Fiona Fang, Technology
Leave a Comment

Exploring the Capabilities and Challenges of ChatGPT

Image source

ChatGPT, one of many new revolutionary generative AI tools, could upskill the population and improve work efficiency. But governments also need to settle widespread concerns to maximize potential while minimizing possible risks.

By Fiona Fang

ChatGPT & Its Popularity

Generative AI describes algorithms that can be used to create new content, including audio, code, images, text, simulations and videos. Released on Nov. 30, 2022, Chat Generative Pre-trained Performer, or ChatGPT, is the latest and most popular AI chatbot by Open AI. Once released, ChatGPT immediately attracted much public attention. The fastest-growing application in history, it reached 100 million monthly active users in January—taking only two months to do what took social media giant TikTok nine months.

Creativity of ChatGPT

What’s impressive about ChatGPT is that it holds the potential to drastically change how people create content. Most AIs over the last couple of decades are able to analyze existing data, find data anomalies, detect fraud, and recommend a movie, according to Gaurav Gupta, a partner of Lightspeed Venture Partners. While these are important and impressive, ChatGPT goes further and is able to simulate engaging in complex and informative human-like conversations.

After digesting massive amounts of human text, essays produced by ChatGPT are “better than most writing professors [get] from students over the decades,” states Naomi Baron, Professor Emerita of Linguistics at American University. However, because generative AI can only create new content based on the previous patterns and relationships of the data it was trained on, the contents created by it lack uniqueness and innovation. Wendy Hall, Regius Professor of Computer Science at the University of Southampton, says that while generative AI tools “copy, predict and cleverly merge ideas semantically,” they don’t create entirely new content.

Accuracy of ChatGPT

Like Apple’s Siri, ChatGPT can answer almost every closed-book question, including general knowledge questions like historical events and mathematical questions like algebraic equations. But there’s a catch: some of these answers might be blatantly false. However, ChatGPT’s developers warn that this AI tool will “occasionally” produce incorrect answers for several reasons. First, it has limited knowledge of events after 2021. For instance, if you ask ChatGPT a question like who won the 2022 World Cup, it will not generate an answer. Second, it may pick up biases present in the large datasets upon which it is trained. With no citations, ChatGPT’s processes also give people no way to verify its sources. The concern is especially alarming in academia as the authority of academic passages relies largely on the authority of their sources. To respond to this concern, leading global universities and schools, like France’s renowned Sciences Po to the Australia’s University of Sydney, have banned its use. When using this tool, users must keep in mind its limitations and always verify its answers with multiple sources.

A Further Step to Improving Efficiency

With the recent advances in ChatGPT, the question is how people will adapt. Giving almost everyone the ability to write well could replace current tasks, such as copywriting, answering customer service inquiries, writing news reports, and creating legal documents. Individuals in these fields, especially those who are at entry level and fail in advancing their skills, will potentially be impacted.

However, it might not be bad news that these emergent AI tools could replace some workers: they could increase efficiency and productivity. AI software platforms, like GitHub Copilot and ChatGPT, could help developers write code faster and more easily. Andrej Karpathy, Tesla’s Director of Artificial Intelligence, tweeted that the former wrote 80% of his code with 80% accuracy, allowing him to simply refine the already-written code.  

Over the next few decades, AI could liberate people from repetitive and mundane work. Teachers could use AI to automate administrative tasks. For instance, AI platforms, like Gradescope, save teachers time when it comes to grading homework and tests. After evaluating both multiple-choice tests and written responses, these AI tools can also track individual knowledge gaps in learning. Teachers then could spend more time focusing on the more subjective areas of education that AI cannot yet quite cover, such as giving feedback on essays, evaluating oral presentations, and creating lesson plans.
Powerful AI tools can also help lawyers work more efficiently. As an online and AI-powered legal research tool, Casetext helps lawyers locate cases quickly and easily. To achieve this function, it analyzes different cases to find those that are most relevant to the case at hand. Currently, over 10,000 law firms rely on Casetext’s Parallel Search to find the most on-point cases faster. As new technical progression upskills the population and further improves efficiency, the next question will be how its full abilities could be utilized to benefit society.

Copyright Concern

To fully realize the potential of AI tools, we must ask what issues need to be solved, and what new systems could be built with them. The biggest concerns that the government needs to settle are around copyright. People are worried about the potential legal risks of using AI-created art pieces. “There are real concerns with respect to the copyright of outputs from these models and unaddressed rights issues with respect to the imagery, the image metadata and those individuals contained within the imagery,” says Craig Peters, CEO of Getty Images. In other words, producing AI-generated images runs the risk of rampant copyright infringement, as AI models are currently incapable of creating art out of nothing; put simply, both their inputs and outputs consist of artworks already created by other artists, thereby leading to a legal gray area in terms of ownership and copyright. Given these concerns, Getty Images has banned the upload and sale of illustrations generated using AI art tools like DALL-E, Midjourney, and Stable Diffusion. Similarly, the widespread use of ChatGPT adds to these worries since it generates passages without references to other sources. With these unsettled concerns, people may not dare to use content created by these generative AIs, and the value of these tools could be diminished.

To fully harness the potential of advanced AI tools such as ChatGPT, the government will need to recognize the potential risks of new generative AI tools and establish new rules. Given that AI-created content commits copyright infringement if their inputs are protected by copyright law, it might behoove the government to pass legislation that limits the use and production of such content.

Final Thoughts

AI will likely revolutionize the job market by taking over menial tasks and creating new employment. Even as 50% of young adults aged 18 to 24 believe that ChatGPT will double their productivity, millennials are worried about job loss. Only time will tell what new employment will be created on the heels of such technological innovation and how each generation will adapt to it. As AI technology continues to advance at a remarkable pace, it’s safe to assume that ChatGPT and other similar tools will become increasingly ingrained in our daily lives and will likely be here to stay for the foreseeable future. □

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s