Generative AI firm Galileo has debuted a tool to help businesses develop trustworthy artificial intelligence (AI) solutions.
The San Francisco-based company on Tuesday (Feb. 6) announced the release of a new retrieval augmented generation (RAG) and agent analytics solution.
As Galileo noted in a news release, RAG systems “have become increasingly popular with developers of LLMs,” or large language models.
“RAG supplements an LLM’s general knowledge with domain-specific context, so the LLM can provide domain-specific results,” the company said.
However, the release added, “the complexity of RAG systems and their many moving parts have required labor-intensive manual evaluation, and their inner workings can be somewhat of a black box for AI builders.”
Galileo said its tool changes this process “by embedding advanced insights and metrics directly into the user’s existing workflow, with easy access through an intuitive Galileo user interface,” offering visibility into each stage of the RAG workflow, and enabling rapid evaluation, error detection and iteration.
“Galileo’s RAG & Agent Analytics is a game-changer for AI practitioners building RAG-based systems who are eager to accelerate development and refine their RAG pipelines,” said Vikram Chatterji, CEO and co-founder of Galileo. “Streamlining the process is essential for AI leaders aiming to reduce costs and minimize hallucinations in AI responses.”
The launch of Galileo’s new offering comes as many companies are — as PYMNTS wrote Tuesday — “fishing with dynamite” when it comes to AI systems.
“That’s because the biggest and most impressive large language models (LLMs), including OpenAI’s GPT-4, are trained on over 1 trillion parameters, and cost hundreds of thousands of dollars a day to run,” that report said. “Using such models for daily tasks with minimal impact or low complexity, or for small-scale personal queries, is, well, a bit overkill.”
While Big Tech’s big AI models have popularized and familiarized the technology across a broad global audience, the future of AI’s commercial applications “likely lies in smaller models that have fewer parameters but perform well on specialized tasks,” PYMNTS wrote.
For AI to be “truly democratized,” that report said, it will need to be built atop smaller, more cost-efficient systems, as smaller models are what companies like OpenAI, Google and Apple are hoping to commercialize.