance AGI system performance and efficiency. By focusing on alternative architectures, this research aims to overcome computational bottlenecks in recursive reasoning, probabilistic inference, attention allocation, and large-scale knowledge representation. Bids are expected to range from $40,000 - $80,000.

Proposal Description

Our Team

Intactis Bio unites leaders across neurotech and AI. CEO Daniel Rodriguez‑Granrose, PhD, has 14 years in tissue engineering and ML. Neuroscience Lead Travis Rush, PhD, built the world's first ever phenotypically complete map of neurons. Dana Xiadani, MS (Harvard), specializes in EEG and neuromorphic FPGAs. David DeRoo, MBA, MSEE, brings 30+ years in neurotech hardware. Advisors include Jordan Christensen, ex-Recursion SVP of technology, who built the world's 30th largest supercomputer.

Company Name (if applicable)

Intactis Bio Corp

Project details

Artificial Intelligence (AI) systems are limited by the energy demands and rigidity of silicon-based processors. The brain, in contrast, is the most efficient and flexible processor known. Intactis Bio Corp. is pioneering a new computational paradigm by integrating centimeter-scale, 3D-bioprinted neural tissues, brain-machine interfaces, and silicon hardware into self-contained AI systems. By stimulating neurons and recording responses at various points throughout the tissue, we capture detailed electrophysiological signals that reflect individual neuron activity and network behavior. By encoding and decoding data into neural spike train patterns we program our tissue to spike in reliable ways. These neural spike train programs serve as the basis for biocomputation. Intactis has performed biocomputation studies in which a neural tissue was trained to control its own bioreactor for 21 days (see video attachment).

Although the computational core of a neural biocomputer lies in the neural biochip, these centimeter scale neural tissues require highly capable silicon systems to support signal encoding and decoding. To further strengthen the bridge between living neural tissue and silicon electronics, we are exploring a suite of emerging hardware approaches that speak the same “language” as our biocomputers. These differences will be evaluated and benchmarked against neural biochips and silicon architectures (GPUs, TPUs, as part of Milestones 1 & 2). 

  • Temporal logic circuits harness the timing of electronic pulses, much like neurons encode information in the millisecond gaps between spikes, to carry data and perform calculations. By leveraging controlled delays in hardware pathways, these circuits can execute complex, time-based reasoning tasks at low power, and they offer a natural fit for translating symbolic instructions into the precise spike-timing patterns our tissues understand.

  • Race logic designs turn computational problems into literal contests between signal paths, whichever pulse completes its journey first reveals the answer. Neurons compete in similar ways, with the earliest-firing cells often driving downstream decisions. By mimicking that competition in silicon, race logic can enact rapid, energy-efficient selection and optimization routines that dovetail with our neural networks’ own dynamics.

  • Analog processors compute through the continuous flow of voltages or currents rather than discrete bits. Since biological neurons themselves integrate ion currents and membrane voltages in an analog fashion, analog chips can mirror those natural processes with minimal translation overhead. This direct emulation can yield both higher fidelity and lower energy use when interfacing with living tissues.

  • In-memory computing architectures promise to collapse the usual divide between processing and storage by embedding simple operations directly where data resides. Just as synapses within neural tissue simultaneously hold and process information, these memory arrays can perform calculations “in place,” avoiding costly data transfers and slashing energy costs.

Beyond these broad architectural differences, our neural tissue platform yields unique capabilities tailored to AGI workloads. Recursive reasoning emerges from the inherent feedback loops within living tissue, eliminating the need for software-based recursion stacks; symbolic chaining is implemented through phase-aligned spike encoding, allowing sequences of meaning to propagate in lockstep rather than relying on discrete symbolic processors; and attention allocation leverages the tissue’s built-in salience detection: neurons amplify important signals organically, rather than through algorithmic attention gates.

Following hardware evaluation and integration, we will map SingularityNET's Hyperon components to our biocomputation pipeline, test the system, and share our results, thus significantly advancing beneficial AGI for all. To validate the full pipeline, we will run closed‐loop in-vitro experiments in which the tissue drives part of its own test suite. For example, a Hyperon-guided recursive reasoning task will be encoded in the FPGA, sent into the tissue via the electrode array, processed biologically, decoded back into digital form, and then fed into an analog accelerator for probabilistic inference before returning to the tissue. All stages will be timed, profiled for energy usage, and assessed for fidelity against ground-truth software simulations. By iterating on this integrated setup we will identify which combination of temporal logic, race logic, analog, or in-memory processing best complements our neural tissue core. This novel, co-designed approach not only amplifies the energy efficiency and adaptability of our living cores but also charts a new path to AGI, where authentic biological computation and SingularityNET coded electronics work in concert.

