Experts Warn of Misinformation Risks While Acknowledging AI’s Limitations
SINGAPORE: ChatGPT, the innovative chatbot capable of writing essays and tackling complex questions, may not be on the verge of replacing humans anytime soon, according to an expert in artificial intelligence (AI). On Tuesday (Dec 20), Toby Walsh, a leading researcher in the field, discussed the implications of AI advancements during an interview with CNA’s Asia First.
“ChatGPT can draft business letters, write film scripts, and answer exam questions. However, it lacks true understanding,” Walsh noted, emphasizing that while the AI can often provide accurate information, it does not genuinely comprehend the content it generates. “At the end of the day, it operates on probabilities rather than understanding, which means it can also produce completely incorrect information.”
Walsh expressed minimal concern over machines overtaking humans, stating that AI lacks sentience or consciousness. However, he warned that reliance on such tools could lead to complacency, making people overly dependent on AI for tasks they should undertake themselves. “The real risk is not malevolent machines taking over the world, but rather the subtle ways we might inadvertently relinquish responsibility to these tools,” he cautioned.
OpenAI launched ChatGPT for public testing on November 30, and it gained over a million users within a week, demonstrating significant interest in AI’s capabilities.
Chatbot’s Role in Education
The impact of AI tools like ChatGPT on education is already sparking debate. Walsh raised concerns that educators might need to rethink traditional methods of assessing student knowledge. “If students can simply use ChatGPT to generate essays, how do we teach them to write effectively?” he questioned.
In contrast, Jonathan Sim, an instructor at the National University of Singapore’s Department of Philosophy, argued that educators should not shy away from integrating AI into learning environments. “We should teach students how to use these tools effectively to enhance their learning,” he stated.
Sim has designed an exercise for his students where they use ChatGPT to generate essays, followed by critiquing the content. “I would grade the essays generated by ChatGPT at best a B. It offers a valuable learning opportunity to understand why it’s not an ‘A’ essay,” he explained, emphasizing that this process could help students improve their writing skills.
Misinformation and Ethical Concerns
Despite its potential, Walsh highlighted that ChatGPT poses a significant risk as a “perfect tool” for spreading misinformation. “Social media is already inundated with fake news, and ChatGPT could exacerbate the issue by generating plausible text at scale and speed,” he warned.
He elaborated that personalized misinformation could become more sophisticated, as scammers could easily tailor messages using information gleaned from the internet. “In the wrong hands, it’s a potentially dangerous tool,” he cautioned.
Walsh, who authored Machines Behaving Badly: The Morality of AI, noted the lack of transparency surrounding how ChatGPT operates. “While we believe that the million users are helping to improve the AI, we don’t fully understand how their queries influence its development,” he added.
A New Way of Accessing Information
Walsh believes that ChatGPT represents a significant evolution in information retrieval. He compared it to how Google transformed search engines after the early days of AltaVista. “Rather than merely following links, if a search engine can provide direct answers, it may become my preferred destination,” he stated, suggesting that AI could redefine how we access information.
However, he cautioned that regulation has yet to keep pace with rapid technological advancements. “While social media is finally being adequately regulated, we need to act quickly to address potential harms from emerging technologies,” Walsh warned. “We must ensure that as these tools evolve, they do not become data monopolies that undermine societal benefits.”
In summary, while AI tools like ChatGPT showcase remarkable capabilities, the conversations surrounding their ethical implications, educational roles, and potential for misuse are crucial as society navigates this new digital landscape.
