Covering Disruptive Technology Powering Business in The Digital Age

AI: A Whiz in (Some) Exams but Woeful in Arts?
July 3, 2024 News AI

Written by: Izzat Najmi Abdullah, Tech Journalist, AOPG.

There is no denying that Artificial Intelligence (AI) has been nothing short of fascinating (well, at least for me). Now, my music can be adjusted based on my current mood, all thanks to the Amazon Echo Dot or the LingLong DingDong (don’t laugh, the name may sound funny to you, but China’s answer to Alexa is no joke). Depending on where you are, this technology is truly game-changing. These assistants, yes Siri, you are included, have managed to make my mundane daily routines all fun and easier.

But here’s the thing: Despite the convenience and efficiency AI brings, have we already pushed it to its limitations of capturing true creativity that can fuel innovation or artistic expression?

Or perhaps more importantly, can it develop the kind of intuition that allows us to navigate complex situations and make nuanced decisions?

Is AI now capable of replicating or even replacing the human touch?

AI’s Progress in Analytical Tasks

Based on recent events, perhaps the answer is no, or at least “not yet” for now.

It is true, that AI systems have demonstrated significant prowess in handling analytical tasks, particularly those involving large datasets and pattern recognition, but what about human exams?

China’s gaokao, a notoriously difficult college entrance exam, has served as a proving ground for academic excellence for generations. This year, however, it wasn’t just students that were sweating over the questions. AI models took a crack at the test, with the aim of gauging their progress in tackling complex challenges.

In the language tests (mainly English and Chinese), almost all of the AI models aced with flying colours, with the top-performing AI model, Alibaba Cloud’s Tongyi Qianwen 2-72B, managing to score 303 points out of a possible 420. This seems pretty impressive and respectable for something that doesn’t really have a “brain” right? Yes, but wait till you see the results for tests involving mathematics.

Despite impressive scores in the Chinese and English sections, all of these models flunked the math section, with the highest score reaching a mere 75 points out of 150!

Teachers who were involved in grading the AI-generated responses noted a distinct lack of depth and human touch. They mentioned that the essays were all pretty well-organised, but many of them lacked the persuasive power of human “touch.” These AI models lack techniques such as using relevant examples, citing references, and employing metaphors—all of which were absent in the outcome of their writings, leaving the responses quite sterile and unconvincing. Some AI models also faced issues with adhering to the word limit in essays, often exceeding it to some extent. While human candidates might lose points for not meeting the required word count, AI’s tendency to over-generate content indicates a lack of precision and control in its responses, a problem that needs addressing as AI continues to develop.

The Struggle with Creativity

Answering tests and adhering to the required test instructions is not the only problem that AI can’t seem to solve as creativity remains another significant hurdle for them. The ability to generate original ideas, use metaphors, and understand nuanced language is still largely known to be beyond the reach of AI systems, but wouldn’t it be funny if AI lost in a competition that is specifically for AI-generated photos?

Miles Astray, the photographer behind the winning (and later disqualified) entry titled “F L A M I N G O N E,” intended the entire experience to be a thought-provoking experiment. The image itself captured a flamingo mid-scratch, its head and neck obscured by its body, creating a surreal, almost headless appearance. Astray submitted it to the AI category of the 1839 Awards, a relatively new photography competition.


Astray submitted this real image to an AI photo competition, and it won before being disqualified.

Astray wasn’t the first to use a real photo in an AI competition. In 2023, artist Boris Eldagsen took the opposite approach, winning a real photography competition with an AI-generated image. Astray, explaining his motivation, mirrored Eldagsen’s sentiment: “We’re not really ready for this technology. We’re not really keeping up with how fast it is moving.”

“F L A M I N G O N E” initially won over the judges and even secured the People’s Vote Award before the truth came out. Astray, despite the deception, wasn’t trying to be malicious. He in an interview, wanted to highlight two key points: The enduring power of nature’s creativity compared to AI and the growing difficulty of distinguishing real content from AI-generated content.

So, Will AI Surpass Humans?

The question of whether AI will eventually surpass humans in creativity or complex problem-solving is a matter of considerable debate. On one hand, AI technology is continually advancing, and future iterations may develop more sophisticated models that can better emulate human thought processes. The rapid pace of AI research suggests that significant improvements are on the horizon, potentially enhancing AI’s capabilities in areas currently dominated by human cognition.

However, on the other hand, there are inherent aspects of human cognition that may prove difficult for AI to fully replicate. Take, for example, creativity, as it is deeply linked to human emotions, cultural experiences, and individual consciousness – elements that are challenging to encode into an algorithm. The ability to draw on personal experiences, interpret cultural nuances, and inject emotional depth into creative works is a uniquely human trait.

Thus, while AI continues to make impressive strides, it may still fall short in some of the capabilities that humans can do.

For now, the most promising path forward is a symbiotic relationship between humans and AI, where each complements the other’s strengths. AI can assist in processing vast amounts of data, identifying patterns, and performing tasks with precision, while humans can provide the creativity, intuition, and nuanced understanding that AI currently lacks. This collaborative approach leverages the best of both worlds, driving innovation and enhancing decision-making across various fields.