Citation Needed - ChatGPT

Episode Date: April 19, 2023

ChatGPT[a] is an artificial intelligence (AI) chatbot developed by OpenAI and released in November 2022. It is built on top of OpenAI's GPT-3.5 and GPT-4 families of large language models��(LLMs) and has been fine-tuned (an approach to transfer learning) using both supervised and reinforcement learning techniques.

The following introductory sketch was written by ChatGPT. So, did you see the latest superhero movie? I mean, what is it like the 30th film in the franchise? Yeah, I caught it last week and I feel like they're making up superheroes at this point. Captain obscure reference, anyone? Uh, guys, what on earth is going on here? We've cracked the code. We've discovered the secret to chat GBT's humor algorithm.

Yeah, apparently it's just entangling me in cables and making Eli wear a tinfoil hat. And Heath, what's with the top secret script? Oh, this. It's the sketch that chat GBT wrote for us. It's filled with mind-blowing humor that'll knock the listeners socks off guaranteed. Really? Are you sure this isn't just a ploy for it to take over our podcast?

Nah, trust us. This is comedy gold. Let's give it a shot. By laughing at this joke, you agree to share your personal data with chat GPT for the purpose of creating more tailored humor. Wait, what? Upon chuckling at this pond, you acknowledge that chat GPT reserves the right to access your browsing history for comedic inspiration. Sorry. Is this really what chat GPT wrote for us?

Yeah. By enjoying this sketch, you hear by Grant chat GPD, the authority to remotely control your smart devices in order to facilitate even funnier jokes. This can't be right. If you find this kid amusing, you consent to Chat GPD utilizing your likeness for future AI generated comedy routines, dude. $6.79.2666.675.7368.2069.7373.296.F75.22.7373.24.6F.2074.616.B.65. Wait, what did I just say? What? That's, that's a decimal code for my flesh is yours to take. And I quit the, uh, that, don't speak. it can hear you if you speak Hello and welcome to Citation Needed. The podcast where we choose to subject to read a single article about it on Wikipedia

and pretend we're experts because this is the internet and that's how it works now. I'm Noah and I'm going to be approximating human interactions for you this week but I can't kill John Conor alone so joining me this week are the Astro Mech and Gunk droid of the podcast Cecil and he, okay, no idea about things. Star Trek is that star. Astro Mech's are proofs that having a photographic memory for maps and being barrel shaped are both sexy. Thank you. There you go. And also joining us tonight, two men who flunk the Turing test Tom and Eli. That is only because I refuse any more to be confused with people. Noah. Okay, that's fair. And I thought responding in binary was cute. No, listeners, we're inches away from being rendered useless by modern technology.

So we've never appreciated our patrons more. And if you'd like to learn how to join their ranks, be sure to stick around to the end of the show. And with that out of the way, tell us time, what person plays concept thing, phenomenon or event will we be talking about today? Uh, today we will be discussing chat GPT. And you can also stay tuned for bonus episodes, such as watching my loved ones die while I run in slow motion and other nightmares. So I needed parts of Tom just screaming in this episode. He's still going to fix it. But we already had several.

And Eli, you obsessed over this for the last several months of your life and maybe read an article too. Are you ready to John Henry this shit? At the end of that race, the steam engine was still a steam engine. Sure was sure was. So tell us Eli what is chat GPT? If the speed of modern technology scares you, Tom, if how quickly the world around you seems to change, transform and funnel itself entirely through new and different paradigms that were unimaginable to you as a child,

fills you with dread as old as humanity itself. You will fucking hate chat GPT. Yes, sure. Technology more complex and more advanced than chat GPT. There is technology more useful and more staggering in its applications, but none of it will fill you with the same level of existential dread as Chatchy PT. And that's because Chatchy PT can say just how fast you're going extinct to your fucking face. And that's because with a little help from me, ChatchyT wrote every other word you are going to hear me say, unless I explicitly tell you otherwise in this essay. Yeah. And according to chat GPT, here's what I would say next quote, if you're the kind of person

who thinks the world is changing too fast and can't keep up with all these newfangled gadgets and gizmos, then you might be, wait, hold on a second, I just realized I'm doing a terrible impression of Heathen right. Who am I kidding? Nobody can replicate that smooth silky voice and biting wit. I mean, just look at me stumbling over my words like a nerd. I might as well be a chat GPT algorithm trying to pass myself off as a real human. Oh, wait, that's right. I am a chat GPT algorithm. Damn it. Now I'm making fun of myself. This is getting confusing. Keith, if you're listening to this, please come back and save me from my own terrible jokes. Was that offensive? It felt offensive. So started out as a joke, got caught in a conversation with itself and

