( Note: At least a dozen people have asked me in the last month or so when I plan to write about ChatGPT, the new technological gizmo on the block. There is a lot of noise about it, for sure. There is enough measure of amazement, intrigue, suspicion, skepticism, and uncertainty about ChatGPT’s capabilities and the future it holds for us. There are lots of questions as well: Will it change the way we learn? will writers become redundant? will ChatGPT take away all incentives to think and create? will it replace Google? Is ChatGPT intelligent? Quite frankly, all these questions, doubts, and curiosity about ChatGPT don’t have any concrete answers as yet. Yes, one thing is for sure, ChatGPT is a quantum leap in one of the most researched areas of AI, which is Natural language processing, and from what we have seen in these initial few months of ChatGPT 3.5, the results are unbelievable, almost God-like, and mysterious in its execution. I have been thinking and reading about AI for many years now, as a techie and as a common man, trying to understand how AI is shaping the way we are organizing ourselves as a species. There is virtually no field of human endeavor that has remained untouched by the AI juggernaut, and I don’t think it is possible to reverse the flow now. However, it is time, we think deeply about what it means to be human, and if that is done, then whatever happens in the field of AI can augment what we do. Questions such as what is intelligence, creativity, understanding, consciousness, etc should be widely asked, not merely by researchers, scholars, and philosophers, but by each one of us. Unless we are clear about what makes us human, even a whiff of technological change can easily sway our sense of who we are. In my opinion. If we are to use technology well, we have fundamentally understand ourselves first not in some abstract philosophical sense, but as a member of species among thousands of others on this planet and what it means to live here. In the three parts of this essay, I intend to clarify my own thoughts on such matters, and hope that it may be useful to a few who are on a similar journey as me)
Part 1
If not anything else, ChatGPT, the new chatbot released by the Open AI foundation last November, has created a buzz, like no other tool in recent times. The last I remember such a buzz about a software tool in our daily life was when Google released its search engine at the turn of this millennium, with its sleek interface with nothing but a text box to type in anything we wished to search for. It was a transformative moment in human civilization. Before Google, there were search engines ( AskJeeves was one of my favorites) but nothing came close to the power and scope of what a Google search engine could provide. Within years, the world adopted Google as its trusted, undisputed source of information, and Googling as a verb became part of the standard English lexicon. The prevalent belief is that If google cannot find something we want, that information either does not exist or is blocked by google for whatever reason. Results from Google, for billions of people, have assumed the validity of a religious dictum.
I have been following the work of OpenAI for years now. The charter they have set for themselves is to work on cutting-edge AI ( Artificial intelligence) problems and make the solutions available to the world to benefit from it. In other words, they would not overtly focus on monetizing their solutions but would allow players in the market to use their models to build products and services that can benefit humanity, and of course, make people and the world richer. AI as a field is a fascinating one. In the next part of this essay, I will discuss more on the issues that AI researchers have been grappling with for decades, especially in the process of encoding “intelligence” into systems, and the tremendous strides they have made in this field in the last twelve years. But for now, let us focus on what ChatGPT does. It is a tool that can predict text based on all the texts that have ever been fed into its software model. Nothing more and nothing less. In other words, when we type in something in the ChatGPT window, it scans through a vast number of probable words ( or tokens as it is called in AI) and spits out the one that fits the best with the style, tone, and structure of the original text pasted into it. Now, the reason it is able to generate a response so semantically complete and contextually relevant in most cases is due to the awesome beauty of the neural network that powers it, and due to the algorithms that run on these networks, ingesting and processing all the data that was fed into it. In short, ChatGPT has been trained on data and learned to correct its predictions from the gigabytes of data it was privy to, just as the neurons in the human brain continuously recalibrate themselves based on the inputs it receives. Unlike the human brain, which in creative conditions has the capacity to make unexpected connections, a deep learning tool such as ChatGPT can only predict from what it has “seen” or “learned. So this brings into question the nature of intelligence, creativity, understanding, and its measurement, and what this means for Human beings and AI systems. Does ChatGPT understand what it spits out, or is it just a mechanical computation of probabilities refined to such a fine extent that its responses look intelligent and not really so? We will talk more about this in the later installments of this essay.
Since ChatGPT’s release in November, an increasing number of people have predicted the demise of creative writing by Human beings. Everywhere I go, and any discussion on ChatGPT I am part of invariably focuses on how this tool could turn writers obsolescent. Each time I hear such an observation, I chuckle. I am reminded of a short story by the great English writer Roald Dahl, published in 1958 called “ The Great Automatic Grammatizer”. The story was way ahead of its times, and at this distance, Dahl’s premise seems vindicated with the advent of ChatGPT. It was staggering to imagine how Dahl’s imagination could conjure up the details of a writing machine, a kind of computing contraption, whose mission is so prescient of that of today’s ChatGPT considering that research into AI only began only in the 1950s. Dahl is a fabulous short storyteller and a genius at building tension and letting the steam blow off in the last few lines. His stories often have unexpected or macabre endings. Though Dahl wrote primarily for a young adult audience, his work has acquired a universal appeal and continues to be published and stocked in bookstores even today. He had an uncanny knack for exploring the not-so-nice emotions of the human heart and the unintended consequences of human acts. In the Great Automatic Grammatizer, he tells a story that has the act of writing as its theme. This story is true to his oeuvre. There are only two characters in the story – both of them a little eccentric and a trifle devious perhaps – and their reaction to the writing machine they have built is so close to how we feel today about ChatGPT, this new tool on the block that promises so much and at the same time is a bit scary for what it presages.
