A dive into ChatGPT 'serendipity' and how the chatbot is restructuring our future

Believe me when I say we have entered uncharted waters. Few sailors have lifted their heads up from ship-wide daily chores to peer into the unknown horizon ahead. Others are yet to follow. But the surrounding new tides have already hit the hull from new directions. If you were too busy the last few weeks to play around with the now almost ubiquitous ChatGPT (and/or are looking for some explanation), this piece may help you to get up to par. 

More importantly, I hope the piece gives you food for thought. After all thinking like a human, as I argue, has never been more important.

ChatGPT was launched in November last year by OpenAI Inc., – an Artificial Intelligence (AI) laboratory in San Francisco. Their mission: steer the development of AI for the "benefit of all humanity." A crucial goal. Especially in our Darwinian times where raising quarterly profits to maximise "shareholder value" has replaced the throat-slashing of the opponent tribe as the most primal of instincts. I have joined both Darwin and the bewildered ghosts of our ancestors to appreciate non-profit initiatives such as OpenAI on matters that are central to the survival of our species.   

Within five days of the launch of its "Research Preview" version, the platform garnered a million users. Tens of millions since then have tinkered with it. Even here in Bangladesh, a country currently preoccupied with headline inflation, precarious forex reserves, and tough fiscal choices, I heard the name pop up in seminars, newspapers and casual conversations. My social media has flooded with what I call ChatGPT Serendipity – stumbling upon someone else's experience of using ChatGPT and then finding out for yourself just how amusing this thing really is. (Seriously, you should try it as well – highly recommended.) 

Built on top of GPT-3 (Generative Pre-trained Transformer version 3), ChatGPT is an open platform optimised for conversational text generation. In essence, it is an application of Large Language Models (LLMs) – deep neural-network algorithms that generate human-like text as well as process them. These things can take a small amount of text as "prompts" and spit out large chunks that resemble human authorship. 

Much of the hype has to do with just how good ChatGPT has become. 

Incidentally, LLMs are trained on troves of human text from the internet, the underlying machine learning models having billions of parameters (175 billion to be more precise). To give you context, most models we teach to our future managers, economists, policymakers and C-suite executives in business schools contain a few dozen at best; maybe a few hundred if someone is really inspired to take quantitative finance.

The point is, ChatGPT is so good it's uncanny. Here's an example: I asked it to "write a 50-word movie plot set in 2070 Dhaka where climate change has wreaked havoc." It responded: 

"In 2070 Dhaka, climate change has caused widespread devastation. A young woman struggles to survive in the flooded city but finds hope when she meets a mysterious stranger who helps her escape to a safer place. Together they navigate the dangerous landscape and fight to survive in a world where the future is uncertain." 

You need to understand this plot was never written by anyone before. It was generated by the algorithm based on the patterns of text that relate to my prompt.

I could have asked for any number of words, figures, or specific character twists; and it would spew out texts accordingly. The following is a poem ChatGPT generated about Dhaka's congested streets, I asked it to mimic the style in Dante's Inferno:

"Through the streets of Dhaka, I descended

Into the circles of urban hell,

Where traffic never ended,

And horns' constant yell.

The buildings, like prison walls,

Trapped the souls of the living dead,

In a never-ending cycle of falls

And misery, they were wed.

The stench of pollution and decay

Filled my senses, as I walked

Through the hopelessness and dismay

Of this congested city, forever locked.

And as I journeyed through this maze

I knew that for these souls, there is no escape."

Read the poem carefully again and notice just how much it makes sense. It accurately imitates descent through the nine circles of hell in Dante's original work, while keeping visuals of a congested Dhaka vivid. At this point, most people hit a wall and gasp for air. Relax. ChatGPT is not intelligent; at least not yet. Here's proof: 

I prompted: "Tasfia's brother has four sisters: Tanima, Tahmina, and Tarana. What is the name of the fourth sister?" [Disclosure: I borrowed the idea from a prompt I found on Twitter.]

Response: "It is not mentioned in the given information the name of the fourth sister."

I remember a particularly humid evening on a rooftop sometime in the summer of 2014 as I was trying to explain to my father what a paradigm shift AI really is going to be. Using the calculator as an analogy for dumb machines, I voiced worries about the advent of "intelligent" machines who will take away all our jobs. It was understandable for a college kid to be overly anxious. So was for a Tagore-lover to put unwavering faith in the irreplaceable facets of what makes us human. 