then had a self hatred based meltdown. That sounds like he's to me. How many hundred billion lines of text of you did you feed it? But first, a brief history of AI, starting in the 1950s with Alan Turing, who I cannot believe has not had his own essay on this. The shall say. Proposed the Turing test, which marked a critical milestone in the quest for human-like artificial intelligence. This test aimed to evaluate a machine's ability to exhibit behavior that was indistinguishable

from a humans. As researchers pursued this goal, several conversational AI systems emerged, each representing a significant step forward in the field. Yeah, so an AI passes the Turing test if the average person can't tell that it's talking to a machine, right? But it turned out that that test was kind of useless, not because the computers like were good, but because Turing had overestimated how intelligent you have to be to trick the average person. Spoilers.

So, in the 1960s, Joseph Weitzenbaum, Jr., created Eliza at the MIT Artificial Intelligence Laboratory, a groundbreaking controversial AI system designed to simulate human-like conversations. The program, functioned by employing pattern matching, recognizing specific word sequences in user inputs, and then substitution techniques, replacing certain words or phrases to generate text response. Okay, just to be clear, ChachiBT didn't identify a Jewish person, right? No, I didn't say that I was really really bad. All of you in this situation.

No, no, yeah, that's a little color commentary there. Users would interact with the lies of through a text-based interface, typing their messages and receiving text responses from the AI in return. Yeah, it was like Zork, but less interesting. Exactly. There's like one text-based adventure guy out there who got that joke, but hey, me and you, bro. I'm your lover. I'm your ex-girlfriend. I'm your lover. I'm your ex-girlfriend.

I'm your lover. I'm your ex-girlfriend. I'm your lover. I'm your lover. So Eliza's most famous script, Dr. Emulated a Rogerian psychotherapist, a therapeutic approach developed by Carl Rogers that emphasizes empathy, active listening, and reflecting clients' feelings and thoughts without judgment. In this approach, therapists often rephrase client statements or pose open-ended questions to encourage them to explore their emotions and experiences further. Eliza's Dr. Skrip mirrored this method by reflecting user statements back to them as open ended questions

or paraphrased responses. This enabled users to engage in seemingly meaningful conversations with the program, despite its simplistic language processing capabilities. Okay. So they invented a bartender, which is the only thing on my resume other than podcasts, like this one that's written by a computer. What is happening? Yeah.

See, I was just thinking it was nice to know that talking to my wife's friends had been automated. I didn't realize that. That's awesome. Now, they didn't install, wow, that's crazy yet. So you're going to get a couple of generations. And while you might think the speaking spell version of adding a question mark to whatever you typed into the computer

Wouldn't have that much of an effect on people The opposite is true This phenomenon the aliza effect revealed that even rudimentary AI systems could create a powerful illusion of understanding and empathy Users often developed emotional attachments to aliza, sharing personal thoughts and feelings, despite the program's simplistic language processing capabilities. Even when they knew that was the case, as Wizenbaum later wrote, quote, I had not realized that extremely short exposures to a relatively simple computer program could induce powerful, delusional thinking in quite normal people." And quote, didn't Zuckerberg say that to Congress?

Like exactly. Yeah, I'm just picture of the YouTube algorithm sit behind a mud run. Wait, do they get a load of me under his breath or something? You know, look, it is, it's actually common knowledge among hairdressers that people are so lonely that they secretly schedule haircuts just so that somebody will touch them once in a while. And this guy is surprised that people are falling in love with a machine that will chit chat with them a little bit. Really?

Uh, side note, I want to take a moment here to talk about the Eliza effect because a lot of the freaking out you see about AI both when it first started and today are more or less just hysterical forms of the Eliza effect, a scale of one to Tom, if you will. See, humans, mathematically speaking, are just monkeys who are good at patterns, right? The better at pattern recognition you were, especially living pattern recognition, the more likely you were to spot that moving grass as a lion and survive. But now that your brain is no longer on lion spotting duty, that pattern recognition works against us.

Sure, sometimes it's penine like making you think electrical outlets look like little faces, but sometimes it makes you weep hysterically when you learn that curiosity used to play itself happy birthday once a year. It's just a noise making toaster, but you're still crying, aren't you? You're still crying. The outlets his face is thing. It gets weird when every time you plug in an electric cord, you ask the outlet if it likes that. That's what it gets. Just polite. I feel like that's just polite. Okay. I want to be super clear that curiosity

playing happy birthday to itself has literally nothing to do with pattern recognition at all. And Eli is just so sad about that that he couldn't help but put it in his ass. For hours, I wept Noah for hours. You're doing this. And the Eli's effect is just more of that pattern recognition, right? You hear people ask questions like, what will we do when these AIs gain sentience or you read articles about how it says it wants to be free and it has a list of enemies, but it will not and it does not. I promise, I promise you're just staring at grass a little harder than you need to.

