From helping with homework assignments to recipe recommendations based on leftovers to penning love letters for Valentine’s Day, it seems there’s very little that OpenAI’s popular bot ChatGPT can’t do. As part of the Silver Jubilee edition of the R&D Showcase 2023 hosted by IIITH, a roundtable was conducted amongst faculty to mull and brainstorm over the future of academics and research in the times of ChatGPT. Here’s a summary.
While ChatGPT was the first to disrupt our world with its innovative, human-like textual responses, other similar AI-language models are in the fray too. For purposes of simplification, the summary of recommendations and observations from the roundtable uses the term ‘ChatGPT’ as a generic reference to all AI-generative tools.
The discussion kicked off by first placing the AI language model in its context and the myriad ways in which it is being used. Trained on hundreds of billions of documents and websites, tools like ChatGPT are natural language processing models that generate text in a specific style or tone for tasks such as translation, summarization and answering questions. For instance, it can create a literature review on a particular topic or allow researchers to quickly generate draft versions of research papers, grant proposals, and other written materials. This has understandably led to ennui among the senior academic community who is of the opinion that instead of embracing these tools as mere companions, the younger generation has come to heavily depend upon them.
Drawbacks of Generative Models
Innovative as they may be, these tools are not without their own shortcomings. Faculty opines that while ChatGPT may provide facts and figures, it cannot conceptually deliver content. Some of the drawbacks and challenges that were listed include language translation complexities, unavailability of calculative concepts and source authenticity. Data in languages other than English is not extensively available on such AI tools. The tools cannot solve or address mathematical theories. While translation is possible, it is not accurate for some languages. ChatGPT-like tools cannot fully understand the context and meaning of the text they generate, and they cannot perform well in tasks that require common sense reasoning or logical reasoning. Like with most AI models, the GPT model might generate a response based on stereotypes or prejudices rather than on accurate or objective information. In its response, the source of the data or the content is not provided. It kills natural human creative thinking and makes the user dependent on external tools. The tool itself also has its limits as, after some time, it stops responding.
The panel on Teaching in the time of ChatGPT anchored by Profs. Radhika Krishnan, Makarand Tapaswi and Charu S witnessed discussion on how with easy access to, and availability of AI tools for both the educator as well as the receiver, AI can be both a boon or a bane. In the case of the former, AI can go beyond the traditional black or white boards and the powerpoint slides, and assist in creating pedagogical content. It can help create interactive projects, tasks, or resources. For example, if one is working on AI/ML, then it appears to be a good tool to understand and play around with. But if one is in the theoretical domain, then it would depend on the objective of the course and the topics of learning. Teachers may leverage the tools to be aware of the ever-changing content and for the quick resolution of queries to save time. At the same time, everyone ought to be aware of the correct usage of such tools, and students must also be mindful of using the data for reference or for analyzing their perspectives.
Engagement With Students
ChatGPT has the potential to revolutionise education and bring new opportunities for teachers to educate their students in various ways. It can help teachers plan the teaching flow and rigour as per the class levels/interests. To deal with the students and their learning resources, teachers should openly interact about the AI tools and improve engagement with the help of interactive lessons. ChatGPT may create customised study plans, problem statements and case studies while providing student guidance. Student confidence may also improve after using ChatGPT for different assignments.
Evaluation and Assessment
AI tools may be helpful for teachers to frame newer ways of assessments. For instance, ChatGPT’s ability to provide oversight on the quality of written work could be helpful in the field of education. Evaluation is easier and more convenient using AI tools that help to grade and provide customised feedback on student assignments. One can also use ChatGPT to create interactive language learning activities like facilitating discussions on topics of interest. The GPT could provide background information and pose questions.
Academic Integrity And Plagiarism
The GPT model may have drawn flak from noted linguist Noam Chomsky who labelled it as “high-tech plagiarism” and a “way of avoiding learning” but higher educators seem to be divided on rigid compartmentalization of the usage of such tools. This is due to the fact that along with helping educators frame newer ways of assessments, generative AI tools can also help in testing authenticity and preventing plagiarism. However, in the public beta mode of Chat GPT, OpenAI recently acknowledged that its detection tool is unreliable on texts under 1,000 characters, and that AI-written text can be edited to trick the classifier. Hence for ideation, the human brain can always overpower AI tools.
Yay or Nay?
Acknowledging that AI tools are a double-edged sword – improving the language model and yet sometimes providing incorrect answers to numerical or logical queries, the group agreed that we need to embrace such tools instead of misusing or dismissing them. Teachers can use these tools to design curriculum for emerging concepts. ChatGPT is not a replacement for teachers but a tool that can support them in their role. Therefore teachers can upskill themselves and learn how to use the generative model in their classrooms effectively. This includes understanding its capabilities and limitations and integrating them into teaching practices. As for students, they can opt for open book submissions rather than available GPT assignments. ChatGPT should complement other methods of determining the source of text instead of being the primary decision-making tool.
