Artificial intelligence (AI) has become essential in the ever-evolving field of data science to derive meaningful insights from complex datasets. YouTube content creator and AI enthusiast Andy Stapleton recently took a closer look at the capabilities of three leading AI tools: Julius AI, Vizly, and the latest version of ChatGPT. His goal was to determine which tools excel at data analysis and understand their limitations. Here’s what he discovered:
The experiment begins
Stapleton began his experiments with a simple dataset: public health data. He fed each AI tool the same data and gave them the same prompt: “Here’s some public health data. Please provide some insight into what this data shows, such as graphs or other visualizations that you think would be useful.”
Julius AI was put to the test first, quickly generating Python code to analyze the dataset and creating visualizations that included distributions of hospital codes, admission types, illness severity, and length of stay. “Julius AI provided a comprehensive initial analysis,” Stapleton notes. “It’s clear that it can handle large datasets and generate useful insights efficiently.”
Tool Comparison
Stapleton then tested Vizly with the same dataset and prompt. Vizly created a similar visualization but offered its own overview of public health data analysis. “Vizly chose slightly different parameters for their analysis,” Stapleton said. “One notable feature is the interactive charts, which reveal additional information when users hover over data points.”
Eventually, Stapleton turned to the latest version of ChatGPT, which not only generated visualizations but also provided analysis plans and interactive graphs. “The interactivity of ChatGPT’s visualizations is what sets it apart,” Stapleton says. “It lets you explore the data in a more dynamic way, which is really helpful for doing deeper analysis.”
Going deeper
Stapleton then asked each tool to show the distribution of hospital stays broken down by time period. All three tools performed well, but Vizly’s interactivity again stood out. “Vizly’s graphs were the most user-friendly,” Stapleton said. “You could zoom in and really explore the data.”
For the next test, Stapleton introduced a more challenging dataset from his doctoral research on organic photovoltaic (OPV) devices. This dataset was unstructured and contained metadata and raw data. Julius AI impressed, despite the complexity of the data, by correctly identifying and plotting the IV curves for the OPV devices. “Julius AI’s ability to self-correct and find the data it needed was impressive,” Stapleton said.
Vizly struggled initially, but after several attempts, it was eventually able to identify the IV curve data. Meanwhile, ChatGPT quickly processed the unstructured data and accurately plotted the IV curve, even calculating the efficiency of the OPV devices. “ChatGPT’s inference capabilities are excellent,” Stapleton concluded. “It can handle complex data sets with ease.”
Testing the limits
To push the limits further, Stapleton tested the AI tool on images of silver nanowires and single-walled carbon nanotubes. Both Julius AI and Vizly attempted edge detection, with mixed results. ChatGPT was not able to directly measure the diameter of the nanowires, but it did provide valuable guidance on using other tools, such as ImageJ, to make accurate measurements. “Having ChatGPT provide actionable advice is a huge advantage,” Stapleton noted.
Final thoughts
After extensive testing, Stapleton found both Julius AI and ChatGPT to be the most effective tools for data analysis. “For anyone working with large, complex datasets, Julius AI and ChatGPT are invaluable,” he said. “They complement each other perfectly, making data analysis more accessible and efficient than ever before.”
Stapleton’s in-depth look at AI tools for data analysis highlights the transformative potential of these technologies: As AI continues to evolve, tools like Julius AI, Vizly, and ChatGPT will play a key role in helping researchers, analysts, and businesses extract new insights from their data.
