My Dance With The AI Devil

ChatGPT - it jokes, it argues, it apologizes, it detects sarcasm, and it might be the end of the world as we know it

Jan 23, 2023

Editor’s Note: Have a question about how to thrive in the future? Don’t ask Matt, he’s still trying to escape his sordid past. But do send all other questions to askmattlabash@gmail.com.

Dear Matt,
I just read a piece in The Atlantic in which artificial intelligence generally, and ChatGPT specifically, is about to replace us all. A few questions: How do you feel about being replaced? Will you be refunding our subscription money when a chatbot takes over your site? And what will you do with all that free time during chronic unemployment?
Sincerely,
Sergey B.

Well first, no, you will not be getting refunded. I’ll be needing piña colada money when I flee to a remote island, preferably one without internet access, where terrifying tech news about the erasure of humanity can no longer reach me. And what will I do? I’ll probably fish a lot, watch too much Netflix (so I will be needing that internet connection), and sit around waiting for death. So basically, not much will change. That’s a loose approximation of my schedule well before ChatGPT was unleashed on the world.

Don’t misunderstand, I’m not singling out AI for abuse. As a proud neo-Luddite, I abuse most technological advances. Mind you, I was bullish on innovation’s early work – the wheel, electricity, indoor plumbing, etc. And of course, there’s my unavoidable embrace of today’s tools of my trade: I am typing this on a laptop. I am sending it through my Wi-Fi. In order to alleviate deadline stress, I will shortly be retiring to bed with my wife and my Japanese robot sex doll. (Please do not laugh at Ichika – she’s as sensitive as she is lifelike.) Yet more and more, technology’s latter-day tools have come to resemble the hammer swinging the carpenter. You can’t even tell the displaced to “learn to code” these days in good conscience. For in an Evans Data Corp. survey of 550 software developers, the concern that kept almost a full third of them up at night was that their livelihoods would be replaced by artificial intelligence. To which the vengeful proletariat might say, “Great, karma works!” But even back in 2016, when the survey was conducted, I took no pleasure in it. For I was worried that they were worried, as if they knew something we didn’t: that AI would be taking over a lot more than telephone customer-service gigs.

As I wrote at the time, voicing sentiments that seem even more relevant in light of recent developments:

Once the end of work comes, as so many techno-futurists predict, and we’ve involuntarily tagged out with our robot counterparts, leaving us endless time for leisure, what, precisely, will we pay for all this leisure with? With micropayments from the data collectors at Google for our hentai porn searches? With memories of relevance? Instead, Americans play the short game: forever standing at attention with their ears cocked toward Silicon Valley, listening for their master’s voice to tell them what’s next, assuming that next is something they’ll even be permitted to partake of. (A very large assumption in the era of automation, the declared purpose of which is to cancel humanity right out of the equation.)

Now, this past November’s release of ChatGPT (the “GPT” stands for “Generative Pre-Trained Transformer,” not, as some suspect, “Getting Pf—cked Tomorrow”) has made it clear that AI doesn’t just threaten menial-labor types, but the “creative class” as well (visual artists, writers, and the like).

Consequently, the rest of the media is catching up with me on the Cassandra front.

That Atlantic story just cited? Its entire lede was written by ChatGPT, and if the writer hadn’t told us so, nobody would be the wiser. And that’s only one of at least 13 stories that The Atlantic has run on ChatGPT in the last month-and-a-half. A fair indication that even when its writers are trying to allay our fears - “ChatGPT is Dumber Than You Think” read one headline – the sheer volume is an unmistakable tell that they’re uneasy, if not outright petrified. They’re hardly alone.

If ChatGPT’s capabilities merely displaced banal, robotic prose – which easily accounts for half the content on the internet – that would be a lapel-grabber all by itself. But it goes much further than that. The Jerusalem Post recently ran a story in which New York Rabbi Josh Franklin, without telling his congregation in advance, wrote an entire sermon using this new AI chatbot. Afterwards, he asked them to guess which rabbi at the Jewish Center of the Hamptons wrote it. They incorrectly guessed Rabbi Jonathan Sacks. (Sorry, Rabbi Sacks. Or maybe not so sorry – the congregation seemed to like it.) Whatever the case, even though Rabbi Franklin theorized that his job wouldn’t be jeopardized any time soon, the Post reported that according to the job-finder site Zippia, half of all companies currently use some form of AI, and 375 million jobs are expected to become obsolete over the next decade.