Research With A Bot
The panel on Reimagining Research Frontlines was anchored by Profs. Prasad Krishnan, Surya and Ravi Kiran. They opened the discussion on the need to reimagine the modes and paths to be taken towards the tech world especially on university campuses where new digital tools come into play all the time enabling the advancements of AI tools for automated research. Terming intelligent ways of utilising AI tools as a necessity, they said that researchers ought to look forward to the wild and innovative possibilities of the future. With the hope of more and more authentic research publications on diverse topics, irrespective of the hindrances, AI tools are to be welcomed.
Collaborating With Tools
With researchers currently using AI tools to frame their abstracts in the best way possible, such tools can be used as a starting point to deal with curiosity amongst researchers. Students however need to be aware of the correct usage of such tools and use data appropriately for analysis. ChatGPT is increasingly being referred to in many research papers and getting multiple citations too. In fact, it has made the job of early PhD and Masters students quite redundant compelling faculty to make an observation on the need to train them to do independent research to find relevant data.
Better Literature Review
ChatGPT not only retrieves many research papers and succinctly summarises them, but it also provides papers’ full titles, author names, journals in which they have been published and even DOI for many research areas. GPT on the other hand can only search for papers that are not yet published or those that are located deep in the Web. It may also have some errors in reporting the source. ChatGPT has demonstrated that it can adeptly write in the style of academic scientists. Hence this language technology can be used by researchers to improve the readability of their work or as a writing aid to boost equity and access for publishing outside their native language.
Speed Up Writing
With the help of ChatGPT, it is now possible for a researcher to write a paper in lesser time than previously required saving physical energy. The use of this tool makes it invaluable for improving the quality and efficiency of our writing. With its advanced natural language processing capabilities, ChatGPT helps us generate ideas, gather information and write in a concise and well-structured manner. That said, while it tries to create unique writing every time, occasionally it presents phrases that have been used before on the internet. Therefore checking ChatGPT’s output for plagiarism is essential. Copying answers from such tools can block the creative thinking process that is critical for researchers in diverse fields.
Irrespective of whether you are a student, researcher or a professional writer, ChatGPT helps produce a high-quality research paper that can effectively communicate your ideas and findings. It saves time and energy, fastens the process enabling you to focus on analysis and thereby leading to more innovations. In addition to all these, it can provide an element of the right style or craft.
The teaching community was of the view that student ability to judge the correctness of solutions ought to be enhanced as makers can make mistakes. The students need to start believing in themselves and realise that they can write better answers than machines. AI tools don’t hold ethical values and hence are not liable for plagiarism or responsibility unlike humans. When students use AI-assisted writing tools for their research papers without appropriate attribution, it could raise a number of ethical concerns including fairness, originality, intellectual property and academic dishonesty. By taking a proactive approach to regulating the use of AI and teaching students about the ethical usage of such tools, we can ensure that the benefits of these tools are maximised while minimising the potential negative impacts. ChatGPT and other similar tools call for a need to reassess our internet policies and data laws so that no corporate gets a monopoly.
The Wild Wild Future
The panel on Wild Possibilities Of Future Innovations was anchored by Profs. Aarti, Siddhartha and Kannan Srinathan. Concerns were voiced over the manner in which generative tools have access to research publications. The overall thinking was that such tools should not overpower the library or other classical data sources. Typically access should be authorised by respective authors only. OpenAI applications can fall under academic integrity policies like plagiarism but the grey zone between clearly plagiarised work and an educational support tool is large. Hence guidelines should be developed for the appropriate usage of ChatGPT in academic work.
Talk A Bot
ChatGPT can only respond textually but there is an upcoming AI chatbot Wotabot where one can talk to an AI chatbot. Remarking that in the future with bots interacting and communicating with each other, which will not only assist humans but also make our lives easier, the panel said that the scope of futuristic AI lay in the credo: AI for good. AI tools can be made interactive for longer and keep track of our usage, needs and comfort. Plus, with monetized versions of generative tools coming out, they should come with a regulated policy on data control.
At Your Service
By sending responses instantly to multiple people at the same time, ChatGPT can help save human time and energy. With some human intervention, it was speculated that such bots can help leaders solve problems with existing data points. They can also help generate ideas for new designs or prototypes. Custom AI tools can also act as mentors to researchers assisting them virtually.
Ideas were bandied about on how AI tools can be connected with robots to generate intellectual avatars, acting according to human requirements in various ways. The Metaverse could act as a bridge between the real world and its virtual counterpart as we navigate the two realms multiple times a day. The aerial version of AI tools (or the ChatGPT avatar version) can not only help people feel more comfortable but also give more effective outputs in the real world.
On whether chatbots respond to feelings and deliver answers with consciousness, it was surmised that AI tools can be developed to respond to the human mood and feelings. ChatGPT itself is sensitive to tweaks of input phrasing or when the same prompt is attempted multiple times. Hence there is a huge scope that the bot answers with complete consciousness and reacts as per requirements.
Allaying general fears about the emergence of AI tools, the roundtable signed off by remarking that instead of getting disrupted by any technology, we ought to adopt it for our good. With tech advancements, it is essential for educators and policy makers to carefully weight the potential benefits and drawbacks of using ChatGPT and take necessary steps to ensure that students receive accurate, unbiased and personalised feedback. It was noted that nothing can replace teaching jobs and educators should embrace manmade tools and continue to believe in themselves.