Artificial Intelligence #93: Could AI be exposing flaws in the examination system?

In my last post,?

ChatGPT: Taking AI from statistical to a linguistic model and making AI more inclusive we discussed why LLMs like chatGPT could make AI more inclusive from a recruitment and skills perspective.?

Continuing that theme,we see an outpouring of concern for the use of chatGPT in education

Already, chatbot has passed the final exam of an MBA programme designed for Pennsylvania’s Wharton School,?

But, could AI be exposing flaws in the examination system?

Most people would agree that the current process of examinations needs to be reviewed in some way.?

The experience of the Chemistry nobel laureate is a case in point!?

"At school I had a teacher that didn't like me and I didn't like him. At the end of the year he decided to fail me. The ironic thing is that the topic was chemistry. I have the distinction of being the only chemistry laureate who failed the topic in high school!" - Tomas Lindahl, awarded the 2015 Nobel Prize in Chemistry: https://bit.ly/2RdmSNe

Rashik Parmar makes an interesting observation

The three laws of digitisation continue in any domain:?

1. What ever can be digitised - will be.?
2. Once digitised the value trends to become free.?
3. The new data from this Digitisation will open new pathways to value.

So, we could ask the question: if exams are digitised - how can we open new pathways to value for exams? ‘

Here are three possible ways:

a)? ?Prompt engineering as a quiz: In essence, LLMS connect everything to everything else and then enhance that output through? Reinforcement learning from human feedback .?Hence, given a concept, chatGPT is essentially providing ideas / connections/common sense on a massive scale.?So, chatGPT is an ideas generator.? If it is taken as a starting point - the process of enriching that output could itself be a test.?In the simplest sense, that's a form of prompt engineering.In this case, the exam is - how effectively can a student enhance the output of chatGPT

b) ? Reflective process: chatGPT is one of the best reflective mechanisms - for example chatGPT os one of the best journals. So, the exam is not one specific exam but rather like reflective coursework done over the duration of of the course??

c) ? Multiple perspectives: If generative AI is a real brain, I think it displays the condition of? Synesthesia. ? Synesthesia is a perceptual phenomenon in which stimulation of one sensory or cognitive pathway leads to involuntary experiences in a second sensory or cognitive pathway.For example, letters or numbers are perceived as inherently colored. Consider the text to video abilities of runway.ai Overall, there are endless possibilities if you creatively combine modalities to create an exam

Today,? exams are seen as sacrosanct - but in my view, they are not inclusive. UPSC in India , Gaokao in China and PSLE in Singapore and leetcode evaluate people in a specific way. They create a few winners but lose out on a wider range of talent in my view.

So, we could ask:

Could AI/chatGPT/LLMs rethink exams to make them more inclusive, overcome the skills shortage and bring back creativity to jobs and economic growth ?

The views expressed here are my own and not associated with any organisation I am associated with