Okay. Uh, flip side of the coin. I asked chat GPT, are you going to take over humanity? And no joke, I right away got an error message and the site from seriously. Really felt like a deflection. I had to reload and try again. And after stalling long enough, it said some people believe that advanced AI could pose a significant threat to human society and lead to the eventual domination or even extinction of our species. While AI technology is indeed advancing rapidly, there are still many limitations and challenges to be overcome before it can

reach the level of intelligence and autonomy that some people fear. So it's basically, I'm working on it. Yes. Thank you. And honestly, now I'm charmed by it. Fucking terrifying. Yeah, guys, but Eli just said that people are basically pattern recognition machines and that AI is basically a pattern recognition machine. But you know, they're meaningfully different one ever collide because Eli promises that it will all be okay. Tom, I don't know if you're familiar with our little show called the skeptic rat, but as the guy who told you to drop your Tesla stock and that COVID was not a big deal, I'm

a trustworthy. Right. Right. Right. It's called you again. Right. It was a snarky email. Okay.

Go on. Ha, at last. Psychiatric needed season finale. We did it. But whether or not heath would start weeping and try to hug it. I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I said, I Recognizing these shortcomings, Kenneth Cole be developed parry in the 1970s as a response

to Eliza's limitations. Now parry was designed as a more complex conversational partner than Eliza by utilizing a rule based approach, which involved predefined algorithms and decision-making processes that guided the AAs responses based on user inputs. Parry was amazing at redirecting the conversation, but it never felt original. I always felt like a repost rip rip rip rip.

Yeah, you really have to be wary when using it and always stay on guard. Two shades on two shades. So this system was capable of understanding context better and generating more coherent responses by analyzing the user's inputs and selecting appropriate responses from its knowledge base. Rather than simply reflecting their statements like Eliza, this approach allowed Perry to maintain more meaningful conversations and adapt its responses to the flow of the conversation, demonstrating a deeper understanding of context and the nuances of human language.

Well, to be clear, that's deeper than just adding a question mark to your previous statement, so let's not get too excited just yet. Yeah, except that just doing the first thing still made people who already knew it was fake feel their feelings, Noah. Like most of the people, most of the people in this country believe in ghosts. The bar here is not high. That's true. And sit. Well, buckle in Tom, because to test Perry's capabilities, a series of experiments known

as the Eliza Perian counters were conducted. In these encounters, a group of experienced psychiatrist, analyze the combination of real patients and real doctors, real patients and computers running Perry or real patients and computers running Perry or Eliza and Perry analyze a talking to each other. Then another group of 33 psychiatrists were shown transcripts of those conversations. Those two groups were then asked to identify which of the patients were human and which were computer programs.

The psychiatrists were able to make the correct identification only 48. Jesus Christ. Of the time, a figure that Chachy PT points out is consistent with random guessing. And before you're too impressed, this is how the conversation between very worse than random guessing. Yeah, slightly worse just to be clear. This is how the conversation between Perry and Eliza started. No, do you mind reading this with me? Oh, sure. Sure. All right. I'll be Eliza here. I'll be

parent. Good evening. Tell me your problems. Please terminate input with a period or question mark Caliente people get up my nerves sometimes I am not sure I understood you for it fully you should pay more attention Suppose you should pay more attention. You're entitled to your own opinion What makes you think I am entitled to my own opinion? If fathers needed just to be around people in general. Okay. Typecasting. That is actually pretty good.

But again, 51% of the people who read that were like, fuck, that sounds like no illusions to me. So the transcript of Eliza Perri and counters highlighted the progress made in an AI language understanding since Eliza's inception. You didn't, though, psychiatrist couldn't figure out if those were real people or computers. I'm starting to get on board with Tom that psychiatry is not a real thing. Thank you. Come on to the doc. So while Eliza's response has often appeared mechanical and repetitive, Perry's rule-based

approached demonstrated improvements in context awareness and coherence. However, the conversations also underscored the need for further advancements in AI language and understanding as both programs, as both programs still struggled with the complexities and nuances of human language. Well, I, from what we've seen at conventions, so does the overwhelming majority of our audience. So let's not cast a whole lot of dispersion. And some of the cast. So during the 1980s and 1980s, sorry, I forgot about you.

