Galileoa pioneer in the sector of generative AI for enterprises, has presented Galileo Lunaa groundbreaking suite of Evaluation Foundation Models (EFMs) that guarantees to remodel the way in which firms evaluate their GenAI systems. With Luna, Galileo goals to deal with the critical challenges of speed, cost and accuracy which have hindered the widespread adoption of generative AI in production environments.
“Galileo developed Luna to deal with the constraints of current GenAI evaluation methods, that are slow, expensive and sometimes inaccurate,” said Vikram Chatterji, co-founder and CEO of Galileo, in an interview with VentureBeat. “The motivation got here from the necessity for low-cost, highly accurate, and ultra-low-latency evaluations in production environments.”
The development of Luna represents a big milestone for Galileo, which has been on the forefront of enterprise GenAI since its inception in early 2021. The company's commitment to pushing the boundaries of AI evaluation is obvious within the nearly year-long intensive R&D process that led to the event of Luna.
Special models redefine speed, cost and accuracy
At the center of Luna's innovation are its purpose-built small language models which can be rigorously tailored to specific evaluation tasks comparable to hallucination detection, context quality assessment, data leak prevention, and malicious prompt identification. This specialized design enables Luna to deliver unprecedented performance in three key areas: speed, cost, and accuracy.
“Luna outperforms GPT-3.5 by way of speed, cost, and accuracy through several innovations,” explained Chatterji. “Luna uses specially designed small language models tailored to specific evaluation tasks, significantly reducing computational effort and value. This design decision enables evaluations which can be 97% cheaper and 11 times faster than those performed with GPT-3.5.”
But it's not nearly speed and value. Luna also offers industry-leading accuracy, outperforming previous methods by as much as 20% in detecting hallucinations, quick injections, personally identifiable information (PII), and more. “Multi-headed small language models and advanced techniques like intelligent chunking ensure Luna models retain context higher and supply more accurate assessments,” Chatterji added.
Revolutionizing evaluation without ground truth data sets
One of probably the most notable elements of Luna is its ability to operate without traditional ground truth datasets. By leveraging pre-trained evaluation models tuned to varied domain-specific datasets, Luna eliminates the time-consuming and dear technique of creating custom test sets. This innovation streamlines the evaluation process and reduces the reliance on large-scale human-generated data.
The potential uses of Luna are many. Chatterji particularly emphasized this system's importance in industries that require high reliability and speed in evaluating AI. “Luna is especially powerful in large-scale enterprise applications where volume and throughput are required (i.e. hundreds of thousands of queries monthly). We see Fortune 100 firms in healthcare, finance and telecommunications finding Luna particularly useful,” he said.
Adaptation and continuous development in view of the rapid progress in GenAI
Use cases range from real-time monitoring of AI outputs and detecting hallucinations in AI-generated content to making sure the security and quality of chatbot interactions. And with Galileo's Fine Tune product, Luna might be customized to specific customer requirements, achieving accuracy levels of 95% or more for critical tasks in industries comparable to pharmaceuticals and financial services.
As the generative AI landscape continues to evolve rapidly, Galileo stays committed to staying on the forefront of innovation. Chatterji emphasized that Luna will scale in three key ways: by expanding support for more forms of evaluation tasks, constantly improving accuracy, and further reducing cost and latency.
“Galileo is committed to pushing the boundaries of what is feasible in AI evaluation and helping enterprises bring trustworthy AI into production,” said Chatterji. “As the generative AI landscape continues to evolve, Galileo stays committed to providing its customers with cutting-edge evaluation capabilities that make AI practical for enterprises and construct trust with consumers.”
With the launch of Luna, Galileo has cemented its position because the leading provider of enterprise GenAI evaluations. More and more firms wish to leverage the ability of generative AI. Luna's ability to deliver fast, cost-effective and accurate assessments shall be a critical consider driving widespread adoption and unlocking the total potential of this transformative technology.