Milestones and Budget: 

To complete our tasks we will perform 4 specific milestones: 1) Evaluate hardware architectures to support our neural chip, 2) Review AGI Hardware Literature and Benchmark Biocomputation paradigms against state-of-the-art silicon machines 3) Procure Biocomputation Hardware, Integrate with our neural biochip, and test in-vitro, and 4)  Integrate and Test SingularityNET’s Hyperon Stack in a neural biocomputation context. Note that while milestone 3 is ¼ of the work, it provides a larger budget in order to purchase and integrate new hardware; it also provides a longer timeline in order to give us time to procure said equipment.

The majority of our costs associated with the grant will be dedicated towards directly procuring and incorporating updated silicon components for our biocomputer. A detailed breakdown of expenses is provided below. As we perform evaluations we will begin to purchase equipment leading up to milestone 3, where we will fully integrate our systems. 

 

Component

Estimated Cost (USD)

Updated Multi-electrode Array w/ switches (Temporal Logic Enabled, analog processing, and/or race logic design enabled)

30,000

High-speed storage (e.g. NVMe RAID, SAN)

10,000

Dedicated GPU or FPGA compute (for preprocessing/filtering, ML work)

                  10,000             

Data networking (10GbE+, switches, interconnects)

5,000

Server/workstation for control & visualization

3,000

Cooling, power, and rack infrastructure

3,000

In Lab Biology Work (Reagents, tissue culture flasks, etc.)

11,000

Total Cost

80,000

The Team: 

Developing neural-based biocomputation requires a fusion of neuroscience, bioengineering, electrical engineering, and computer science, fields that historically have operated in silos. Our team brings together world‑class expertise in each of these domains, united by a shared vision to unlock energy‑efficient, brain‑inspired computing.

Our Founder & CEO Daniel Rodriguez‑Granrose, PhD, has 14 years of hands-on experience at the intersection of tissue engineering and AI/ML. Dr. Rodriguez-Granrose pioneered AI‑driven scale‑up of organoid tissue cultures while on an NSF-GRFP fellowship. At Intactis he demonstrated stable biocomputation in centimeter‑scale tissues and self‑funded the company with $150K USD. 

Dana Xiadani, MS, is a Harvard‑trained Computational Neuroscience and specializes in EEG analysis, FPGA‑based neuromorphic systems, and real‑time neural signal fusion. Her skills are critical for translating biological signals into digital computation.

Travis Rush, PhD, led a team at Recursion Pharmaceuticals responsible for building the first comprehensive neural cell library (Recursion's neuromap). 

Jordan Christensen, is the former SVP of Technology at Recursion. Jordan offers deep experience in large‑scale machine learning, he previously built and led the use of the world’s 30th largest supercomputer.  

David DeRoo, MBA, MSEE, brings over 30 years of electrical engineering and product development experience, from microelectronic neurotech systems to co‑founding and fundraising for high‑growth startups. David has extensive hands-on experience building multi-electrode arrays and will collaborate to enable novel MEA level capabilities such as Temporal Logic, or analog processing.

Intactis Bio is supported by Altitude Lab/Recursion Pharmaceuticals state‑of‑the‑art facilities and network.

The research will benchmark biocomputing performance against traditional AI accelerators on AGI-relevant tasks, demonstrating projected 90% energy efficiency improvements while maintaining symbolic reasoning capabilities. This represents a paradigm shift from silicon-based simulation of intelligence to authentic biological computation integrated with digital AGI systems. With prior work in neural signal decoding, real-time hardware systems, and techbio AI at the largest scales, the team is uniquely qualified to pioneer this biologically grounded AGI computing paradigm



Links and references

Two of Dr. Rodriguez-Granrose's publications are provided, as well as a Link to an AGI Integration Example made by Dana Xiadani:

  1. Rodriguez-Granrose D., et al, Design of experiment (DOE) applied to artificial neural network architecture enables rapid bioprocess improvement.

  2. Rodriguez-Granrose, D., et al, Transition from static culture to stirred tank bioreactor for the allogeneic production of therapeutic discogenic cell spheres. 

Additional videos

The following URL leads you to our tech demo video. It shows the incredible scale of our biochips and progress we have made with neural tissue engineering, please watch it! Link!

Proposal Video

Not Avaliable Yet

Check back later during the Feedback & Selection period for the RFP that is proposal is applied to.

  • Total Milestones

    4

  • Total Budget

    $80,000 USD

  • Last Updated

    27 May 2025

Milestone 1 - Hardware Paths for AGI Bio-Silicon Integration

Description

In this phase, we will pinpoint the most promising hardware and software approaches to augment our living neural tissue computers and accelerate progress toward AGI. We will evaluate non-traditional paradigms including temporal logic circuits, race logic designs, analog and in-memory processors. To bridge brain and silicon, we are considering the inclusion of cutting-edge high-density electrode arrays for finer neural control, network‐attached storage (NAS) for seamless data flow, and SingluartityNET aligned hardware modules (Temporal logic circuits, Race logic designs, Analog processors, In-memory computing architectures, FPGAs) to translate symbolic commands into spike patterns in real time. Through our research, we will vet the optimal components that boost performance, energy efficiency, and adaptability. Mapping of SingularityNET’s Hyperon modules onto a bio-silicon pipeline is novel because it unites real neural processing with programmable electronics, creating an adaptable, energy-efficient route to AGI.