If you can get hold of the short story and read it, nothing like it, but for others here is the synopsis of the story: It begins with a conversation between Adolph Knipes, a young and talented electrical engineer, and his boss Mr. Bohlen, the owner of the engineering firm. Mr. Bohlen appreciates Knipe’s seminal contribution to building a powerful arithmetic calculator that has won great acclaim and commercial value, but for some reason, Knipe doesn’t look happy or even remotely satisfied with the achievement. Bohlen is worried that the young man may be thinking of quitting the firm and offers him a two-week vacation hoping this would bring back some cheer in the young man’s face. Snipes reluctantly accepts the offer. Once home, he gulps a mouthful of whiskey and hits the typewriter hoping to finish a short story he had started. It is at this point we come to know that Snipe actually cared very little for his engineering prowess, and all he desired in life was to become a successful writer of short stories. We are told that he has written five hundred and sixty six stories over ten years, and not one was accepted by any publisher. Bending over the typewriter, lost in thought and depression, he suddenly has a flash of insight, a vague formula in his mind that could solve his problem. Can he build a machine that can write for him? Toying with the idea, he thinks, machines can do anything that can be resolved into mathematical terms, but they are incapable of original thought. But who cares? The calculator he had helped build had memory, and the machine was governed by rules that act on memory. Slowly, the insight begins to concretize into specifics. After all, the English language is governed by rules, which for the most part, remain valid and cannot be violated. And therefore, if the storylines, genres, characters, vocabulary, and all other ingredients of writing, including passion, intensity, and emotions can be fed in as inputs, then, it is quite possible to churn out any type of story one needs.
Knipe feverishly works on the specifications and in fifteen days presents his proposal to his boss, who, not surprisingly, is skeptical of the whole idea. Knipe explains the rationale of such a machine to him. He tells Mr. Bohlen that magazines in the country that had wide circulation, have a distinct type of story they publish on a weekly or monthly basis, and the writers of such stories get paid ridiculously high sums for churning out such stuff. Knipe lays out the economics of generating such stories in-house through his proposed machine and selling them to the magazines at low cost, and how the sheer volume of content sold this way will more than makeup for the costs of building such a machine. After all, who wants customized, man-made products? It is the age of industrialization and everything from buckles to shoes, to chairs, is machine-made. Strong argument! Mr. Bohlen’s eye light up, and he gives the go-ahead. Within weeks, Knipes builds the machine with levers, switches, bulbs, and pedals occupying a huge space in the backyard of the building, and it takes about three iterations for Knipes and Bohlen to refine the machine’s output to match the desired quality of writing. In less than thirty seconds, a story is ready for a famous women’s magazine with a curated storyline that can appeal to readers of that magazine Mr. Bohlen is impressed. He smells money. Furthermore, he accepts Knipe’s idea of setting up a literary agency to brand and sell these stories under fictitious author names and to include his own name as one of the authors. Knipes sets up an office that takes care of publishing the stories under different names ( including theirs) and in no time become household names as prodigious authors. The short-story business turns out to be a huge success, and Bohlen spurred by the good name and prestige that comes with being recognized as a published author, instructs Knipes to expand the machine’s capabilities to write full-length novels. Knipe willingly obliges again and is able to achieve that tweak in no time. To write the novels, Like driving a car, Mr.Bohlen has to sit in the middle of tubes and gears and pedals, to control the flow of emotional flow of the story. After fifteen minutes of intense physical activity and a lot of whirring noise from the surrounding metal, the machines spit out page after page of the novel that Bohlen ordered the machine to write. On reviewing the output, Bohlen is a little alarmed that the novel is a little too pulpy and fruity for his taste, to which Knipe responds “ you have pressing a little hard on the passion-control buttons”. The novel is published, and a dozen more novels are generated and released into the literary world – all to good reviews and acclaim
At this point, Knipe suggests to Mr. Bohlen that as their next steps in owning the entire literary landscape of the country, they should now attempt to persuade established writers to stop writing – by paying them a handsome yearly amount, of course – and instead, lend their names to the machine-generated stories and novels. After all -Knipes rationalizes – mediocre writers will soon find that the stories and quality of writing the machine produces is far superior to their own efforts, and therefore wouldn’t object to selling their rights to us; and the few who will decline this offer and are truly creative, and whose work cannot be duplicated by the machines, will not matter, because, they will be so few in number, and our business model can easily absorb that loss caused by their creative writing. What an ingenious idea!
The story ends in typical Roald Dahl style – A third-person narrator, who is perhaps the last of the creative writers alive ruminating on the possibility that his nine children may have to starve to death if he does not sign the contract. The last line of Dahl’s story ends with this prayer by the writer “ Give us strength, O Lord, to let our children starve”. An unexpected prayer but at the same time echoes the painful cry of a creative man whose work is usurped by a machine, and he doesn’t want to succumb to the juggernaut of automation.
Dahl’s story also is also a meditation on the nature of intelligence and what it means. Can machines be truly intelligent as human beings can be? The threat of machines taking over has been there for nearly a century now. With each passing day, the proponents of AI seem to get close to the goal of achieving a general-purpose intelligence that human beings are effortlessly capable of. Is ChatGPT an intelligent tool, or a tremendous computational algorithm that predicts with a high degree of precision in most cases? All these are pressing questions, and it is important that all educated men and women think deeply about these matters. I will take this up in the next part of this essay.