ChatGPT is making waves within the technology space and has sparked debate amongst various researchers about its impact on students, industries, the labor market, etc. While there are numerous articles and research papers published on ChatGPT, there may be a necessity for higher understanding to grasp how scholars perceive this latest technological disruption.
With each passing day, we’re seeing increasingly research articles being published. While some research has been critical of ChatGPT, some have highlighted its positive impacts only, and few have analyzed each the professionals and cons of this latest technology. We’ve analyzed 100 research papers and case studies from top journals to take a look at ChatGPT from the lens of a researcher and tried to narrow right down to the areas, which in keeping with researchers shall be highly impacted. This has been done by reviewing all of the articles and categorizing them based on the researchers viewpoint in regards to the impacted sectors and fields.
ChatGPT is an Artificial Intelligence chatbot that has been developed by OpenAI. It was released in Nov 2022 and is built on OpenAI’ s GPT-3.5 and 4.0 large language models (LLM). Since its launch, ChatGPT has been the topic of debate amongst various scholars, particularly around its impact on students and various industries and sectors.
While AI has been around for a while now, ChatGPT took everyone by storm. Its ability to put in writing human-like text content has everyone talking about it. While it’s still latest, more researchers and analysts are discussing about its pros vs cons, but nobody is denying the proven fact that it’s here to remain and can impact everyone in various capacities.
ChatGPT presents many opportunities in several fields, however it also brings its own challenges. While there are some areas where ChatGPT can totally disrupt the way in which of working, there are other areas where it’d pose challenge in term of authenticity, ethics, and fabricated work. Because it got here into picture, researchers have focused on different fields where ChatGPT’ s application can bring a significant change. These include Medical Research, Academic Research, Healthcare, Education, Social Media to call a number of. There are various ways in ChatGPT can really eliminate quite a lot of redundant processed which on one side is nice but may have impact on the labor markets.
All these research studies give us an insight into how our lives shall be impacted in future via tools like ChatGPT and that is only the start. This research analyses top cited articles from various journals and tried to know a researcher’s perspective of ChatGPT and whether they give the impression of being at it in a positive, negative or a neutral way. It also tries to narrow down probably the most critical area in keeping with researchers, where ChatGPT can bring a change or have maximum impact.
:
ChatGPT is a highly advanced language model developed by OpenAI that relies on the GPT-3.5 architecture. It was launched in November 2022. This model is a highly effective artificial intelligence (AI) system that may understand and generate natural language text at a level that’s comparable to that of human beings. ChatGPT is designed to perform a wide selection of language tasks, including text completion, text classification, and text generation. Additionally it is able to serving as a conversational interface, allowing users to interact in natural language conversations with the model.
The architecture of ChatGPT relies on the transformer-based model, which is a highly effective deep learning architecture that has been utilized in many natural language processing tasks. The transformer-based architecture is very effective in processing large amounts of sequential data, which makes it highly suitable for language modelling tasks. The transformer-based architecture is very effective in capturing the contextual dependencies between the assorted elements of a sentence, which enables the model to generate highly coherent and contextually relevant responses.
ChatGPT has been trained on an enormous amount of information, which incorporates billions of tokens from quite a lot of sources, equivalent to books, web sites, and other text-based resources. This huge corpus of information enables the model to generate highly accurate and coherent responses. The model has been trained using a mix of unsupervised and supervised learning techniques, which enable it to learn from the vast amounts of information that it has been exposed to. The model is designed to repeatedly learn and adapt to latest data, which enables it to enhance its performance over time.
Just after 2 months of it’s launch, ChatGPT had over 100 million lively users. To provide an idea of the quantity, TikTok which is a very fashionable social media platform, reached 100 million users after 9 months of coming into the market.
There’s a giant difference between ChatGPT and search engines like google. While search engines like google indexes webpages on web to assist the user find the data they asked for, ChatGPT doesn’t has the power to look web for information.
ChatGPT uses the data it learned from training data to generate a response, which might result in incorrect data because it’s training itself on data being fed and never taking a look at resources being updated and being made available online.
While we’re seeing increasingly articles and papers on ChatGPT, AI and NLP, the target of this study was to review these articles and check out to see ChatGPT’ s impact from a researcher’s viewpoint. This helped us understand how researchers are interpreting ChatGPT’ s potential impact and the areas to lookout in near future as more impact of ChatGPT may also result in a shift within the labor market of that specific sector.
This can also be necessary as the largest concerns immediately with ChatGPT are two. Jobs losses as a consequence of AI replacing humans doing redundant tasks after which the moral concerns in field of academics and research. There’s a general thought process that ChatGPT can have biggest impact on academia and research, but our study goes into the small print on top articles by different researchers and tries to unearth, what are the areas which may need the largest impact in keeping with researchers.
Since ChatGPT was just launched in November 2022, we are going to surely find many other areas where research studies will occur and impact of AI shall be evaluated. This study shows the present trend based on available data.
For our research, we reviewed top 100 articles from leading journals. These were sorted based on the variety of citations. ChatGPT being a latest phenomenon available in the market, we didn’t find lot of citations for ChatGPT based articles. In total there have been 371 citations from 217 articles published between 2022 and Apr 2023.
Out of those articles, we reviewed the highest 100 ones and first, categorized just by taking a look at the title of the articles whether or not they are written with a positive, negative, neutral (no clear positive or negative overtone) or debating (positive vs negative) tone. There’s a growing interest on this topic and we are going to soon see more variety of articles be updated on ChatGPT in coming time.
