DSAIL (Domain Specific AI Language) is the core technology that Jaxon AI is expanding into a wider enterprise market. This technology seeks to address the illusion and inaccuracy challenges in artificial intelligence by combining IBM Watson's underlying models. DSAIL operates with the aim of minimizing the risk of AI illusions.

Illusions occur when an AI system produces an inaccurate response to a query. Inaccuracy can be caused by various factors, such as incomplete training data and lack of validation.

The purpose of DSAIL is to help reduce the risk of illusions. Cohen explains that DSAIL accepts natural language input and converts it into binary language format. It then undergoes a series of checks and balances, such as Boolean satisfiability, to ensure that the AI response meets all constraints before being returned. This is done to limit uncertainty and improve the credibility of the AI system in applications.

One common method used by multiple vendors to reduce illusions is Retrieval-Augmented Generation (RAG). In the RAG model, LLM also has access to knowledge bases to help obtain accurate answers.

Cohen explains that RAG is a technique used by DSAIL to address illusion problems, but it is only part of the method. He points out that when using DSAIL technology, the output generated by RAG technology still needs to undergo a series of checks before being returned to the user, further limiting illusions.

IBM's Watsonx serves as the cornerstone of Jaxon AI's system.

Jaxon uses models from IBM's Watsonx model library as building blocks for its AI system.

Cohen explains that IBM's StarCoder model is specifically used for the code generation step in Jaxon AI. Jaxon utilizes StarCoder's capabilities to automatically generate initial code for AI projects based on collected designs and requirements, as a step in Jaxon's overall approach to building customized AI systems.

StarCoder LLM is an open-source project launched in May, with support from ServiceNow and Hugging Face. Savio Rodrigues, IBM's Vice President of Ecosystem Engineering and Developer Advocacy, said that IBM is actually one of the founding contributors to the StarCoder project. He also noted that IBM is working closely with Hugging Face to bring open models to enterprise users.

It is clear that IBM has multiple code generation LLM tools in its Watsonx library. While StarCoder has broad capabilities, IBM's own models focus on specific use cases. IBM has already used its own code generation LLM to assist with COBOL code migration and building quantum computing applications.

IBM is building an ecosystem plan to embed Watsonx into software vendor tools.

The AI and LLM technology market is a highly competitive field, including major players such as OpenAI, Microsoft, Google, and Amazon Web Services (AWS).

IBM is looking for opportunities to enter the market, particularly by helping developers and independent software vendors (ISVs) like Jaxon AI through a project called IBM Build.

Rodrigues explains that IBM Build provides partners with access to Watsonx, technical support, and marketing support. The overall goal is to provide organizations with reliable and trustworthy AI foundational models with consistent pricing, performance, and availability.

"We know that our customers trust IBM's approach to AI, from the way we train models to the legal checks we perform," Rodrigues said.