Recent studies by researchers at West Virginia University (WVU) are shining a beaming spotlight on the latest addition to the ChatGPT universe, the Code Interpreter plugin.
But. The findings present a dual-edged sword.
While underscoring its immense potential for STEM education, it highlights the limitations in the realms of bioinformatics and biomedical research.
Unlocking STEM’s Potential
Launched in December 2022, ChatGPT with its new plugin has swiftly captivated diverse sectors, from corporations to educators. Gangqing “Michael” Hu, a prominent voice in the Department of Microbiology, Immunology and Cell Biology at the WVU School of Medicine, recognizes the boon this development has been for education.
“Code Interpreter makes coding in the STEM fields more accessible,” he observed.
The tangible benefits are hard to ignore. The plugin democratises the world of coding, ensuring that those even without a background in science can easily delve into computer programming.
Furthermore, it’s cost-effective, nurturing a newfound enthusiasm among students to venture into data analysis and expand their learning horizons.
However, with every rose comes its thorn. The onus remains on the users to correctly interpret data and discern the veracity of the results. Moreover, a deep understanding of interaction with the chatbot becomes paramount.
Bioinformatics: A Bridge Yet to be Crossed
The realm of bioinformatics, where biology gracefully waltzes with computer science, expressed palpable anticipation for the Code Interpreter plugin.
They hoped for a tool that would simplify complex processes. Regrettably, the current iteration doesn’t quite fulfil all expectations.
Hu’s team, comprising experts from diverse universities, ascertained the plugin’s inability to cater to bioinformatics needs entirely. They flagged significant concerns: the absence of internet access to download genome data, the plugin’s limitation to the Python programming language, and a lack of software packages tailored for bioinformatics.
The absence of parallel processing further exacerbates its inefficiency with large datasets.
Yet, amidst the critiques lies a silver lining. The Code Interpreter offers a formidable solution to a prevalent challenge – ChatGPT’s hallucinations. Hu explained, “For questions that can be addressed through coding, the code itself serves as the source or citation. That is a significant step forward.”
The Path Forward
Despite the shortcomings, the spirit of innovation remains unbridled.
Suggestions for augmenting the Code Interpreter abound, from enhancing storage capacities to incorporating stringent privacy and security protocols, compliant with regulations like HIPAA.
Hu, ever the optimist, envisions a future where these impediments are mere footnotes in the annals of AI history. “AI is a fast-moving field. I hope by that time OpenAI may overcome some of the limitations,” he commented, already planning to introduce the plugin in his upcoming classes.
With experts like Hu relentlessly pushing boundaries and ushering in advancements, the promise of AI in diverse fields remains undiminished.
As Hu himself succinctly puts it, “There are certainly many other innovative uses awaiting to be discovered.”