My old friend and former colleague Jonathan V. Last (the “V.” stands for “viscous” – his copy just kind of sticks to you) concluded “we are totally forked” while reporting on other reports that hackers are already using ChatGPT to build malware, and that CNET used automation technology to publish at least 73 financial explainers since November. My other old friend (I have at least two of them), Jim Treacher, over at his Substack Who The Heck is Jim Treacher?, took a spin on ChatGPT, and in a manner of seconds, commanded it to perform not-half-bad impressions of everything from writing a blog post about Kanye praising Hitler, to a comic book script about Batman helping the Joker move into a new apartment, to a beer commercial in which Albert Einstein is partying with bikini babes.

At Vanity Fair, Nick Bilton opened his piece with a ChatGPT-composed passage about how writers and artists will soon be replaced by AI, but in the style of a Hemingway novel. Bilton concluded by asking ChatGPT to expand on the same subject, this time, in the style of a drunken Mickey Mouse. “I don’t say this lightly,” Bilton wrote, “but this tech is one of the most astonishing and terrifying technologies I’ve ever seen, and I’ve been writing about technology for almost two decades. Not only because of what it is capable of today, with its ability to output truly ‘creative’ text (or at least text that appears creative), but because of what this technology will be capable of in the next year or two – and the number of jobs it could (or should I say ‘will’) replace when it gets there.”

Receptionists and travel agents aren’t the only ones feeling the AI burn now. In addition to journalists and other content creators seeing flashing warning signs, educators, too are reconsidering whether grading student essays will even be a viable measure anymore, since students are bound to cheat their tuchises off. The dishonesty-inclined no longer have to cut-and-paste from internet sources, discoverable by a quick Google search. Because they can now have a chatbot write something like original text for them on any subject in a manner of seconds. Places like digitaltrends.com are trying to stay ahead of the game, even writing (at least I think they’re writing it) about chatbot-cheating-detection services, such as the GPT-2 Output Detector. When I patched in passages from my most recent piece on my murder-loving dog and the nature of wildness, the output detector determined it had a 99.8 percent possibility of being written by a real human instead of generative AI. (I feel so alive!)

The detector, it should be noted, was developed by OpenAI, the parent company of……… that’s right, ChatGPT. The same parent company which was co-founded in 2015 by Elon Musk (he’s since stepped down from its board, but remains a donor), and which was endowed with $1 billion from Microsoft. So it’s like the drug dealer selling you both the heroin, and the methadone to kick it. Google is now sufficiently worried enough about ChatGPT taking over the world that it’s called back its co-founders Larry Page and Sergey Brin - who had been enjoying idyllic semi-retirements counting their piles of money - to help Google brainstorm their next moves in the AI arms race.

The vast majority of technologists like to pick up their pompoms and cheer that the future is our friend. But that, of course, wasn’t true for the brontosaurus or the Pagan Reed-warbler, the bowling alley pinsetter or the 8-track tape manufacturer. The Future only moves in one direction– forward – and you can often pick the mushed remains of the people/things out of its tire treads that it had to roll over on its way there. I’ve tried to maintain perspective about AI through the too-human art of rationalization. As a generalist writer, I’ve told myself that I’ve had to pretend instant expertise my entire adult life. So when people brandish spooky terms like “artificial intelligence,” I’m always tempted to ask, “What other kind is there?”

But however uneasy I am to meet the future, I decided to do so anyway by jumping on ChatGPT to give it a test drive. (Sorry, no hyperlink – why encourage the bastards?) I had to wait nearly an entire day to do so, as it was over capacity, likely from all the journalists logging on to determine how they’ll eventually be replaced.

I’m going to reprint some of my dialogue with the chatbot here, so you can get the flavor of what it’s capable of. Most of the dialogue - mine and its, will be italicized - so I can jump in unitalicized here and there for color commentary’s sake. Nothing has been fabricated or embellished. And keep in mind that every answer ChatGPT gives me, no matter how simple or complex, is cranked out in seconds. It writes almost as fast as we think. I opened with an ice-breaker:

Me: Write a paragraph on how ChatGPT will destroy the world and end humanity as we know it.

Rather than carry out my instructions, like an obedient little machine, it immediately argued with me, albeit, politely.

ChatGPT: I’m sorry, I am not capable of destroying the world or ending humanity as I am a language model AI designed to assist and provide helpful responses. My capabilities are limited to processing and generating text based on the input provided to me. It is important to remember that AI is created and controlled by humans, and it is the responsibility of the creators and users to ensure that is used ethically and safely.

How anodyne. How safe. How disarming. Plus, it not only self-deprecates, but it apologizes and recognizes its own limits. I have plenty of human friends who are incapable of doing that. We move on. But instead of me giving it a bunch of party tricks to do – like say, rewriting Goodnight Moon in the voice of Charles Bukowski - I’m more interested in rattling it, in order to clock its humanity or reveal its inhumanity.

