The Capacity of ChatGPT and Alternative Large Language Models in Identifying Fabricated News. An Analysis!

Generating texts akin to human-written ones and performing essential language-based functions is a technique called Large Language Models (LLMs), which progressed from the natural language processing technique (NLP). Upon the advent of Chat GPT, an OpenAI LLM, such models have become publicly accessible and highly sought-after. LLMs have proven capacity to generate essays, programming code, coherent writing, and define complex terminology. According to a new research, this ability of theirs has paved a new way for debunking fabricated news, false information and finding the solution to contemporary problems of our world.

This technology is beneficial, as was deemed by a researcher named Kevin Matthe Caramancion of the University of Wisconsin-Stout, who aimed his study at figuring out the extent to which the most up-to-date LLMs could identify the authenticity and credibility of provided news. In an online world overflowing with varied opinions and swift distribution of tampered news stories, the need to draw out what’s real and what is fake must be emphasized and can be challenging. However, Caramancion’s findings showcased an overview of LLMs future in tackling false online data and fake news.

In an interview with a media outlet, Caramancion expressed that his interest in this research stemmed from the general question regarding the magnitude to which LLMs could detect fake news. His target was to be thorough and concise in their capability to pick out facts from misinformation. To further increase the validity of his study, he used a control simulation and reputable services for reviewing authentic facts as a standard for drawing out comparisons. He collected 100 pieces of news from those fact-checking services, ran them through those models in controlled environments, and categorized their feedback into one of three groups: True, Partially True/False, and False. He checked the efficacy of the LLMs’ responses against the validated facts that the services had provided them.

As aforementioned, in the age of social media and the internet, computer scientists have been attempting to develop a reputable fact-checking system or medium so that people do not have to go through the hassle of individually dissecting fabricated online news from what’s authentic. However, those tests have been in vain since we have yet to access a tool that challenges fake news. Caramancion’s objective was to find whether the LLMs we already have access to could perform this task.

He tested the effectiveness of those LLMS: OpenAI’s Chat GPT-3.0 and Chat GPT 4.0, Microsoft’s Bing AI, and Google’s Bard. He ran fact-checked news through them and drew comparisons between the responses of all models. Results showed that GPT-4.0 executed best, but all models fell short of human verifiers, placing great importance on the continuing necessity for human skills. Caramancion suggested AI developers could further advance their technology in accordance with their focus on interconnecting their tech with human cognition.

Read next: ChatGPT Chronicles: A Hilarious AI Banter Blitz!
Previous Post Next Post