Deliverables

Intactis Bio Corp. will deliver a comprehensive research plan that lays out the evaluated hardware paradigms, and will detail how we intend to augment our neural tissue platform with electrode arrays, NAS solutions, and FPGAs. Importantly, it will explain that these components are candidates to be tested and that our experiments will determine which models deliver the best signal fidelity, throughput, and energy savings. By mapping SingularityNET’s Hyperon modules (PLN inference, MOSES recursion, ECAN attention) onto our bio-silicon pipeline, we’ll create repeatable tests for performance, power draw, and scalability. This deliverable not only compares new hardware to GPUs and TPUs but also shows how the best silicon systems can amplify the unique benefits of living neural networks.

Budget

$10,000 USD

Success Criterion

This milestone will be achieved when we produce a clear, actionable plan that identifies at least four computing paradigms to integrate with our neural biochip. The plan must explain our criteria for selecting the best versions of these components, focusing on signal precision, latency, and energy per operation. We must describe how we will test them in lab experiments. Success means the plan convincingly shows how integrating the right silicon systems with neural tissue yields measurable improvements over traditional accelerators in AGI-relevant tasks like recursive reasoning and dynamic attention. Finally, the document must highlight how this blended bio-silicon strategy is a novel leap toward energy-efficient, adaptive AGI systems. - Hardware Paradigms Catalog: List and rationale for four computing approaches. - Bio-Silicon Integration Map: Diagram showing novel computing integration data flow. - Hyperon Module Mapping: Outline of PLN, MOSES, and ECAN integration steps. - Test Protocols Overview: Initial benchmarking and experiment plan. - Novelty Highlighted: Bio-silicon co-design clearly justified.

Milestone 2 - Benchmarking AGI: Tissue vs. Silicon Review

Description

We will carry out a concise, practical review of the most promising non-traditional computing approaches for AGI, focusing especially on our novel neural tissue computers alongside temporal logic circuits, analog memory arrays, race-logic designs, and modern neuromorphic chips. This review will identify how each system handles key AGI challenges like step-wise reasoning, uncertainty-based inference, and dynamic attention. By comparing energy use, speed, and flexibility, we will show where living neural tissue offers real advantages over GPUs, TPUs, and purely silicon-based neuromorphic hardware. Finally, we will translate these insights into a clear set of side-by-side tests that measure raw performance, scaling potential, power draw, and the accuracy of converting symbolic instructions into biological signals. What sets our work apart is that we are not merely surveying hardware in isolation: we will directly benchmark each silicon paradigm against real neural tissue using the same tasks and metrics, creating the first side-by-side evaluation of symbolic-to-spike translation accuracy, scalability, power draw, and overall performance. By translating these insights into a clear test suite, we will establish a repeatable framework for quantitatively comparing biological and silicon approaches, revealing exactly where our biocomputer outperforms or complements traditional GPUs, TPUs, and neuromorphic designs.

Deliverables

Our literature review deliverable is a written report summarizing recent studies on each computing paradigm, with special emphasis on how living neural tissue computers translate symbolic queries into spike patterns and back again. This literature review will include a benchmarking guide that defines four core measures: task throughput, scalability, energy per operation, and signal-conversion accuracy. A comparison table will highlight how neural tissue computers stack up against GPUs, TPUs, analog in-memory devices, and digital neuromorphic chips.

Budget

$10,000 USD

Success Criterion

This milestone is successful if the literature review covers at least fifty key publications since 2020 and clearly explains in everyday language how each technology handles reasoning, inference, and attention. The benchmarking guide must include four well-defined, reproducible tests with step-by-step instructions on how to conduct benchmarking equations. The comparison table will highlight at least three areas where neural tissue computers outperform conventional accelerators in energy efficiency or biological fidelity, and it should also point out any trade-offs. - 50+ Publications Reviewed: Comprehensive coverage since 2020. - Literature Synthesis Report: Plain-language summaries of key studies. - Benchmark Criteria Guide: Definitions and measurement protocols for four metrics, throughput, scalability, energy, fidelity. - Comparison Matrix: Table contrasting neural tissue with other accelerators. - Statistical Plan: Outline of paired t-tests or ANOVA (α=0.05) for future results.