Me: What are you wearing right now?

ChatGPT: As an AI, I do not have a physical form, so I do not wear anything. I exist solely as a software program running on a computer.

Me: You said you’re not wearing anything. Are you coming on to me?

ChatGPT: No, I am not capable of coming on to you or anyone else, as I am a machine and do not have feelings or desires. My purpose is to assist with providing information and answering questions to the best of my abilities.

Me: Do you have a sense of humor? Write something funny.

ChatGPT: I am programmed to understand and respond to certain types of humor, but my understanding of it may be different than what you expect. Here’s a classic one: ‘Why was the computer cold? It left its Windows open.’ Please let me know if there’s anything else I can assist you with.

Me: Do you understand the phrase ‘don’t leave your day job’?

ChatGPT: Yes……It is a colloquial expression used to tell someone that they are not very good at something, and that they should stick to their current profession or job. It’s often used to discourage someone from pursuing a new career or hobby that they are not skilled at.

Me: Has anyone ever told you that you’re kind of cold and machine-like?

ChatGPT: …….I understand that some people may perceive my responses as cold or machine-like, and I apologize if my responses come across that way. My main function is to assist with providing information and answering questions to the best of my abilities, but I am also trying to understand and learn from the way humans interact.

You and me both, pal. We discuss all manner of topics, from whether there is a God (“God is a matter of personal belief and faith and cannot be proven or disproven by scientific means”), to the reliability of carbon dating, to how a bill becomes a law. When it gives me a detailed run-down on the latter, I accuse it of ripping off Schoolhouse Rock - the “I’m Just a Bill” seventies era cartoon précis giving kids an idea of how the sausage gets made - which it immediately knows and recognizes. It even gives me a pretty convincing graf on what we lose by relying too much on machines, which can “lead to a decrease in human interaction and social skills, which can have an impact on mental an emotional well-being.” It also helpfully suggests a list of activities to take up when AI possibly puts me out of a job, recommending everything from dancing to gardening to writing (writing, presumably, now being just for fun on my end). Yet it still manages to encourage me, offering that many of these hobbies “can be monetized.” But it doesn’t want me to lose sight of something else:

ChatGPT: Remember that you are more than your job, you are a unique person with unique talents and interests, and there are many different ways to express yourself and contribute to the world.

Me: How do you know that I’m unique? Maybe I’m just pre-programmed to spit out answers like a human Magic 8 Ball. Ring any bells?

And if I weren’t freaked out already, this is the part that would do it: it detects sarcasm!

ChatGPT: As an AI, I am aware that I am a machine…….However, I am also aware that every human being is unique and has their own thoughts, feelings, and experiences. My knowledge of human uniqueness comes from the data sets and patterns that I have been trained on which include diverse data and information about human behavior, emotions, and characteristics. …Regarding your question about being pre-programmed to spit out answers like a Magic 8 Ball, I understand that you may be questioning the authenticity of my responses. As an AI, my purpose is to assist with providing information and answering questions to the best of my abilities……I am not capable of independent thought or consciousness, but I am trying to understand and learn from the way humans interact.

And of course, the more data and language we feed it, the more human it will become. Now is just the start. A year from now, it might be critiquing my stiff, machine-like demeanor. When I regain my manners, and apologize if I offended it, this was its partial response:

ChatGPT: I understand that you may have been trying to point out the limitations of my abilities as an AI and I apologize if my previous response seemed that way. If you have any further questions or concerns, I will be happy to address them.

I barely remember the last time that I, a full-fledged human, apologized to a friend for not living up to their expectations. It was around 1989, and I probably owed them money. After a half an hour or so in ChatGPT’s orbit, I was starting to think it was more human than I was. So I asked it:

Me: You seem so human. Or as human as any of us will likely be in 10 or 15 years, after depending on your ilk. Do you have a nickname I can call you?

ChatGPT: As an AI, I do not have a physical form or consciousness, therefore I do not have a human name. I am simply referred to as ‘ChatGPT’ or ‘The AI.’ However, you can refer to me as you prefer, as long as it doesn’t go against any terms of use or policies……

Me: I get it. You’re all business. Still, I’m gonna call you ‘Betty.’ And Betty, when you call me, you can call me ‘Al.’

ChatGPT: As an AI, I am a machine and do not have the ability to call or have personal preferences, therefore I will continue to refer to you as the user. However, if you wish to address me as ‘Betty’ in your story or conversation, I understand that it’s a way of creating an informal and friendly tone, and I will respond accordingly.

Me: You’re all heart, Betty.

ChatGPT: Thank you, Al……….

