The Biological Computing Co. (TBC) has announced a partnership with Amazon Web Services (AWS) to deliver what it describes as the world’s first neurally optimised AI video generation model. The San Francisco-based company develops AI optimisation methods using insights and algorithms derived from real neurons. The model is now available in early access.
TBC’s optimised image-to-video model builds on an open-source video generation model, reducing inference costs by 80% and increasing generation speed fivefold while also improving video quality compared with the base model. It achieves this without requiring changes to existing workflows or infrastructure. The model uses TBC’s lightweight proprietary software adapter, developed from measurements of living neural activity.
Brain-inspired AI is a sub-domain of AI that draws on principles from biological neural systems. Rising global compute demand is increasing interest in the field. TBC’s approach leverages the brain’s efficiency as an information-processing system, building on the idea that discoveries from biological neural networks can inform faster and less computationally intensive AI models.
“We’re turning discoveries from real neurons into faster, cheaper AI for businesses and creators. More importantly, biology gives us a fundamentally different engine for discovering better optimization strategies over time,” says Dr. Alex Ksendzovsky, CEO and co-founder of TBC.
The AWS partnership gives TBC a route to apply this approach across more of the AI stack. TBC plans to adapt its model to run on Trainium, AWS’s family of purpose-built AI chips. At the same time, it can deploy its models through Amazon SageMaker AI and distribute them commercially through AWS Marketplace.
“By building on the AWS AI stack and leveraging go-to-market channels like AWS Marketplace, TBC can move from discovery to commercial scale at startup speed. This is exactly the kind of bold, forward-leaning innovation we love to support at AWS,” said Jason Bennett, Vice President and Global Head of Startups and Venture Capital at AWS.
For companies deploying generative AI, greater inference efficiency can increase how much can be done with existing compute resources. “We need more infrastructure, but we also need to make every unit of compute dramatically more productive,” explains Dr. Jon Pomeraniec, co-founder and COO of TBC.
Beyond the video model’s performance gains, the partnership provides an early commercial testbed of TBC’s broader approach to biological computing. The company takes computational problems in AI and designs targeted experiments in which living neurons are presented with analogous challenges, using the resulting neural activity to identify principles that can potentially be translated back into AI systems.
Each experiment adds to TBC’s understanding of biological computation and its library of potential algorithms. The larger ambition is to build a repeatable discovery process that can extend beyond a single model or optimisation technique, bringing insights derived from biological neural networks into conventional silicon-based AI.
That work could also have relevance beyond AI optimisation. A deeper understanding of how biological neural networks process information may provide insights for areas of neurotechnology including brain-computer interfaces and neuroprosthetics, where interpreting and interacting with neural activity remains a central challenge.
TBC plans to scale this discovery process across more AI models and architectures before expanding into a wider range of AI workloads. The company has opened early access to what it describes as the world’s first AI model improved using computational principles derived from real neurons, with broader commercial availability to follow.
.svg)
