In recent years, artificial intelligence advancements have occurred across multiple disciplines. However, a lack of communication between domain experts and complex AI systems have posed challenges, especially in fields like biology, healthcare, and business. Large language models (LLMs) such as GPT-3 and GPT-4 have made significant strides in understanding, generating, and utilizing natural language, powering the development of agent models for a multitude of applications.
Researchers from the Hong Kong Polytechnic University have presented a solution to address this challenge named LAMBDA. This open-source, code-free multi-agent data analysis system enables more straightforward interaction between AI capabilities and domain knowledge. It breaks coding barriers, blends human intelligence with AI, and restructures data science education. LAMBDA’s key attributes include reliability, i.e., it can solve data analysis tasks consistently and accurately, and portability, i.e., compatibility with various LLMs.
LAMBDA consists of two agents, the “programmer” and “inspector.” The programmer creates code according to user instructions and datasets, and the generated code is executed on the host system. If the code faces errors, the inspector proposes improvements, which the programmer then uses to rectify the code, after which the code is re-submitted for reevaluation.
Experimental results revealed that LAMBDA performed relatively well in machine learning tasks, with high accuracy rates on various datasets for classification tasks and low Mean Squared Errors for regression tasks. This system successfully overcame coding obstacles with no human intervention while linking data science to non-coding skilled human experts.
In conclusion, the researchers introduced a novel, open-source, code-free method called LAMBDA for data science applications, combining human intelligence with AI. Experimental results demonstrated its effectiveness in data analysis tasks. The researchers suggest that future improvements may include planning and reasoning techniques. The system can bridge the gap between data science and domain experts who lack coding skills, potentially leading to more developments and findings in the field. This innovation has the potential to make data science and data analysis more accessible to domain experts who lack advanced coding skills, fostering more innovation and discovery in the future.