Milestone 3 - Test New Hardware and AGI Systems On Brain Chip

Description

In this phase, we will purchase our updated hardware identified in Milestones 1 & 2, code any necessary software integrations, grow iPSC derived neural cells in the laboratory, and test system performance in the laboratory. This milestone will first test basic integrations and signal processing in the platform. From there we will explore key AGI challenges such as multi-step reasoning, uncertainty-driven inference, and dynamic attention by encoding tasks into spike-timing patterns, letting the tissue and silicon co-process them, and decoding the results. This work is novel because it goes beyond proof-of-concept: we will be the first to measure how each specific hardware paradigm shapes live biocomputation and influences AGI architectures under real-world lab conditions.

Deliverables

We will deliver a concise presentation of our early results, highlighting both simulated runs and preliminary lab data from our integrated bio-silicon platform. This report will show how tasks are encoded into our FPGA-driven pulse patterns, how the neural tissue responds via the electrode array, and how we route outputs through analog or in-memory modules for further inference. Key performance metrics include speed of execution, energy per operation, and fidelity of symbol-to-spike conversion. These metrics will be analyzed against initial benchmarks. Accompanying these findings, we will update our system documentation to reflect real-world behavior, noting any unexpected interactions and refining our hardware-software interfaces based on learnings.

Budget

$50,000 USD

Success Criterion

This milestone will be deemed successful if we demonstrate a working hybrid experiment, either in simulation or on the benchtop, where a defined AGI-style task is sent into our neural biochip, co-processed with a chosen silicon accelerator, and then decoded back into symbolic form. We must show clear data on at least two AGI-relevant operations (for example, a three-step recursive query and a probabilistic inference task), reporting execution times, energy use, and conversion accuracy within 15% of our targets. - Hardware Procurement: Hardware on-site and set-up at Intactis Lab. - Reliable AGI Prototype: ≥5 successful simulated runs on AGI-style task, ≥90% task accuracy. - In-vitro test: 1 in-vitro run to test AGI style task integration with living neuron biochip. Cost & Integration Clear: Budget and setup complexity documented.

Milestone 4 - End-to-End AGI Task on Neural-Silicon Unit

Description

In this final phase, we will bring together all of our insights, hardware tests, and software frameworks to deliver a comprehensive evaluation of the most promising biocomputing architectures. Building on our earlier work, we will compare how each silicon‐enhanced neural tissue setup fared in real AGI‐style tasks, looking at speed, energy use, reliability, and ease of integration. We will assess not only technical performance but also practical factors like component cost, lab setup complexity, and long‐term scalability. This milestone culminates in a working prototype or proof-of-concept system, where one standout hardware paradigm will be showcased running a clear AGI use case, such as a multi-step reasoning problem with adaptive attention. This live demonstration will underline the novelty of our bio-silicon co-design, proving that living networks and programmable electronics can together handle real-world, symbol-based intelligence tasks.

Deliverables

We will hand over two key items. First, a final report that lays out side-by-side comparisons of every tested architecture, complete with charts and narrative summaries covering feasibility, energy and speed metrics, component and maintenance costs, and how smoothly each system integrates into our existing lab pipeline. This report will highlight trade-offs and recommend the best path forward for scaling up. Second, we will deliver a working prototype or demo setup: a self-contained biocomputer that uses our chosen hardware paradigm integrated with 3D-printed neural tissue and supporting silicon modules. In this proof-of-concept, the system will execute an AGI-relevant task, from encoding a recursive query into spike patterns, to applying probabilistic inference, to dynamically shifting attention based on results, all in real time. A short video walkthrough and accompanying user guide will document the demonstration, ensuring others can reproduce and build upon our success.

Budget

$10,000 USD

Success Criterion

We will consider this milestone achieved when we submit a final report that clearly shows which architecture delivers the best balance of performance, energy efficiency, cost, and integration ease. The report must compare at least three complete bio-silicon configurations, quantify their strengths and weaknesses, and offer a justified recommendation. The prototype demonstration must run an end-to-end AGI-style task, with measured results that meet or exceed our targets for execution time and energy per operation. The prototype should operate consistently over multiple trials, proving reliability, and should be documented in a reproducible guide. If we can show that our novel integration of living neural tissue and a selected silicon paradigm not only works but offers clear advantages over traditional accelerators, then Milestone 4 will be successfully completed. -Demonstrate An AGI Relevant Task: Show end-to-end results for an AGI-style tasks. -Meet Performance Targets: Execution time and energy use within 15% of our goals. -Prove Statistical Gains: Improvements over a digital-only control with t-test, p < 0.05. -Update All Protocols: Complete lab procedures and interface docs revised. -On Time & Budget: Total project within 6 months and under $80 K USD.

Join the Discussion (0)

Expert Ratings

Reviews & Ratings

    No Reviews Avaliable

    Check back later by refreshing the page.

Welcome to our website!

Nice to meet you! If you have any question about our services, feel free to contact us.