The AI Around the Corner: The End of Civilization as We Know It?
The Creative Conundrums From AOL to AGI, Dial-Up to Deep Learning
Once upon a 1998, a film called You've Got Mail charmed the VHS players of rom-com enthusiasts and technophilic yuppies worldwide with its engaging story and the undeniably profitable chemistry between Tom Hanks and Meg Ryan. The film revolves around two business rivals, Joe Fox (Tom Hanks) and Kathleen Kelly (Meg Ryan), who fall in love through their anonymous online personas, NY152 and ShopGirl, respectively. While their online love life flourishes, they become bitter business rivals when Joe opens a Fox Books “just around the corner” from Kathleen’s The Shop Around The Corner. The film boasts clever storytelling and a nuanced exploration of identity in a burgeoning digital world, amidst the growing influence of the internet.
Fast forward to the present day (you might need to switch to tape #2) and we're living in a world where artificial intelligence (AI) can pass the bar exam, create award-winning art in less than a minute, and, much to my terror as a software engineer, write passable code. Assuming we can survive what AI has in store for us (some don’t think we can), today’s digital war zone makes the concerns of You’ve Got Mail look like prelapsarian child’s play by comparison.
That said, although it may be difficult to imagine in a time when brick-and-mortar bookstores like Barnes And Noble may, or may not, be gasping a last dying breath at the hands of Amazon, and proprioceptive Disney robots are rollerblading around like T-800 murder machines in the making—back in 1998, little bookshops, like The Shop Around The Corner, were as threatened by the “big, bad chain stores” as people of many industries are feeling threatened by AI technologies today.
In a world where we're increasingly bombarded with news about the potential disruption (or destruction) of AI technologies, it's only natural to seek solace in the simpler times like You've Got Mail. But this charming tale of love and dial-up modems offers more than just comfort and nostalgia to people of a particular age (like myself)—it provides valuable lessons on coping with loss, adapting to change, and finding authenticity in a rapidly evolving time.
The End of Western Civilization As We Know It
You know what this is, you know what we're seeing here? We're seeing the end of Western civilization as we know it.
— Frank Navasky (Greg Kinnear), You've Got Mail
The apocalyptic potential of AI technologies has some experts clamoring for caution. Last month, tech leaders like Steve Wozniak and Elon Musk, among thousands, signed an open letter calling for a six-month pause on AI development more advanced than current capabilities. The intention of such a hiatus would be to allow for the establishment of safety and oversight protocols.
Eliezer Yudkowsky, a decision theorist at the Machine Intelligence Research Institute, believes that even this six-month moratorium is not enough. Resembling Frank Navasky's (Greg Kinnear) doomsaying technophobia for anything more advanced than a blender, Yudkowsky sincerely argues that creating superhuman AI without proper safeguards would result in total annihilation for humanity.
While Frank Navasky's eccentric fears are humorously exaggerated in the film, as he collects typewriters and shuns computers with a visceral disgust that most of us might reserve for cockroaches, it is unclear to the average person whether predictions of an AI apocalypse, like Yudkowsky’s, are valid or not. As laypeople, we can only observe while enigmatic AI experts engage in intellectual duels, like massive abstruse kaiju, effortlessly wielding jargon like “inscrutable arrays,” “orthogonal agent” and “paperclip maximizer.”
Despite the use of esoteric and theoretical language, there are clearly tangible concerns arising already. Reports of AI teaching itself unintended skills and scammers using AI technology to mimic real people’s voices in fake kidnapping and emergency scenarios, have created a truly justifiable sense of urgency around regulating AI. Action is being taken at varying levels of seriousness in disparate parts of the world, such as China setting new security rules and Italy banning ChatGPT. Europe and the US are still considering their options.
Some, such as Google's CEO Sundar Pichai, argue for wide-ranging global regulation of AI, while others, like Yudkowsky, suggest measures as extreme as being “willing to destroy a rogue datacenter by airstrike.” Unfortunately, I think this could mean a serious setback to my aspirations of creating an AI service that can convert images of any object into images of cheese.
It appears unlikely that the majority of us will have any significant influence over necessary regulatory legislation or actions within the critical timeframe for decision-making—that is, before the clock runs out. On a brighter note, we may be on the cusp of developing a sufficiently intelligent AI to make the choices for us.
Go To The Mattresses
Minus specifics, it's hard to help. Except to say, go to the mattresses.
