Embraceable Doomsday: Or How to Make AI Your Best Friend, Maybe -- Umberto Tosi

It's getting so there's no future in futurism. The future nips at our heels faster than we can predict it.

Take the latest chatter about ChatGPT. Professors say it's the end of the world. Standard fare for Scici writers - warning of misuse, while speculating on our replacement by the instant proliferation of this AI phenomenon. 

ChatGPT explains itself succinctly as any human:

"ChatGPT is a large language model developed by OpenAI that can be used for natural language processing tasks such as text generation and language translation. It is based on the GPT-3.5 (Generative Pretrained Transformer 3.5) model, which is one of the largest and most advanced language models currently available.

All this is free of charge, mind you. (Or course, nothing is truly "free" in tech world.) Free means getting billions of users to provide content for free - content in quality and quantity promising to be worth trillions of dollars in return. Data is to Internet enterprise as petroleum is to the old industrial economy. 

Continuing, Chatty explains: "One of the key features of ChatGPT is its ability to generate human-like text responses to prompts. This makes it useful for a wide range of applications, such as creating chatbots for customer service, generating responses to questions in online forums, or even creating personalized content for social media posts."

It already has attracted hordes of users and caused a stir among professionals, social critics academics in the scant time of its launched on November 30, 2022.

 Audrey Niffeneger

Many educators and technology watchdogs are sounding alarms amid the buzz over ChatGPT: NPR cites Edward Tian, a Princeton computer science and journalism major as saying: "I think we're absolutely at an inflection point, ,,,This technology is incredible. I do believe it's the future. But, at the same time, it's like we're opening Pandora's Box. And we need safeguards to adopt it responsibly."  

Stephen Marche, author of "The Next Civil War," checked out ChatGPT and declared that "the college essay is dead" as a pedagogical tool (ironically in an Atlantic essay.)

"So what?" Astrophysicist Neil DeGrasse Tyson commented in a recent interview about ChatGPT. "You can cheat, but at the end of the process you'll be an idiot who didn't learn anything." The result confirms what Tyson famously said a decade ago: "When students cheat on exams it's because our school system values grades more than students value learning."

"Are you Skynet?" A BBC Channel 4 interviewer asked ChatGPT with serio-sarcasm. The chatbot denied this with a boilerplate definition of its limitations and parameters.

Chatbot maintains utility and political neutrality as a natural language tool that - (for the present) - can't exceed the pre-training of its creators. The response was at best naive and at worst deceptive, in that machine learning would require such a chatbot to learn from its own interactions beyond initial programming. Plus, ideological neutrality isn't neutral. It's a position, most frequently favoring the status quo, for better or worse.

So far, chatbot's originators avoid the idea of regulation, checks and balances.  More frightening: consider the sources. Volatile, egocentric billionaire Elon Musk is an OpenAI backer.  

I asked ChatCPT to "write a 500-word blog presenting the pros and cons of ChatGPT for fiction writers." Out of curiosity and laziness, I hoped the result would be good enough to develop into a credible post at least
What followed was informative but mundane.

"ChatGPT, (which stands for Generative Pre-trained Transformer), is a language model developed by OpenAI that has the ability to generate human-like text. It can be used for a variety of compositional, editing and research tasks, drafting fiction or nonfiction.

 Alan Turning - Alessandro Nassiri / Wiki commons

"One of the biggest pros of using ChatGPT for fiction writing is its ability to generate realistic and coherent text." 

 

I raised my eyebrows. Good fiction needs to hold its readers with more than realism and coherence.

ChatBot added: "This can be incredibly helpful for writers who are struggling to come up with ideas or who want to explore different plot lines and character development. ChatGPT can generate entire scenes, dialogue, and descriptions that can serve as inspiration for a writer's work. It can also help writers who have writer's block."

This was getting to sound more like a pitch, even though the ChatGPT is free.

Copy editing, or proofreading would be more accurate terms than  "writing" for its text processes. Though useful and time-saving, it's no substitute for the complex interaction between professional editors and authors thus far.

ChatGPT's responses aren't shy of self-criticism. "One of the cons of using ChatGPT for fiction is that it can be difficult to control the output..."

It's garbage-in; garbage-out. Also, it can be almost laughably literal-minded. It's text are strictly based on the input, no more, no less.

 Neil DeGrasse Tyson

Note fellow bloggers: Metricsmule SEO expert Eric Palazzolo offers a practical YouTube tutorials on using ChatGPT to generate posts online. It compares Chat's results with Google searches. "Google is still king with search," he says, "But ChatGPT did a good job. ChatGPT gives you a leg up" when it comes to convenience, he adds.

Palazzolo points out that ChatGPT's  lack nuance, tone, style, and even elementary complexity, but offers strategies to work around this. "ChatGPT will get you only part way to what you want," he says. "That can save you a lot of time."

"Use ChatGPT for outlines... but don't depend on it a hundred percent," he advises. "Add your personal elements." He points out that using ChatGPT's texts verbatim probably would get your work rejected as plagarism by Google's smart SEO systems.

I'll go a step further and advise anyone using this tool to translate whatever one gets from ChatGPT - no matter how impressive -  into one's own personal voice. Each of us has a unique voice and the ability to write in other voices - of characters, for example. But the end result is what separates craft from art, what's routine from what will be compelling.  

Instead of fearing AI, we should embrace it as a tool, say many professionals who have explored ChatGPT. I agree. The quality of this AI tool's responses depend on the quality of the requests we give it - the more detailed and specific the better. This is known as "prompt engineering" - an AI speciality that is rapidly gaining followers. It's not mysterious. Think of having a bright, eager assistant who needs well defined instructions in order to do your bidding, with lots of clarifications and guidance, the more the better. 

