“ChatGPT will redefine the future of academic research. But most academics don’t know how to use it smartly,” Mushtaq Bilal, postdoctoral researcher at the University of Southern Denmark, tweeted recently.
Academia and artificial intelligence (AI) are increasingly intertwined, and as AI continues to advance, it is likely that academics will either continue to embrace its potential or voice their concerns. about its risks.
“There are two camps in academia. The first is for early adopters of AI, and the second is for professors and academics who believe AI corrupts academic integrity,” Bilal told Euronews Next.
He places himself resolutely in the first camp.
The Pakistani-born, Denmark-based professor believes that if used thoughtfully, AI language models could help democratize education and even pave the way for more knowledge.
Many experts have pointed out that the accuracy and quality of the output produced by language models such as ChatGPT is unreliable. The generated text can sometimes be biased, limited or inaccurate.
But Bilal says understanding these limitations, coupled with the right approach, can ensure that language models “do a lot of good work for you,” especially for academia.
Incremental incentive to create a ‘structure’
To create a structure worthy of academia, Bilal says mastering incremental prompting, a technique traditionally used in behavior therapy and special education, is fundamental.
It involves breaking down complex tasks into smaller, more manageable steps and providing prompts or clues to help the individual complete each successfully. The prompts then become progressively more complicated.
In behavior therapy, progressive prompting allows individuals to develop their sense of trust. In language models, this allows for “much more sophisticated responses”.
In a Twitter thread, Bilal showed how he managed to get ChatGPT to provide a “brilliant preview” for a newspaper article using incremental prompts.
In his demo, Bilal started by asking ChatGPT about specific concepts relevant to his work, then asking about the authors and their insights, guiding the AI-driven chatbot through contextual knowledge relevant to his essay.
“Now that ChatGPT has a fair idea of my project, I ask him to create a canvas for a newspaper article,” he explained, before stating that the results he obtained would probably earn him “20 working hours”.
“If I just wrote a paragraph for each point in the outline, I would have a decent first draft of my article.”
Incremental incentivization also enables ChatGPT and other AI models to help “make education more democratic,” Bilal said.
Some people have the luxury of discussing potential academic outlines or angles for scientific papers with professors at Harvard or Oxford, “but not everyone does,” he explained.
“If I was in Pakistan, I wouldn’t have access to Harvard professors, but I would still need to brainstorm ideas. So instead, I could use AI applications to have an intelligent conversation and help me formulate my research.”
Bilal recently made ChatGPT think and talk like a Stanford professor. Then, to verify the authenticity of the output, he posed the same questions to a real Stanford professor. The results were amazing.
ChatGPT is just one of many AI-powered apps you can use for academic writing or to mimic conversations with renowned academics.
Here are other AI-driven software to help you with your academic endeavors, handpicked by Bilal.
In Bilal’s words: “If ChatGPT and Google Scholar got married, their child would be Consensus – an AI-powered search engine.”
Consensus is similar to most search engines, but what sets it apart is that you ask Yes/No questions, to which it provides answers with the consensus of the academic community.
Users can also query Consensus about the relationship between concepts and the cause and effect of something. For example: Does immigration improve the economy?
The consensus would answer this question by stating that most studies have shown that immigration generally improves the economy, while providing a list of the academic papers it used to arrive at the consensus, and finally sharing the summaries of the best articles he has analysed.
The AI-powered search engine is only equipped to answer on six topics: economics, sleep, social policy, medicine, mental health and health supplements.
Elicit, “the AI research assistant” according to its founders, also uses language models to answer questions, but its knowledge is purely research-based, enabling “intelligent conversations” and brainstorming with a highly knowledgeable source. and checked.
The software can also find relevant articles without perfect keyword matches, summarize them and extract key information.
Although language models like ChatGPT are not designed to intentionally mislead, they have been proven to generate text that is not based on factual information and include false citations of articles that do not exist. .
But there is an AI-powered app that gives you real quotes from actually published articles – Scite.
“It’s one of my favorites for improving workflows,” Bilal said.
Similar to Elicit, when asked a question, Scite provides answers with a detailed list of all articles cited in the answer.
“Also, if I make a claim and that claim has been refuted or substantiated by various people or various newspapers, Scite gives me the exact number. So that’s really, really powerful.”
“If I were to teach a seminar on writing, I would teach how to use this app.”
“Research Rabbit is an amazing tool that ACCELERATES your research. Best of all, it’s FREE. But most academics don’t know that,” tweeted Bilal.
Called by its founders “the Spotify of research”, Research Rabbit makes it possible to add academic articles to “collections”.
These collections allow the software to learn about the user’s centers of interest, generating new relevant recommendations.
Research Rabbit also allows the scientific network of papers and co-authors to be visualized graphically, so users can follow the work of a single subject or author and dive deeper into their research.
ChatPDF is an AI-powered application that makes it easier and faster to read and analyze newspaper articles.
“It’s like ChatGPT, but for research articles,” Bilal said.
Users start by uploading the PDF research document to the AI software and then start asking it questions.
The app then prepares a brief summary of the article and provides the user with sample questions they could answer based on the full article.
What promise does AI hold for the future of research?
The development of AI will be as fundamental “as the creation of the microprocessor, the personal computer, the Internet and the mobile phone”, wrote Bill Gates. in the last post on his personal blog, titled “The Age of AI Has Begun”.
“Computers haven’t had the effect on education that many of us in industry hoped for,” he writes.
“But I think in the next five to 10 years, AI-based software will finally hold the promise of revolutionizing the way people teach and learn.”