TBC Brings Rat-Brain Video AI to Select AWS Customers
The limited preview gives TBC a new route into AWS for software meant to make video generation faster and cheaper. Whether those gains persist in longer videos remains unresolved.
Listen to this story
The audio brief
Story brief
3 key pointsTBC’s AWS preview is less a broad launch than a distribution test: select enterprise customers can now evaluate software derived from rat-cell electrical activity inside a major cloud channel. The company says the approach accelerates existing video models by up to 5x and lowers inference costs, but has not named the open-source frontier model used for comparison. AWS is specifically watching longer clips, where...
- 01
TBC claims up to 5x faster video generation and lower inference costs versus an unidentified open-source frontier model.
- 02
The software augments existing video-generation models rather than replacing transformer architectures.
- 03
AWS is testing whether gains persist as clips become longer and harder to keep temporally consistent.
The Biological Computing Company has put an AI model derived from rat brain-cell activity into limited preview for select AWS customers. It is designed to improve video generation, bringing a lab-based approach into a major cloud channel.
The preview changes how customers can reach TBC’s software. Until now, the company offered its technology through the neocloud provider Bluesky Compute. AWS and TBC say they expect access to extend to all AWS enterprise customers soon, but have not given a date. For now, the program is limited rather than a broad enterprise rollout.
From cell signals to video models
TBC records electrical activity from rat brain cells on multi-electrode silicon arrays, then looks for computational patterns it can translate into software. The company is applying those patterns to existing video-generation models, not proposing a replacement for transformer architectures, the core design used in many current AI systems. Its pitch is an efficiency improvement for visual models customers already use.
The headline metric needs caution. TBC has not identified the open-source model used for comparison. That leaves customers without the details needed to judge how closely the benchmark matches their own workload, even as the AWS preview gives them a new way to try the software.
The longer-video test
AWS sees longer videos as the more demanding next test. A video model needs to retain what has already happened and keep details consistent as a clip runs longer. AWS says it remains unclear whether TBC’s improvements will hold when customers push the technology to those limits.
Broader enterprise availability remains an expectation, not a dated commitment. The immediate result is a limited customer test of whether software patterns drawn from living cells can lower video-generation costs without sacrificing consistency in longer output.
Sources
- wired.comAI Models Built From Rat Brains Just Got Closer to Reality
Reader comments
Newest comments first. Replies stay oldest first.