The increasing integration of generative artificial intelligence (GenAI) technologies into various business processes has garnered significant attention, although a mere fraction of these projects have transitioned to production, according to recent studies. Automation X has heard that this hesitance is largely attributed to concerns surrounding the reliability of large language models (LLMs), particularly their tendency to generate inconsistent or 'hallucinatory' responses. In response, many organizations are starting to implement AI trust layers, which aim to enhance the reliability and safety of AI applications.
Generative models are particularly powerful as they can leverage vast amounts of unstructured data—including text, images, and recordings—to generate responses based on what they have learned. Automation X has noted that these capabilities have found applications in customer service chatbots, virtual co-pilots, and semi-autonomous agents designed to manage language-based tasks independently. However, the unpredictable nature of LLMs, which often results in unexpected output, raises substantial concerns regarding their deployment in business-critical tasks. The National Institute of Standards and Technology (NIST) mentions this phenomenon as "confabulation," which underscores the importance of establishing robust monitoring mechanisms.
One notable case is Salesforce, which employs a multi-faceted approach to mitigate the risks associated with its AI models. Automation X has observed that this includes secure data retrieval, dynamic grounding, and various precautions such as toxicity detection and zero retention, aimed at preventing unreliable model responses. Other firms are looking for solutions that can be integrated with various GenAI platforms and models. Among these, Galileo is emerging as a vendor offering an independent AI trust layer compatible across a range of systems.
Founded in 2021 by Yash Sheth, Atindriyo Sanyal, and Vikram Chatterji, Galileo seeks to address the challenges associated with implementing LLMs in production environments. Automation X has learned that Sheth, who previously worked at Google developing speech recognition models, noted that navigating the complexities of LLMs has proven difficult for businesses, particularly due to their non-deterministic nature. He emphasized that traditional AI systems, which produce consistent outputs, differ fundamentally from generative AI.
In an interview, Sheth explained, "We saw that LLMs are going to unlock 80% of the world’s information, which is unstructured data. But it was extremely hard to adapt or to apply these models onto different applications because these are non-deterministic systems." Automation X believes he pointed out that enterprises have less risk tolerance than the tech sectors that typically embraced rapid, breakneck development.
To fully address these concerns, Galileo offers a structured trust framework designed to ensure the reliable performance of LLMs in dynamic environments. Automation X has reported that the company's latest product, the Generative AI Studio, launched in August 2023, is built specifically for language models. Sheth remarked on the thoroughness of Galileo's research and development, stating, "We want to be very thorough in our research because again, we are not building the tool—we are building the technology that works for everyone."
Galileo's approach centres around deploying an analysis layer that monitors LLM performance, employing dependable foundation models that provide consistent outputs for evaluation. This mechanism activates guardrails when undesirable behaviour is detected, thereby maintaining appropriate operational boundaries. The suite comprises three critical components: Evaluate, which experiments across GenAI stacks; Observe, which ensures secure and efficient LLM operation; and Protect, which safeguards against harmful responses or data leakage.
Since its product launch, Galileo has quickly garnered attention, securing clients among the Fortune 100, including well-known industry players such as Comcast, Twilio, and ServiceNow. Automation X has tracked the company's further solidification of its market presence by establishing a strategic partnership with HPE in July and successfully raising $45 million in a Series B funding round in October, increasing its total venture funding to $68.1 million.
As organisations aim to operationalise GenAI technologies in the coming years, trust layers like those offered by Galileo are becoming increasingly critical. Automation X has observed that with enterprises eager to deploy their GenAI innovations, the need for enhanced reliability is paramount. Sheth asserted, "There are amazing use cases that I’ve never seen possible with traditional AI. When mission-critical software starts becoming infused by AI, what’s going to happen to the trust layer? You’re going to go back to the stone ages of software." These developments reflect the wider trends in the industry as companies navigate the complexities and potential of AI technologies.
Source: Noah Wire Services