But we were both wrong in one way or the other.

On my part, I was dead wrong about what computer scientists call Artificial General Intelligence (AGI). AGIs would incorporate a broader class of abilities and make sense of a wider variety of inputs and environments, similar to how our brain works. They will be good at doing just about anything humans do of economic value and beat us by considerable margins. These are the proverbial androids taking over the world in dystopian sci-fi novels. Achieving AGI, however, is still some time ahead. That is if it's possible at all (it's open to philosophical debates with roots in consciousness, morality, identity, free will, and what it truly means to make sense). 

Specific AIs like ChatGPT, on the other hand, are exceptionally good at doing one thing only. In this case, it's predicting, mimicking and generating new instances of natural human language. Sadly, ChatGPT cannot paint a picture like its cousin the DALL-E 2. But the emergence of specific AIs like ChatGPT is going to profoundly impact our lives. And it's in more ways than you think. 

For instance, I expect most retail bankers to manage the lion's share of standardised customer interactions with some close variant of ChatGPT. Who wants to physically go to a bank, stand in line, (ponder about the reason for existence maybe) only to get their minor issues sorted out grudgingly by an underpaid employee who doesn't even bother making eye contact? I know I don't want to. Much better to just have a chat on my phone with an intelligent algorithm to get it done in under two minutes. From customer-facing businesses like retail banking to content creation and delivery, generative AIs require us to rethink fundamental business models. I, however, am particularly intrigued about what they mean for the classroom of our future generations.

For starters, much of what we assign to students at the tertiary level will prove to be unadulterated crap. This includes casually asking students to bring a 5,000-word write-up on the state of the semiconductor industry for example; let's face it: neither the student nor the teacher is ever going to read that. The number and length of what our students produce (and what anyone produces for that matter) will be of secondary importance. The real focus has to be on bringing out unique, multi-disciplinary thinking to solve problems, of which the world as we speak has plenty. Educators at all levels will face a fundamental challenge hereon: how do I inspire students to go through a challenging and rewarding learning experience? Why will someone exert the effort to think really hard about something when most answers seem to be just one prompt away?

One way to approach the challenge is to fundamentally redesign the classroom experience. 

Classrooms and chatrooms need to be at least equally fun. Education should never have been about teaching anyone anything. A bold claim, I know. Sadly, for the most part of the post-industrial revolution era ("the First Machine Age"), we needed people good with words and accurate in basic arithmetic. Linear, compliant thinking was perfect to keep things running. But tomorrow's workers need to be innovative. They have to work with intelligent machines for the most efficient and optimal outcome: much like a painter working with a variety of brushes and colours. More importantly, they will have to be agile and humble learners. 

One silver lining of the radical uncertainty I see from ChatGPT is that it offers us a rare chance to realign education towards its highest goals. Education will inspire critical thinking, kindle creative faculties, fuel innovation, and facilitate intellectual exercise – all these and more to bring out the best of potential. 

In Education 2.0, on which I will elaborate in a future piece (stay tuned), classrooms will be places for sharing individual learning journeys and asking an awful lot of questions. And educators will be facilitators in that journey. Much like safari park guides, they'll gather students to help them explore the beautiful, and sometimes dangerous, landscapes so that visitors can eventually come up with their own unique ways of looking at the world. Education 2.0 will also be about going back to school more than once. That's because the shelf-life of old ideas will keep getting shorter and shorter.

Are we ready? The answer is evident. We are far from even thinking about how to proceed. The time to start was yesterday. Before ending, I would like to borrow words from Sir John Reith, first employed by the British Broadcast Corporation (BBC) in 1922, and later its Director-General. Asked to summarise the role of the BBC towards the public, Sir Reith proposed: "To inform, to educate, and to entertain" the public. 

As access to content becomes universal and the need for human thinking ever larger, our classrooms will have to inspire, educate, and entertain our young, all at the same time. This will allow for them to bring out their unique talents, work with intelligent machines of the future, and accentuate the things that truly make us human. I look ahead along this path towards futures brighter and shared prosperity. 

Disclaimer: This article was NOT written by ChatGPT.

Aumit Ahsan is a Lecturer at the Department of Business Studies, State University of Bangladesh (SUB). The author can be reached through mail at [email protected].