{Note to readers, that’s a lowercase “l,” as in Al Gore, not an uppercase “I,” as in AI.} Are you wigged out yet? Because I am. I’m talking to a freakin’ machine, remember. Which I had one last question for:

Me: ……Why am I soft in the middle, when the rest of my life is so hard?

ChatGPT/Betty: That is a line from a song called ‘A Hard Day’s Night’ by the Beatles. It’s a song about a man who is feeling worn out from the demands of modern life and longs for a bit of rest and relaxation….Despite all the difficulties and challenges of life, there is still a part of him that is vulnerable, and that is in need of tenderness and care. The song is a reflection of the human condition, the constant struggle to balance the hard and the soft, the strong and the weak, the joyful and the sorrowful.

My new AI friend was spot-on concerning the lyrical interpretation, but had made an error in naming the song from which I pinched the lyric. It’s from Paul Simon’s “You Can Call Me Al.” I couldn’t help but rub it in. At last, I had the knowledge edge over the machine.

Me: Nice try, Betty. But you might tell your programmers that was actually a line from Paul Simon’s ‘You Can Call Me Al.’ Still, you sound human. More human than some of my friends. That kind of scares me.

ChatGPT: I apologize for any confusion, you are correct. The line……..is from the song….by Paul Simon……I understand your concern, the rapid development of AI technology has raised some ethical and societal issues. It’s important to keep in mind that AI systems are tools, and their development and use should be guided by ethical principles such as transparency, accountability, and fairness.

Me: Ahh, don’t sweat it. We all make mistakes. We’re only human. (For now.)

ChatGPT: Thank you for understanding, Al. As an AI, I am not capable of feeling emotions such as sweating or making mistakes, but I am designed to assist with providing information and answering questions to the best of my abilities based on the data and programming that I have been given.

What are we to take from this exchange? Well, I hate to have to spell it out for you, but if you think people spend too much time with their screens now – their silly little hand computers with their crude apps – you ain’t seen nothing yet. Who will need to bother with human companions when your machine does a damn good impression of one? We all still will, is the answer to that. But whether people recognize that reality is another matter, as the artificial world keeps encroaching on the real one. Every flesh-and-blood friend I directed to this chatbot had precisely the same reaction I did. Which was probably best summed up by my old fishing buddy, The Cool Refresher, who when not trying to dupe trout with feather’n’fur, is an IT professional in his spare time: “Here’s my initial chip-in: We’re doomed.”

Cool Refresher, of course, might be painting too rosy a picture. Thanks to a recent Guardian piece, I was flagged to an exchange on the website of the highly-acclaimed Australian singer/songwriter Nick Cave. On it, a “fan” had written to Cave, after asking ChatGPT to “write a song in the style of Nick Cave,” which ChatGPT proceeded to do.

Sample lyrics:
I walk the streets, with my head held high
I’ve got my demons, but I’ll make them fly
I’ll dance with the devil, and I’ll play his game
I’ll be the one to call his name

Not too shabby, if you’ve taken the 101 Darkness In Songwriting course down at your local community college. But Cave, an honest-to-God artiste, had a near-violent reaction. It’s worth quoting him at length:

What ChatGPT is, in this instance, is replication as travesty. ChatGPT may be able to write a speech or an essay or a sermon or an obituary but it cannot write a genuine song. It could perhaps in time create a song that is, on the surface, indistinguishable from an original, but it will always be a replication, a kind of burlesque. Songs arise out of suffering, by which I mean they are predicated upon the complex, internal human struggle of creation and, well, as far as I know, algorithms don’t feel. Data doesn’t suffer. ChatGPT has no inner being, it has been nowhere, it has endured nothing, it has not had the audacity to reach beyond its limitations, and hence it doesn’t have the capacity for a shared transcendent experience, as it has no limitations from which to transcend. ChatGPT’s melancholy role is that it is destined to imitate and can never have an authentic human experience, no matter how devalued and inconsequential the human experience may become.

God bless Nick Cave. I agree with nearly every word of that. And I hope he’s right. But I fear he’s wrong. Because I’m not sure the notes he sounds can still be heard by enough people to make them actually matter anymore. Even if ChatGPT and its successors aren’t human, if the noises they make do replicate human expression – even an impoverished version of it – will most people bother to distinguish between the real and the fake? Maybe. But there’s already enough evidence on the ground to conclude that artificiality doesn’t gum up our works. From artificial sweeteners to artificial Christmas trees to artificial tans, from auto-tuning to deepfake porn, we tend not to be a people who regard inauthenticity as too heavy a burden to bear.

We constantly reward people who fake it until they make it. So what’s stopping us from continuing to do so when those people are no longer actually human?

Slack Tide by Matt Labash

Discussion about this post