Jaxon AI Collaborates with IBM WatsonX to Tackle AI Hallucination Issues

In a rapidly evolving AI landscape, Jaxon AI stands out for its commitment to precision and reliability. Initially crafting AI systems for the U.S. Air Force, where accuracy is paramount, Jaxon AI is now venturing into the wider corporate arena. Their innovative solution, Domain-Specific AI Language (DSAIL), aims to tackle a critical issue in artificial intelligence: the occurrence of hallucinations and inaccuracies in large language models (LLMs). DSAIL, integrating IBM’s foundational models, offers a fresh strategy to enhance AI dependability.

DSAIL’s core function is to minimize AI hallucination risks. Hallucinations in AI refer to the generation of incorrect responses due to factors like inadequate training data or lack of verification. DSAIL converts natural language inputs into a binary format, subjecting them to rigorous checks, such as boolean satisfiers, to validate AI responses. This process enhances the AI’s reliability, particularly in applications requiring high trust.

A popular method to counter AI hallucinations is Retrieval Augmented Generation (RAG). This approach, used by several companies, involves accessing a knowledge base for more accurate answers. While DSAIL incorporates RAG, it extends beyond it. Outputs from RAG are further scrutinized through stringent checks before being presented as results, reducing hallucination risks.

IBM’s WatsonX foundation models are pivotal in Jaxon’s methodology. The IBM StarCoder model, particularly, is instrumental in Jaxon’s code generation phase, enabling automated initial code generation for AI projects. StarCoder, an open-source project supported by various industry players, reflects IBM’s commitment to accessible AI.

However, IBM’s WatsonX library offers a variety of code-generation tools beyond StarCoder, each tailored to specific applications. IBM has employed its models for tasks ranging from COBOL code migration to developing quantum computing applications.

In the competitive generative AI and LLM market, IBM seeks to carve out its niche. Its strategy involves aiding developers and software vendors through the IBM Build program. This initiative grants partners access to WatsonX, along with technical and marketing support. IBM’s aim is to provide stable, trusted AI foundation models, ensuring consistent pricing and performance. Rodrigues from IBM emphasizes the trustworthiness of their AI approach, underscoring the meticulous training and legal vetting of their models.