— NY152 (Joe Fox), You’ve Got Mail
Just as Fox Books has the advantage of 150,000 cheap books in “a goddamn piazza where people could mingle and mix and be,” AI brings undeniable advantages as well, like efficiency and accessibility. AI-produced art, music, and writing are democratizing the creative process, allowing virtually anyone with no relevant skills to become an "artist," “author” or “developer.” While the prospect of churning out masterpieces with the click of a button seems tempting, one can't help but ponder the implications. Will we eventually render ourselves obsolete, like the horse and buggy or Frank Navasky’s beloved typewriter?
In the face of such a radical transformation, some fear that AI-generated art will erode the foundations of human creativity. Artists are suing Stability AI, Midjourney, and DeviantArt, alleging their AI image generation tools violate copyright law. Similarly, Microsoft, GitHub, and OpenAI face lawsuits for AI code generation tools. The line between human and machine becomes blurrier, and the prospect of a world without human artistry is unexplored and disconcerting, to say the least. The legal aspect is certainly unprecedented.
For some, it seems that the end of artistic civilization (as we know it) might just be waiting in the wings. Of course, doomsayers may be mistaken; after all, every generation has had its own unique end-of-days concern, like Y2K or bell-bottom pants. Will we lose the skills and creativity that fed AI its data, or create even greater works from new capabilities? What is reasonable concern?
In a dinner party scene, when asked about writing a book, Frank Navasky quips, “Uh, it's crossed my mind. I mean, something relevant for today, like the Luddite movement in 19th century England, or something. We should talk. Call me.” While it may seem like these lines are purely comic relief—to label a 19th-century English movement as "relevant for today" in the context of 20th-century New York City—the film is intentionally highlighting that the fear of obsolescence, and how people respond to that fear, is nothing new.
During the Industrial Revolution, the aforementioned Luddite movement emerged as a response to the mechanization of labor. English textile workers, fearing for their livelihood, smashed the machines that threatened their jobs. But history has shown that, despite a few bumps in the road and some unfortunate incidents involving steam-powered looms, manufacturers rebuilt their machines and society eventually adapted to the change. Nowadays, unless I’m living under a rock, concerns about the over-reliance on power looms seem rare.
By briefly highlighting this historical parallel, the film subtly urges audiences to contemplate technology's broader societal impact. It also clearly foreshadows that the film’s war between tradition and progress is not going to end well for tradition. Kathleen Kelly, following NY152’s advice, “goes to the mattresses” and fights Fox Books by organizing rallies, speaking with the press, and enlisting boyfriend Frank Navasky's support. However, like the Luddites, she must come to terms with the loss when her business closes.
The outcomes for the Luddites and Kathleen Kelly's bookstore are far from ideal. As we delve into the uncharted territory of AI creativity, we might encounter significant shifts in how we perceive and produce what was once considered uniquely human. It appears reasonable to prepare for the worst while acknowledging that history and fiction has shown, in times like these, adaptation could remain a possibility.
It’s Not Personal. It’s Business.
It was business? What is that supposed to mean? I am so sick of that. All it means is it's not personal to you. But it's personal to me. It's personal to a lot of people. What's wrong with being personal anyway?
Whatever else anything is, it ought to begin by being personal.
— Kathleen Kelly, You’ve Got Mail
When it comes to generative AI, it's hard not to take it personally. After all, humans have prided themselves on their creativity and ingenuity since time immemorial. That’s been kinda our thing. Yet, as AI continues to seep into our daily lives, one must wonder, is human creativity dead or augmented? Somewhere between?
AI technology's rapid progress is, like many human advancements, driven by economic motives and caters to consumer demands, potentially creating a divide between human experiences and AI efficiency. Whether it’s garden planning, diet formulation, summarizing lengthy articles (*cough*), reviewing written work, preparing a legal defense or writing code—AI tools like ChatGPT could vastly outperform humans in terms of efficiency and cost-effectiveness. I can’t help thinking that this must be the same giddy thrill that grandma felt when washing machines first hit the market.
However, we should be careful not to be blinded by profit and productivity alone. We have to consider the social and emotional impact of our choices. As Kathleen Kelly so eloquently put it, there is value in the personal. Human creativity is primarily autotelic—its experience holds intrinsic worth independent of the output it produces—regardless of a machine's ability to outdo us.
