The Massachusetts Institute of Technology (MIT) has been using large language models (LLMs) to detect errors in data recorded over time without the need for training.
MIT researchers hope the new method could some day help alert technicians to potential problems in equipment like wind turbines or satellites – which is currently a time-consuming and arduous task.
Go deeper with GlobalData
Engineers often streamline this complex problem using deep-learning models to detect anomalies in data take over time, however training a deep-learning model to analyse time-series data can be difficult and costly.
There are also concerns that the model would need to be retrained after deployment and wind farm operators may lack the necessary machine-learning expertise to do so.
In a new study, researchers developed a framework called SigLLM, which includes a component that converts time-series data into text-based inputs an LLM can process, meaning errors can be more easily identified.
Sarah Alnegheimish, an electrical engineering and computer science (EECS) graduate student and lead author of a paper on SigLLM said: “Since this is just the first iteration, we didn’t expect to get there from the first go, but these results show that there’s an opportunity here to leverage LLMs for complex anomaly detection tasks.”
How well do you really know your competitors?
Access the most comprehensive Company Profiles on the market, powered by GlobalData. Save hours of research. Gain competitive edge.
Thank you!
Your download email will arrive shortly
Not ready to buy yet? Download a free sample
We are confident about the unique quality of our Company Profiles. However, we want you to make the most beneficial decision for your business, so we offer a free sample that you can download by submitting the below form
By GlobalDataThe LLM can also be used to forecast future time-series data points which would enable making predictions and remedying faults more easily in the long-term.
With comparisons being drawn between the LLM and other suitable AI solutions, it is hoped that the framework with improvements could flag potential problems in equipment like heavy machinery or satellites before they occur.
Two approaches for anomaly detection were used in the research project, firstly ‘Prompter’ – an approach where users fed prepared data into the model and prompted it to locate anomalous values.
Secondly the MIT researchers used the approach they named ‘Detector’ whereby the LLM was used as a forecaster to predict the next value in a time series data set, the researchers then compared the predicted value to the actual value and a large discrepancy suggests that the real value is likely an anomaly.
In practice, Detector performed better than Prompter, which generated many false positives.
When both approaches were compared to current techniques Detector outperformed transformer-based AI models on seven of the 11 datasets that were evaluated, despite the LLM requiring no training or fine-tuning.
