ChatGPT is making a bigger splash than any other recent innovation. But educators are worried, and GPTZero – the first detection tool – is already generating interest
Perhaps no other modern technology has disrupted society as quickly as generative artificial intelligence (AI). While creative AI suits like DALL-E have spewed a host of legal issues ranging from intellectual property violations to ownership-related complicationswithin just a year of its existence, the information-dissemination solution ChatGPT has been rocking the boat barelywithin a couple of months.
Mostly, it is the research and academic circle that is seeing red. So much so, that an “antidote” toChatGPT is already doing the rounds! A senior scholar at Princeton University has developed an app called GPTZero, that can identify whether a given chunk of text had been generated by ChatGPT.
Amazing innovation inviting misuse
The future looks all set for a generative AI arms race as both Microsoft-backed OpenAIand Google Sparrowis engaged in a war of information-providing chatbots.Venture capitalists have sensed a huge opportunity and big money is being poured into this space. An estimate by Pitchbook reveals that VC investments in Generative AI have increased by 425% since 2020 to $2.1bn.
Although as a technology ChatGPT is making a bigger splash than any other recent innovation and is outright threatening the ubiquitous Google search into oblivion –educators are already up in arms against its use by students and research scholars. Like all generative AI, ChatGPTwill certainly be a goldmine for forgers and plagiarists. OpenAI, its developer, has already declared a commitment to preventing AI plagiarism and other nefarious applications. Last month, Scott Aaronson, a researcher currently focusing on AI safety at OpenAI, revealed that the company has been working on a way to “watermark” GPT-generated text with an “unnoticeable secret signal” to identify its source.
The real and present danger is the use of ChatGPT by students for assignment purpose. ChatGPT can generate search results on any topic, curate them to near-human perfection, and produce an essay or a report complete with cross-references. And being open-access, it is out there for anyone to try out. It is the perfect recipe for cooking up assignments right from the school level homework up to research papers. One amazing innovation that is lying ready to be misused.
Educators to overhaul teaching methods
Across the US, professors and administrators are starting to overhaul classrooms in response to ChatGPT – indicating a potentially huge shift in teaching and learning. Some are redesigning their courses entirely, incorporating a greater percentage of oral assessments and making submission of handwritten papers mandatory instead of typed print-outs.
The issue is spreading across the globe like wildfire and India is no exception. Recent newspaper reports from Kolkata chronicles how teachers across the cities top academic campuses are being flooded with near-perfect home assignments. They are well written; the content is of high quality… and many of them are identical. Not used to such homeworksubmissions that are too good to be true, Kolkata teachers have finally woken up to the dangers of ChatGPT. Discussions ae already on to adopt measures on lines of the US academia.
The New York City education department and the Seattle Public School systemhas announced blocking ChatGPTaccess on school networks and devices. The statement from the New Yorkdepartment expressedfearsover the tool’s“negative impacts on student learning, and concerns regarding the safety and accuracy of content”. Thewordings aptly reveal both facades of concern – one:preventing unfair practices; two: ensuring that students are fed on correct information only. Universities in France and Australia have also taken recourse to the banningroute.
More conservative approaches include institutions making arrangements to educate students about the new AI tools, their pitfalls, and fair usage. Many USuniversities are incorporating papers or short courses within their curriculum to sensitise students about issues like academic integrity and information filtering.
Here comes GPTZero
And then there is the third approach – harnessing technology itself to counter the evils of technology! Some institutions, like the Harvard University, are looking for detection tools that could help flag machine-generated essays. While existing plagiarism-detection software are working hard to incorporate more refined features to catch AI-authored content, 22-year-old Edward Tian spent his winter break to come up with a dedicated tool singlehandedly that specifically targets ChatGPT. The interest it garnered was huge.
Tian is acomputer science and journalism senior at Princeton University. He calls his tool GPTZero. Within a week of its launchon January 2, around 30,000 people had tried its Beta version – and it crashed. Streamlit, the free platform that hosts GPTZero, has since stepped in to support Tian with more memory and resources to handle the web traffic.Tian was flooded with enquiries from teachers, principals and education administrators around the globe. A lot of teachers also contacted him to share their positive experiences with GPTZero.
GPTZero works because it was trained on similar datasets as ChatGPT, meaning it can measure the comparative complexity — or “perplexity” reading — of a text. “Perplexity” and “burstiness” are two signs that GPTZero utilizes to assess whether an excerpt was written by a bot.
Perplexity is a metric for text complexity; if GPTZero finds the text confusing, it has a high level of complexity and was probably created by a human. If a text is familiar to the GPTZero bot, it is more likely AI-generated.Such text will have minimal
Another reading GPTZero carries out is Burstiness. This refers to the variation in sentence length and several other factors.Burstiness contrasts the various sentence versions separately. Humans prefer to write more quickly, for instance, mixing longer, more complicated words with shorter ones. AI sentences are typically more consistent and mechanically uniform.
Although it has proved highly effective, GPTZero is not error-free. Tian is still working to refine the model’s precision. He plans on continuing to improve his app after he graduates from college. However,Tian is sure to have a lot of competition soon as several players will certainly join the race to create ChatGPT detectors.
Personally, Tian is not against the usage of ChatGPT and other AI techniques. In conversations with the media, he has opined that it is virtually impossible to prevent students from using ChatGPT. According to him, GPTZero is “not meant to be a tool to prevent the deployment of these technologies.”Instead, he thinks students must be explained what the technology is capable of and what it means for the society at-large.
Anyone interested can check out GPTZero at https://gptzero.me/
Know more about the syllabus and placement record of our Top Ranked Data Science Course in Kolkata, Data Science course in Bangalore, Data Science course in Hyderabad, and Data Science course in Chennai.