As AI development continues to make headlines, it's only natural for people to theorize and ponder about its implications. Fortunately, people have already been theorizing for some time now. Perhaps the most famous theory is the Turing Test, proposed in 1950 by the mathematician and computer scientist Alan Turing. The test posits that if a machine can engage in a conversation with a human without the human realizing they are conversing with a machine, then it can be said to possess intelligence. While the Turing Test has been influential, it still leaves room for philosophical debates on the nature of consciousness and intelligence in machines.
The Chinese Room thought experiment by John Searle exposes the limitations of the Turing Test, contesting the idea that computers can demonstrate genuine understanding and consciousness simply by imitating human behavior. In this hypothetical situation, an individual who does not understand Chinese is placed in a room where they receive Chinese characters as input, manipulate them using provided rules, and produce Chinese output to be sent outside the room. While it may appear to an outside observer that the person inside the room understands Chinese, they're merely following instructions. Searle's argument is that a computer, even if it passes the Turing Test, only simulates understanding without truly comprehending the meaning behind its responses. In other words, for machines, it’s possible to “fake it” but never “make it.” Fortunately for them, I don’t think they can experience impostor syndrome.
Amidst these theories lies the concept of Artificial General Intelligence (AGI), or machines that possess the ability to truly understand or learn any intellectual task that a human being can do. AGI would not only mimic human creativity but potentially surpass it. Imagine an inverted Turing Test, where you can tell you’re communicating with a machine because it’s too smart to be human. The relevance of AGI is not just about the capabilities it could bring but also the profound ethical, social, and economic questions it raises.
In the end, there is no shortage of theories and papers on AI / AGI, and exploring them would be far beyond the scope of this rom-com-gone-pseudo-academic newsletter article. Whether AI creates in the same way that humans do, or understands the way we do is not necessarily relevant to the impact it can have.
What's important now is that AI can match or exceed human results, and integration into our lives is unavoidable. As we further develop and use AI, it's vital to balance productivity gains with preserving humanity. We must appreciate our unique creative aspects, even if not financially rewarding, while harnessing AI's potential to enhance and broaden our skills.
Do You Think We Should Meet?
We should meet. And we will meet. But I'm in the middle of a project that needs…
(he pauses to think of the right word)
... tweaking.
— NY152 (Joe Fox), You’ve Got Mail
Just as some critics argue that Joe and Kathleen should not have ended up together, many people will say artists should reject the use of AI, take a page out of the Luddite’s handbook and go to the mattresses. However, it is my belief that the steamroller of technological progress will continue to flatten the landscape, whether we like it or not.
Joe and Kathleen found a way to merge their respective worlds and reconcile their differences and perhaps we too can coexist harmoniously with AI. Rather than being replaced, could artists serve as the guiding hand that shapes AI-generated content, imbuing it with the intangible qualities that define the human experience?
The ChatGPT Experiment
To provide a glimpse into the capabilities of current AI models, I conducted an experiment with ChatGPT for this article.
I started by asking ChatGPT to describe my own writing style, by feeding it what I wrote for last week’s Collective Curiosity newsletter, Code: Of Conduct, it had this to say:
The writing style of this article is engaging, conversational, and informative, using a blend of humor, anecdotes, and real-life examples to make the content relatable and accessible to readers.
After patting myself on the back for impressing an AI language model with my “engaging” and “informative” writing, and marveling at its ability to (generously) refer to my irreverent understatements as “humor,” I experimented with whether it could do my own work for me.
Using the same descriptors ChatGPT gave me of my writing, I asked it to write its own version of the same article you are reading now, with the following prompt:
Write an essay comparing and contrasting the concepts of You’ve Got Mail with the benefits and threats of AI technology. Include all the points listed below. Use a style that is engaging, conversational, and informative, using a blend of humor, anecdotes, and real-life examples to make the content relatable and accessible to readers.
[…list of sections / concepts and brief descriptions of what I wrote in the sections above]
Now, I’m neither a great writer myself, nor a “prompt engineer” and I imagine that any one of the thousands of self-proclaimed “AI Whisperers” out there would scold me for not employing the most effective strategies to get the most out of my prompt, but, the results were mostly just ok.
One example output was the following:
Once upon a time, in the wonderful world of movies, there was a romantic comedy called You've Got Mail, starring Tom Hanks and Meg Ryan. It was about two people who met online while their bookstores were in a battle in real life. This enchanting tale has striking similarities to the awe-inspiring yet complex realm of generative artificial intelligence (AI).
