TLDR: A new technique, neurosymbolic AI, is emerging to enhance the accuracy and reliability of generative AI by combining neural networks with symbolic reasoning. This hybrid approach aims to mitigate the ‘hallucination problem’ in AI, fostering greater trust and adoption in various sectors, including finance and code generation. Companies like Imandra and Amazon are at the forefront of implementing this technology.
A groundbreaking technique known as neurosymbolic AI is poised to revolutionize generative artificial intelligence (AI) by infusing it with mathematical precision, addressing the critical ‘hallucination problem’ that has plagued current models. This innovative approach marries the pattern recognition capabilities of generative AI’s neural networks with the rigorous, logic-based reasoning of symbolic AI, which can mathematically prove the correctness of outputs. The primary objective is to preserve the inherent flexibility of generative AI while significantly boosting the trustworthiness of its results.
According to a July 2025 PYMNTS Intelligence report, building trust is paramount for the broader adoption of both generative and agentic AI. While chief financial officers (CFOs) have shown increasing acceptance of generative AI, they remain hesitant regarding the reliability of agentic AI, underscoring the need for more dependable solutions.
Citi-backed startup Imandra is a key player in this emerging field, offering tools designed to provide ‘mathematical guarantees’ for AI-generated processes. One of its flagship products, Code Logician, is specifically engineered to tackle the proliferation of AI-generated code. When integrated into coding assistants like Cursor, Code Logician constructs a mathematical description of the AI’s code and then uses Imandra’s verification capabilities to ensure its accuracy. A significant statistic shared by Passmore indicates that ‘around 60% of AI-generated code contains bugs,’ and Code Logician has demonstrated the ability to make ‘96% of the code correct within three iterations.’ Imandra is actively expanding Code Logician’s compatibility beyond Python to include Java and COBOL, responding to strong enterprise demand for robust code migration solutions. The company is also venturing into developing agents for geometry reasoning and other complex applications.
Also Read:
- Generative AI Market Poised for Substantial Growth Driven by Diverse Industry Applications
- Addressing AI ‘Hallucinations’ in Legal Practice: The Imperative for Enhanced Human Oversight
In parallel, Amazon is also advancing neurosymbolic AI, applying it to critical areas such as warehouse robots and shopping assistants, where precision and reliability are paramount. The broader industry is also seeing related advancements, with Lambda recently announcing a new platform for ‘Agent-to-Agent testing.’ This platform provides a standardized framework for evaluating the quality of AI agents against crucial metrics like hallucinations, bias, and completeness, further emphasizing the industry’s push towards more reliable AI systems.
