Literal Labs Achieves 50x AI Performance Boost with ‘Tsetlin’ Approach

Literal Labs, a cutting-edge technology company based in Newcastle-upon-Tyne, England, has made significant strides in AI performance by leveraging the power of the Tsetlin machine. The company recently conducted a study measuring AI performance on MLCommons' MLPerf Tiny Benchmark using the ToyADMOS dataset, which is designed to identify irregularities in audio data. This capability is crucial for spotting anomalies in audio, a key aspect of preventative maintenance.

The Tsetlin machine-based AI software model developed by Literal Labs was tested on a standard off-the-shelf ARM-based 32-bit microcontroller. The results of the study revealed that Literal Labs' model outperformed the current state-of-the-art AI software models on the same hardware. The company reported that their model consumed 52 times less energy and provided accurate inference results 54 times faster than existing solutions.

Founded in 2023 by University of Newcastle academics Professor Alex Yakovlev and Professor Rishad Shafik, Literal Labs has quickly gained recognition for its innovative approach to AI development. The Tsetlin machine, a pattern-learning automaton based on propositional logic, offers a more compact alternative to traditional neural network models. This efficiency makes it well-suited for edge AI applications on cost-effective, low-power microcontrollers.

Literal Labs combines Tsetlin machines with voting algorithms, binarization, and compression techniques to build advanced software models. By utilizing this approach, the company achieved a code size of 7.29 kilobytes in the benchmark study, resulting in a remarkable 50-fold performance improvement. The company's dedication to solving industry-critical problems while minimizing costs and environmental impact has positioned them as a leader in the field.

Leon Fedden, the Chief Technology Officer of Literal Labs, emphasized the significance of their findings, stating, "This breakthrough is game-changing for predictive maintenance in sectors such as manufacturing and logistics, where downtime is costly, and efficiency is paramount." Literal Labs has made its whitepaper detailing the benchmarking study available on its website, showcasing the company's commitment to transparency and innovation.