Categorizing these articles just based on the titles doesn’t categorically explains if overall, the researchers look the ChatGPT phenomenon in a positive, negative, or balanced light. So, we first categorized the articles into 4 categories going purely by the title. Neutral and Debating tones will be looked in a combined view as Debating tone signifies that the creator is taking a look at ChatGPT in a balanced way analyzing the positives vs negatives.
was assigned to titles which didn’t very clearly state positive impacts of ChatGPT or had a clearly stated debatable title (eg: Every day briefing: Will ChatGPT kill the essay project?)
was given when the title suggested analyzing a positive impact of ChatGPT (For eg: ChatGPT: five priorities for research).
was assigned to titles which clearly stated some negative impact or downside of using ChatGPT (for eg: Abstracts written by ChatGPT idiot scientists)
We then went through all of the 100 articles and intimately analyzed if the researchers are taking a look at ChatGPT in a positive, negative, neutral or a debating light. We found a niche in any research done to seek out the inclination of researchers towards this topic. This becomes more necessary as recent studies have shown that AI generated content couldn’t be distinguished from human generated one by researchers. In a study by Northwestern University and the University of Chicago, AI and human generated content was evaluated by human reviewers and so they couldn’t detect difference in 32% of the cases. Since ChatGPT was released for general public, there was a debate on its impact on the research quality and we desired to analyse this tool very specifically from a researcher’s viewpoint.
We reviewed top 100 articles published in leading journals and while we had categorized the articled earlier as Positive, Negative, Debating /Neural categories purely based on titles but we then categorized the articles based on the content. After reviewing the 100 articles, below was the categorisation:
What was found is that the majority articles at this stage are debating the professionals and cons of Chat GPT, AI and LLM usually and never explicitly praising or criticizing the product. Most articles are analyzing the impact of AI, LLM and ChatGPT based on how it should be used. For eg. ChatGPT Friend or Foe(Editorial) discussed how it could be each, depending on how it should be utilized.
: we categorized articles as Negative in the event that they are only discussing fallacies of ChatGPT. For instance, ChatGPT listed as creator on research papers: many scientists disapprove discusses how ChatGPT can’t be creator of a paper because the creator must take obligation of their contribution while ChatGPT is just writing the paper based on the pre trained data model or in some case it fabricates the small print also which is dangerous.
: articles have been categorized under positive category in the event that they clearly state only positives of ChatGPT and don’t discuss any shortcomings. Example article is ChatGPT, An Artificial Intelligence Chatbot, Is Impacting Medical Literature Which Discusses the impact of ChatGPT on medical literature, including potential advantages and limitations as well but no negative impacts.
: we have now categorized any article as neutral if the creator has not taken a big positive or negative tone from the beginning. All articles that are debating the professionals and cons or are doing a review of the tool have been categorized as Neutral.
Within the 3 categories defined, we also sub categorized them based on the realm of study and below was the information:
We’ve differentiated Medical Science and Research from pure Research as the previous may be very specific for the articles published in analyzing it’s impact in the sector of medication.
Although there are variety of subcategories, however the literature review found that majority of the articles are skewed towards three different sectors/areas.
While reviewing the highest 100 articles on ChatGPT, we found that 54 of the articles are debating pros vs cons on using it or discussing its positives vs negatives. Other 37 articles speak about ChatGPT in a positive way discussing its potential disruption in various fields. Only 9 articles are completely critical about ChatGPT.
Since this topic may be very latest, we have now still not seen too many negative reviews about ChatGPT now, it could be simply speculated in regards to the potential demerits or exploitation of ChatGPT.
Out of the 100 articles reviewed, interestingly, 42 articles have been in the sector of medical science and research which shows, there is a big interest in how ChatGPT will be utilized as a disruptive force in medical sciences. Lots of the research articles on this field are around using ChatGPT in areas of research, providing real time information, generating patient education materials etc.
The second and third rating areas are as expected, in the sector of academics and research. There are obvious discussions around potential exploitations of using ChatGPT in writing research and scholarly articles. That is more so because ChatGPT is trained on existing dataset and in lots of cases also fabricates which might result in potentially incorrect and invalid research articled. The more worry some situation is that in some recent studies, articles written by ChatGPT couldn’t be differentiated by ones written by humans.
ChatGPT and AI are here to remain and the effect on few sectors goes to be substantial. While there shall be some sectors and areas where the impact may be more, cutting off redundant manual tasks by AI but otherwise ChatGPT shall be more of an aid to existing processes reasonably than a whole job eliminator.
The consequences may even be seen in near future when ChatGPT will further develop. Most articles though see it making huge impact in medical sciences, research and academia. While ChatGPT generally is a real aid in assisting with tasks like generating medical history, preparing patient discharge summaries, creating write ups whereas it could also result in low quality research materials and negatively affect the research skills of research scholars and students.
Overall, we’d like to closely monitor how ChatGPT goes to shape up in near future and be prepared to get used to using AI in our each day lives.
- Anna-Carolina Haensch, Sarah Ball, Markus Herklotz, Frauke Kreuter (2023). SEEING CHATGPT THROUGH STUDENTS’ EYES: ANANALYSIS OF TIKTOK DATA
- Mohammad Aljanabi,Mohanad Ghazi, Ahmed Hussein Ali,Saad Abas Abed (2023).ChatGpt: Open Possibilities
- Jürgen Rudolph, Samson Tan,Shannon Tan(2023): ChatGPT: Bullshit spewer or the tip of traditional assessments in higher education?
- Renana Peres, Martin Schreier, David Schweidel, and Alina Sorescu(2023): On ChatGPT and Beyond:How Generative Artificial Intelligence May Affect Research, Teaching, and Practice
- Ali Zarifhonarva(2023): Economics of ChatGPT: A Labor Market View on the Occupational Impact of Artificial Intelligence
- A. Shaji George, A.S. Hovan George, A.S.Gabrio Martin(2023): A Review of ChatGPT AI’s Impact on Several Business Sectors
