By Nigel P. Daly & Yung-Chun Chang, Associate Professor at the Institute of Data Science, Taipei Medical University (printed in Taipei Times, May 6, 2023; https://www.taipeitimes.com/News/editorials/archives/2023/05/06/2003799240)
We may not realize it, but we rely on AI for our pleasures and distractions. AI algorithms feed us our posts on Facebook, products on Amazon, and movies on Netflix. They have become our easy dopamine “fixes”. Educators are now wondering: will AI chatbots like ChatGPT or Mediatek’s “Taiwanese GPT” [1] become the education equivalent of the effortless, quick “fix” for students?
Our view is that without a new “intelligent education”, we will let AI algorithms become smarter while we become dumber.
In recent months, educators have been mulling three ways to deal with this looming AI threat: ban AI chatbots, do nothing, or incorporate AI usage.
Although some schools have proactively moved to ban ChatGPT [2], there are many ways students can still use AI chatbots and evade detection. The second option ignores the real harm that AI can do to developing minds if left unchecked: students may outsource their mental effort and learning to AI.
The third option recognizes not only the above problems, but also the possibility that AI can even enhance human intelligence.
Dangers of AI
There are potential dangers of AI chatbots such as ChatGPT, Bing Chat, and Google Bard. These chatbots use large language models and advanced natural language processing to provide linguistically accurate and creative responses to users.
AI chatbots can already outperform most humans in writing essays, solving math problems, summarizing readings and books, and even writing code.
Many will be seduced into outsourcing thinking and creativity to AI. But this will make us dumber in the same way that AI in our various technologies constantly “recommends” a never-ending stream of tantalizing distractions. We are already losing this war for our attention, and we are losing our ability to think for ourselves.
And then there are AI “hallucinations”. AI chatbot responses look and sound authoritative but can be factually wrong. In fact, a recent Microsoft report on GPT4, the latest and paid version of ChatGPT, puts the error rate about 30% for many types of inquiry [3].
If students don’t verify the results themselves, they will “create” misinformation one out of every three times. It is bad enough that this can be self-deceptive, but what will happen as more and more AI generated content fills news reports, blogs, Youtube videos and social media posts? Misinformation will proliferate, exponentially.
Everyone, including elementary school students, should understand these pitfalls.
So, where do humans outperform AI?
While AI is excellent at big data analysis and pattern recognition in stable systems, humans are much better at making inferences and dealing with new and unstable situations, relying on human judgement to interpret these situations.
Human judgement, while neither perfect not evenly distributed across people, is the result of 1000s of hours of implicit learning and alignment in how humans tend to interpret and react to certain situations. It depends on both the rational and emotional parts of the human brain.
AI systems, in contrast, are typically programmed to find the most rational and efficient solution, regardless of its impact on humans.
“Intelligent education” in the Age of AI
“Intelligent education” will understand and leverage both human and artificial intelligences.
Traditional education systems, especially examination-based ones like Taiwan’s, focus on lower-order thinking skills, such as understanding, remembering, and applying. This “shallow learning” model emerged from an industrial economic model designed primarily to manufacture factory workers but also to separate the wheat from the chafe to create a small class of university professors and white-collar specialists, like doctors, lawyers and engineers.
This model fails to support a knowledge economy. Most workers in our society are not factory workers. And many white-collar jobs, like legal experts and radiologists, are on the brink of being replaced by more efficient and accurate AI systems.
Education today needs to cultivate people who are able to navigate fast-paced and ever changing environments. These knowledge workers need to be cognitively efficient, adaptable, quick-learning, and collaborative problem solvers who are clear communicators.
These competences require deeper learning, like problem-based learning and inquiry-based learning that encourage exploration, understanding and developing solutions.
Deep learning can be enhanced with the addition of AI chatbots that can help students not just acquire knowledge but also learn how to ask questions and have conversations that lead toward understanding and solutions.
Students can ask AI chatbots to help them understand the problem by asking for as many explanations and examples as needed.
Asking the right questions and following one’s curiosity is at the heart of the scientific process. It is also at the heart of effective AI use. Unfortunately, Taiwan’s education system does not teach students how to ask questions. If anything, it actively discourages it: curiosity often gets trammeled by exams and shallow learning.
But asking questions is the first step towards understanding how to analyze problems. After this, students will need to evaluate answers and iterate levels of knowledge and understanding in order to create innovative and useful solutions.
This “intelligent chat” with AI is a skill that should be formally taught and practiced in schools.
As students learn how to work with AI technology to enhance both their shallow and deep cognitive abilities, they will simultaneously nurture their uniquely human skills of judgement to choose the right solutions for open-ended social situations, and also collaboration and communication skills to create and share these solutions.
There is no shortage of global crises facing humanity: global warming, overpopulation, genetic engineering, data privacy, food supply, expansionist wars, political polarizations, and also AI. We cannot afford to become dumber.
We are hopeful that Taiwan’s education system will become more intelligent and help our younger generations learn how to control and harness the power of AI to develop the super-human intelligence that will be needed to resolve these crises.
Author bios
Nigel P. Daly has a PhD in TESOL from NTNU and researches technology in language learning education. He currently teaches Business Communications at TAITRA’s International Trade Institute in Taipei.
Yung-Chun Chang has a PhD in Information Management from National Taiwan University and is an expert in natural language processing and machine learning for intelligent healthcare, bioinformatics, business intelligence, and language understanding. He is currently Associate Professor in the Institute of Data Science at Taipei Medical University.
References
[1] Huang, E. (April 4, 2023). The story behind Taiwan’s ChatGPT. Common Wealth.
https://english.cw.com.tw/article/article.action?id=3406
[2] Schwartz, E. (February 9, 2023). ChatGPT is Banned by These Colleges and Universities. Voicebot.ai. ChatGPT is Banned by These Colleges and Universities – Voicebot.ai
[3] Bubeck, S. et al. (March 22, 2023). Sparks of Artificial General Intelligence: Early experiments with GPT-4. ArXiv.org. https://arxiv.org/abs/2303.12712