What we say we want isn't always clearly understood by others. Same goes for AI. ChatGPT works best when we follow definitions with examples of what we seek, and request a specific structure and style if possible. 

It's likely that ChatGPT and what follows it will grow exponentially more powerful. It's early days. Given the nature of self-learning AI, if ChatGPT will improve as it deals with more and more inputs, perhaps even gaining the ability to generate characters and communicate with users with more responsive and personal speech - as permitted. (People's limits and references can be a minefield, just like in real life.) But the direct relationship between "prompt engineering" and AI results is likely to endure.

This is only the beginning, however. We're waiting for the other shoe to drop -- or, I should say, shoes to drop. OpenAI is fast developing ChatGPT-4, built on data its is getting from users now and promising to be exponentially more powerful that number 3.  And a bigger boot will drop if Google - with its massive market share and competence - overcomes risks and launches its own natural language AI bot.

OpenAI has a head start, but may be quickly outpaced by competing giants with massive resources. Remember when Yahoo and Dogpile were popular search engines? 

ChatGPT points out that it has been trained on a large dataset of text, which means that it may contain biases and stereotypes present in the data. This can be problematic for writers who want to create diverse and inclusive stories, as ChatGPT may inadvertently perpetuate these biases and stereotypes in the generated text. There's the rub. 

I asked it for quotes from authors who have used ChatGPT so far.

"I apologize, but I don’t have access to quotes from authors who have tried ChatGPT specifically. However, I can tell you that many authors have been using AI-assisted writing tools for some time. ... "

ChatGPT then stumbled into error: with a hypothetical list of "...Authors who have tried AI-assisted writing tools:"
 

"I found the AI-generated text to be a great source of inspiration for my writing. It helped me to come up with new ideas and to explore different plot lines." - John Smith, author of "The Time Traveler's Wife."

Still, I was intrigued. I imagined a best-selling author whose books were all ghosted by an AI bot, as science fiction turning into a real-life nightmare.
 
This is only the beginning. Humans, after all, learn through dialog as we grow up.

I recall the words of an IBM computer professor who taught a programming class I took back in the 1970s. "There are two ways to succeed," the IBM engineer told us, "Be the first at something. If you can't be the first, then be the best."
 
Maybe we can instruct the baby as it grows up. In that vein, I experimented with the "chat" aspect by reminding ChatGTP that "Audrey Niffenegger wrote "The Time Traveler's Wife," not 'John Smith.'"

"I apologize for the mistake," Chatty responded humbly. "You are correct that "The Time Traveler's Wife" was written by Audrey Niffenegger," without embarrassment on being corrected.

I urged it to check three other erroneous author citations in that section.

Again, it retorted: "I apologize for any confusion, the other 3 author citations that I provided in that section were not real authors and their books titles. They were examples used to illustrate possible opinions.
"You did not make that clear," I wrote. I suggested that it label such examples as hypothetical explicitly, or by inference.

"You are correct, I should have been more explicit by calling the list as "some hypothetical examples," ChatGTP answered. I felt embarrassed at the bot's seeming humility even though I knew it was simply part of its programming. The bot's uncanny versimilitude notwithstanding, ChatGPT seems more like an efficient tool than a sentient being or a superbot intent on world domination. 

ChatGTP is still green behind its virtual ears - although like a nervous intern, bound to surprise us in the near future. 

This closing exchange reveals a lot. 

I asked it: "Are you the ghost of Alan Turning?"  

Chatty responded with a routine, sanitized bio of Turing that left out some essentials including the brilliant scientist's being persecuted to death under Britain's then homophobic laws.

"No, I am not the ghost of Alan Turing. I am a computer program called ChatGPT, created by OpenAI. Alan Turing was a pioneering computer scientist, mathematician, logician, and theoretical biologist, who made a significant contribution to the field of artificial intelligence and computer science in the 20th century ..." etc., etc.  

"Thank you," I said, determined to maintain civility for my own sake, whether with a bot or a being. "You might want to work on satire - specifically delivered tongue-in-cheek. I was referring to Turing's 'The Imitation Game' in which he predicted that a computer would one day be able to convince a human evaluator that it was a person."

Again, Chatty responded humorlessly (unless this was pulling my leg in return): "Yes, you are correct. Alan Turing proposed the "Imitation Game" as a test to determine a machine's ability to exhibit intelligent behavior that is indistinguishable from that of a human. The test, now known as the Turing Test, is still widely used as a benchmark for measuring a machine's ability to exhibit human-like intelligence."

All and all, ChatGPT-3's debut is impressive, especially in technological agility. The chat bot, for example, takes only minutes to code a program in C to solve complex mathematical problems. Nevertheless, my own brief encounter ChatGPT3 literally flunked the Turing Test -- unless it has been programmed to dodge Turing Test references by nosy writers. 

 

=========================================================================

Umberto Tosi's novels include his highly praised, Frank Ritz, Hollywood noir detective mysteries The Phantom Eye, and Oddly Dead. plus his story collection, Sometimes Ridiculous, plus book the historial epic Ophelia Rising, High Treason, Sports Psyching and Our Own Kind. His short stories have appeared in Catamaran Literary Reader and Chicago Quarterly Review where he is a contributing editor. His essays and articles have been published widely since the 1970s in print and onlin

--------------------------------------------------------------------------------------------------------------------------

Enjoy Umberto Tosi's Hollywood noir detective thrillers: The Phantom Eye  and Oddly Dead, out in August, 2022.

 

 "Tosi writes with tremendous style and a pitch perfect ear for everything that makes the classic noir detective story irresistible. Philip Marlowe and Lew Archer, make room for Frank Ritz!" - Elizabeth McKenzie, best-selling author of The Portable Veblen and managing editor of Chicago Quarterly Review.