Not so long ago, ChatGPT almost single-handedly broadened the adoption rates and general awareness of LLMs. Shortly thereafter, everyone from enterprise users to consumers were clamoring to avail themselves of its capacity to answer questions and generate human-like textual responses about virtually any subject.
However, contemporary applications of ChatGPT indicate that the model’s summarization, question-answering, and textual generation capabilities (according to user-defined lengths and tones of voice) were only the beginning. Today, it’s able to apply some of these same capabilities to images, facets of Business Intelligence, and recurring vertical specific use cases in education and the legal field.
As such, ChatGPT (and other LLMs) is becoming an increasingly valuable resource in an enlarging number of applications. Its capital advantage, of course, is that it is well versed in the entirety of the internet. However, caveats about hallucinations and inaccuracies still persist and, likely, will never entirely dissipate.
“This should get pretty technical,” commented Emily Vilardi, Foxit Product Marketing Manager, about ChatGPT’s prowess in some of the aforementioned applications. “Obviously, with all AI models, there are warnings about accuracy. But for the most part, questions that should be asked, it should be able to accurately answer.”
Image capabilities
One of the more impressive aspects of employing ChatGPT for use cases outside of its quintessential question-answering and dynamic language generation for users’ topics of choice pertains to image data. Some vendors have implemented bots so its question-answering capabilities hone in on images, which Vilardi characterized as the ability to “chat with images.” In much the same way that one can upload documents and have the model answer questions related to them, users can also upload any variety of images and ask the bot about them.
The specificity of the digital agent’s responses is no less so for the latter than it is for the former. “You can ask any question you want and get an instant response,” Vilardi mentioned. “For example, if you’re in the healthcare space, a technician can upload any MRI images or scans and put in ‘what bone is broke,’ and it gives you the response.” One way organizations can access such functionality is through Foxit AI, an intelligent document management platform.
BI applicability
Employing ChatGPT to analyze images and let users interact with the bot about them in natural language is contiguous to certain aspects of BI. Many vendors in this space incorporate LLMs for semantic search and generating conversational responses about textual and non-textual data. However, Chat-GPT can support BI use cases by assessing data visualizations and answering ad-hoc questions about their content. “Another big use case is with images of charts and graphs and things of that nature, to kind of pull out the more business or financials from that,” Vilardi remarked.
Given the inundation of unstructured data organizations are presently contending with, adoption rates for this use case can potentially equal, if not exceed, that of conventional textual applications of ChatGPT. “If you just took a screenshot of a graph you found and uploaded it, you could get the information of different financials, if it were a business graph,” Vilardi denoted. For example, organizations might upload analytics results, presented as data visualizations of clickstream analysis or social media feeds, and get pointed explanations about the business worth of those images.
Advanced textual mechanisms
In addition to their capacity to analyze and generate responses about images, contemporary ChatGPT applications also offer an array of advanced textual functionality. Some of it, including that which allows users to readily clarify the meaning of abstruse passages in contracts and other documents, are particularly relevant for legal and education use cases.
Such features are inherently “different than a summary,” Vilardi pointed out. “The use case here is more, say, signing a contract and you want to better understand what it’s actually saying. This is largely used in the legal space, if there’s a lot of legal jargon and you want to better explain to your client what’s included or what certain terms mean.” Similarly, students can apply such a construct to sections of a textbook or to understand the connotations of specific terms in them.
Horizontal and vertical expansion
Modern applications of ChatGPT are increasing in utility for image data, BI deployments, and industry-specific processes. This fact is noteworthy because it implies that, although the bot was initially unveiled at the end of 2022, its enterprise value is still increasing. Vendors are still determining the most productive ways to use this tool, and the LLM idiom in general. As such, the scope and depth of this approach will likely become more sizable as the years progress.
