GPUs go biological: BBB unveils Bionode, lab-grown, living neuron compute for AI applications

Graphics Processing Unit (GPU) is made Nvidia,,,,, AMD and sima.aino longer the only way to train and deploy AI.

Bio black box (BBB)a startup created by Baltimore, developed a new AI hardware category, and it emerged with its Bionode platform, a computing system that integrates living, lab-grown neurons with traditional processors.

The company has been operating quietly while filing patents and refining its technology, and it believes its biological computing approach – new neurons are specifically designed for new neurons that use donor human stem cells and mouse-derived cells as computer chips that can provide low-power, adaptable alternatives to conventional GPUS.

“In the past 20 years, three separate fields (biology, hardware and computing tools) have developed to the point of biological computing,” said Alex Ksendzovsky, co-founder and CEO of BBB in a video call interview with VentureBeat.

A member NVIDIA’s Inception IncubatorBBB positioned itself as the advancement and enhancement of major silicon AI chips produced by NVIDIA and others.

By leveraging the ability of neurons to reconnect the body, the company aims to reduce energy costs, improve processing efficiency and accelerate AI model training – promoting increasingly urgent as AI adoption expands.

Despite the incredible premise, this is not science fiction: BBB’s neural chips have powered customers with computer vision and LLM, and have held talks with two partners to license their technology for computer vision applications, although the company refuses to designate its customers and partners with confidentiality agreements as specific customers and partners. It also accepts inquiries from potential partners and clients website.

Mixed biology and hardware

At the heart of the BBB approach is the Bionode platform, which uses wired lab-grown neurons to be connected to the computing system.

“We use multiple models,” Ksendzovsky told me. “One of these models comes from rat cells, and one of them comes from human stem cells that are actually converted into neurons.”

The co-founder said “thousands of” integrated into dishes containing 4,096 electrodes, which form the basis of a Bionode chip. He also said they lived for more than a year before they could be replaced.

The idea was to leverage the natural adaptability of neurons to AI processing to create a hybrid computing system that is different from today’s transistor-based rigid chips.

Ksendzovsky, who has worked with neurons on electrodes since 2005, initially entertained the idea of ​​using them to predict the stock market. His mentor, Dr. Steve Porterthis idea was refuted at the time.

“Why don’t we use neurons to predict the stock market so that we can all be rich?” Ksendzovsky recalls asking Potter, which was unrealistic for him to ridicule it. “At the time, he was right,” Kendzovsky admitted.

Since then, improvements in electrode technology, computing tools and neuronal life span have made biological computing feasible. “Bio networks have evolved into the most efficient computing system ever,” Ksendzovsky explained.

This setup offers two direct advantages:

• More effective computer vision: Bionode has been tested as a preprocessing layer for AI classification tasks, reducing inference time and GPU power consumption.

• Accelerated Big Speech Model (LLM) training: Unlike GPUs that require frequent retraining cycles, neurons can adapt at any time. This may significantly reduce the time and energy required to update the LLM, thus addressing the critical bottleneck in AI scaling.

“One of our biggest breakthroughs is to use biological networks to train large language models (LLMSs) more efficiently, thus reducing the large amount of energy consumption required today,” Ksendzovsky said.

Build a viable living GPU with the help of NVIDIA

NVIDIA’s GPUs have played a role in the rapid development of AI, but their high energy consumption and increased cost have attracted attention to scalability.

BBB sees an opportunity to introduce more effective alternatives when running in NVIDIA’s ecosystem.

“We do not consider ourselves to be a direct competitor to Nvidia in the near future, at least,” Ksendzovsky noted. “Biocompute and silicon computing will coexist.

In fact, according to the co-founder, “we can use our biological networks to enhance and improve silicon-based AI models, making them more accurate and energy-efficient.”

He believes that the long-term vision of AI hardware will be a modular ecosystem, with biological computing, silicon chips and even quantum computing playing a role.

“The future of computing will be a modular ecosystem, with traditional silicon, biological computing and quantum computing each playing a role based on its strengths,” he said.

Although BBB has not disclosed a commercial release date, the company will move from Baltimore, Maryland to the Bay Area as it prepares to expand its technology.

The future of hybrid AI processing

While silicon-based GPUs are still the industry standard, BBB’s fragment concept outlines the future of AI hardware no longer being limited to transistors and circuits.

The ability of dynamic reconfiguration of neurons can enable AI systems that are more energy-efficient, adaptable, and capable of continuous learning.

“We have applied biological computing to computer vision,” Ksendzovsky said.

In addition to improving efficiency, BBB also believes that its biological approach can provide a deeper understanding of how AI models process data.

“We built a closed-loop system that allows neurons to reconnect, thereby improving the efficiency and accuracy of AI tasks,” he explained.

Despite its potential, Kendzovsky acknowledged that moral considerations would be an ongoing discussion. BBB has worked with ethicists and regulatory experts to ensure the responsible development of its technology.

He stressed: “We don’t need millions of neurons to process the entire environment like the brain.

BBB dares to bet that living tissue is not only silicon, but also the key to AI’s next leap.