… [goes on like this for a few more paragraphs]
As illustrated in the above example, and observed by other ChatGPT experimenters, ChatGPT has a tendency to write fairly basic pieces reminiscent of five-paragraph high school essays. It’s not so bad as Vogon poetry by any means but it’s not exactly great writing. As I tried various strategies, I began to see some linguistic patterns tediously repeated in ChatGPT’s prose:
“Once upon a time,”
“Fear not, dear readers”
“Alas, dear readers”
“Picture this:”
“In conclusion,”
I’ll admit that its persistent attempts to begin essays with “Once upon a time” eventually inspired me to change my own introduction to “Once upon a 1998.” At this point, I’m still not sure if that was a clever choice on my part or if ChatGPT beat me into submission to a point where I no longer hear how utterly unoriginal that might sound.
That Caviar Is A Garnish
Despite their impressive feats, large language models (LLMs) like ChatGPT still grapple with basic tasks and may be much dumber than you think. Following the above example, I made dozens more attempts, followed online guides for AI model prompting, and experimented with varying prompt specificity. Despite these efforts, the outcomes remained lackluster, with the AI's comprehension of my intended message limited by the guidance provided.
It's like choosing between "chicken" or "vegetarian" at a dinner event or ordering an off-menu item at a restaurant—you have some control, but can't fully dictate the outcome or ingredients. However, the food prepared may surpass your own culinary skills and be served more efficiently. Even if I believe I can write a more polished final product, it will take me literally thousands of times longer than the AI which pops out a result in a matter of seconds.
To illustrate further on some of the limitations, prompts for an “academic tone” tend to read more like ChatGPT grabbed a thesaurus, rather than gained a higher understanding of the topic under discussion:
The 1998 romantic comedy opus You've Got Mail, with the esteemed Tom Hanks and Meg Ryan as protagonists, delineates a captivating narrative set in the urban milieu of New York City. The protagonists' virtual rapport develops through digital epistolary exchanges, whilst their respective bibliophilic enterprises engage in corporeal rivalry. This cinematic endeavor bears an uncanny semblance to the exhilarating and intricate domain of generative artificial intelligence (AI).
… [make it stop…]
While I genuinely appreciate the apt use of words like "milieu" or "corporeal" in writing, I was fairly certain I wouldn't be the only one fatigued by the "academic tone." It's both astounding and amusing that ChatGPT referred to emails as "digital epistolary exchanges."
Apart from the somewhat mundane or ostentatious outcomes, the results from more vibrant prompts, such as "speak like a pirate" or "write an Eminem rap song," offered some entertaining content, including this snippet:
Yo, listen up, I'm 'bout to drop a tale,
You've Got Mail and AI, we gonna set sail
You've Got Mail, you've got AI,
Two worlds collidin', we're reachin' for the sky
Ultimately, out of all the results I generated specifically for this newsletter article, none of them were something I’d particularly enjoy being accused of writing. To borrow from Joe Fox, anything it generated, at the very least, "needs tweaking." Admittedly, this offers some relief, as spending five minutes instead of five days on this article would rob me of the satisfaction derived from the Sisyphean labor of reworking my writing until I have no energy left.
Despite its limitations, ChatGPT achieves amazing things. Its ability to understand context and blend unrelated ideas is both impressive and fun to explore. ChatGPT is great at helping with idea development and outlining, making it a valuable research tool. I can see how much of a time saver it can be to get a rough structure built out with it. It's excellent at summarizing articles providing assistance when a fresh set of eyes is needed for editing, and I can't possibly ask my wife to read my work "just one more time. Final draft, I promise."
Note: I have also delved into ChatGPT's abilities in generating and handling code. As someone who writes code for a living, I have a lot more nuanced opinions on the matter that I would prefer to explore in a dedicated post in the future.
Conclusion
In conclusion, dear readers (you know I had to do it) the rapid evolution of AI like ChatGPT has already brought about significant changes in various fields. While the prospect of AI competing with human creativity may seem daunting, it is essential to recognize that humans and AI can (read: must) coexist harmoniously and complement one another.
You've Got Mail serves as a reminder that differences can be reconciled and that new relationships can be formed. Similarly, as AI increasingly integrates into our lives, we should strive to find a balance that preserves the essence of human creativity while also reaping the benefits of AI-generated content. By using AI as a tool to augment our skills and knowledge, we can work towards a future where both human and artificial intelligence can thrive together.
Have you come this far and enjoyed what you read?
If so, please consider sharing and subscribing!
It would mean the world to me if you did!