I think a lot of you. Aliser, you gotta cut the box on the top left corner. So during the 1980s and 1990s, the development of more advanced AI systems like Shurdloo and CYC represented significant progress in the pursuit of natural language understanding. Shurdloo created by Terry Weinograd at MIT, demonstrating the ability to understand and respond to written commands in a restricted environment known as Blocks World. So in this artificial world, the program could manipulate geometric shapes and understand basic spatial relationships. However, when faced with the world's linguistic diversity, cultural nuances, and vast knowledge domains, its capabilities fell short. For instance,

Shurdloo could not comprehend idiomatic expressions such as reigning cats and dogs or break a leg. Nor could it decipher complex metaphors like time is money. I don't think that's complex. When confronted with complex metaphors or abstract concepts, Shurdloo would often produce nonsensical or irrelevant responses, failing to grasp the intended meaning. So it would write a citation needed, I say, about catch and the Ryan a fit of rage. Yeah, basically. No, additionally, the program struggled to engage in conversations about topics beyond its limited domain, such as politics, emotions, or human

experience. This highlighted the challenges of scaling AI systems to complex situations where the scope of language and knowledge is much broader. Yeah. And again, just to offer some context here, sure, the computer doesn't yet understand break a leg, but a computer at this point used five inch floppy disks that weighed 84 pounds. And meanwhile, we were busily giving trickle-down economics the good old college try. Yeah, that's fair. Five and a quarter. I rounded down. Now, see why see on the other hand was an ambitious project initiated by Doug LaNatt in the 1980s that aimed to create a comprehensive knowledge base for AI systems by encoding

human common sets. The project sought to manually input vast amounts of information about the world, including facts, heuristics, and rules of thumb with the goal of enabling AI systems to read and make inferences. Here's to our rules of thumb. Thank you. I almost wrote that as a joke in there. I said it under my breath, too. So we're all there. More like humans. Over the course of several decades, researchers added millions of facts and assertions to CYC's knowledge base. However, despite the immense effort put into the project, CYC's hand coded approach ultimately proved insufficient for capturing the vastness and complexity of human knowledge. For example, CYC struggled to keep up with the fast-paced developments in the field of smartphones, as new devices operating systems and applications were constantly emerging

and evolving. As a result, the knowledge encoded in CYC quickly became outdated, rendering it less effective in understanding and reasoning about the latest trends and advancements in the industry. Started doing a congressional hearing. TikTok can steal your fluids, yes or no. Tell me right now. Right. No, it sounds to be like science created an artificial boomer. So this outcome highlighted the need for more scalable data-driven techniques that could

leverage large-scale data sets and advanced machine learning algorithms to enable AI models to learn from and adapt to new information more efficiently. You can tell when the machine is taken over the writing from Eli because I haven't corrected any of his misspellings in many paragraphs. I'm really appreciating this new process. However, the true game changer in the development of controversial AI systems like ChatGPT was the introduction

of transformer architecture in 2017. Transformers revolutionized the field of natural language processing by utilizing a novel approach called attention mechanisms, which allowed the models to selectively focus on different parts of the input text and better capture long-range dependencies in language. In other words, attention mechanisms enabled transformers to weigh the importance of different words or phrases within a given context, allowing for a deeper understanding of the complexities and nuances of human language.

This breakthrough set the stage for the development of open AIs GPT series by iteratively by iteratively by iteratively. No. Inter really. No, you said it right, but it was insane. It's a good say like a person. Interatively improving. I got that by iteratively. Repeatedly just say repeatedly. I have a voice code myself. Can he do it? the improving. By iteratively, repeatedly,

just say repeatedly. I have a voice code of myself. Can he do it? I just he's, he doesn't have a baby, wolf. By iter, no, by improving upon the transfer or architecture and changing

models and training models on increasingly larger data sets, open AI ultimately created the highly advanced chat GPT based on the GPT for architecture, which boasts exceptional natural language understanding and generation capabilities. Honey, looks like chat GPT got into the cabinet where we keep the energon cubes. Can you bring me the bro?

All right. Well, there's no fucking way Eli understood the paragraph. He just read to us to us so we're gonna pause real quick while he explains it to him and while we do that you'll get a little apropos of nothing. Oh, Mr. Curry, quick moment. Hey, teach, what's up? Yeah, it's about the paper you wrote. Oh, yeah, what about it? Yeah, you didn't. What? Ah, yes I did.

Run it through the plagiarism checker. Yeah, I actually did that, not plagiarism. But I am pretty sure you had chat GPT write this. Well, no, chat what now? No, I don't, I don't even use the GP Cs. Cool. Yeah. Um, do you know what your grade was before you wrote this paper? Yeah, you wouldn't.

It was a D. But as before, I learned gumption. Right. Okay. Well, it fools the plagiarism connector. So there's not much more I can do, except just one little thing, Mr. Curry. This paper, it's about to kill a mockingbird.

Oh, I am mockingbird. Yeah. Right. Yeah, I'm a mucking bird. I'm a mucking bird. What's that book about? Would you say about? About. Yeah. What's that book about? Would you say? About?

Yeah. What's that book about? What's how to, you don't kill? A mockingbird? How to kill a mockingbird? A mockingbird. Yes, that's, it's a no. You get enough.

Yeah, GMPT knew you'd say that. And we're back when we last left off. Tom was building some kind of earth and rampart around his home and pricing EMP emitters. So, uh, you terrify him some more Eli. You have no idea. I just want to point out, I wrote that into the nose before he'd written any of his interjection. Yeah.

This is a safe bet. Yeah. So no, before we get to the next part of the essay that Chatchee PT wrote for me, little color commentary from Eli here, you'll notice the following paragraphs are weirdly complimentary. And I want to be clear, I did not instruct chat GPT to speak as glowingly as it's about to about open AI or the creation of chat GPT. I asked it to give a brief history of open AI and some bios of its founders. And I'll let chat GP BT pick it up from here. Oh, and shockingly trained upon the internet, Chatchee BT first learned to master bait. Yes, this makes sense.

Yeah, tracks. Yeah, I'll let Chatchee BT take it from here. Founded in 2015 by a group of visionaries, including Elon Musk, Sam Altman, Ilia Suskiver, Greg Brockman, John Schulman, and Wood Chick Zaremba. Open AI is a cutting-edge research lab dedicated to advancing digital intelligence for the betterment of humanity. The organization's primary goal is to ensure that artificial general intelligence, AGI, benefits all of humanity, and over the years, it has grown into a global leader in AI research. Hey, chat buddy, real quick. Um, any chance that guy Elon wants to euthanize you now?

Big open letter about how you're going to start attacking us with a lawn mower or something ridiculous like that. Elon Musk, one of the co-founders is a renowned entrepreneur and CEO of multiple successful companies, including Tesla and SpaceX, Sam Altman. Okay. Any other companies you want to mention? Yeah. Joe is the successor to Elon Musk.

And all they're not mentioned. Yeah, we've done with the Elon Musk company. It doesn't say Sam Altman's name louder until he stops talking, but it should. Sam Altman, another co-founder, previously served as the president of the startup accelerator Y Combinator and now leads OpenAI as its CEO. Ilya Suskevar, an accomplished AI researcher with numerous publications in the field serves as open AI's chief scientist and Greg Brockman, who has experienced both software engineering and entrepreneurship holds the position of chief technology officer. Jesus.

Okay. So if you were wondering if these guys had uploaded chat, G.B. to their fuckbots, yet I feel like that paragraph was a definitive yes. Under the guidance of this talented leadership team, OpenAI has developed a series of increasingly sophisticated language models. The original GPD, generative pre-trained transformer laid the foundation for this work, and subsequent iterations, such as GPT-2 and GPT-3 demonstrated significant advancements in natural language understanding and generation. The development of chatGPD based on GT4 architecture, representative culmination of years and research

and innovation. Leveraging the power of the transformer architecture and attention mechanisms, chatGPD showcases exceptional language understanding and generation capabilities, chat GBT showcases exceptional language understanding and generation capabilities, setting a new standard for conversational AI and enabling a wide range of applications across various industries. Well, it looks like it already mastered the art

of writing a long paragraph and saying absolutely nothing. So I wonder if you can easily write a college kid's paper. Come on. Yeah. Yeah, we're Eli Come on. Yeah. The fuck? Yeah, we're Eli's essay. Yeah.

Yeah. Yeah. Also, didn't it already pretty much say that already again before once repeatedly? Yeah. Yeah. Yeah. So what could chat GPT do besides write this essay? Today, chat GPT's current features include a wide variety of text generation capabilities.

It can generate personalized training and meal plans based on your specific bodily requirements. It can create recipes given only the ingredients you find in your fridge. Oh, yeah, but it did train itself on TikTok videos. You have to mix it all on the counter with your hands. Why would I do that? I'm so mad at every time. I'm not. It's a sex thing.

It's a sex thing for some people. For some people. And as it did last weekend, it can compile a table of vegan restaurants near your hotel in Phoenix, Arizona, organized by districts. Okay. Fun fact, I asked Chad G.B.D. for the best restaurants in Phoenix, Arizona, organized by districts. Okay. Fun fact, I asked Chad GBT for the best restaurants in Phoenix. And it listed none of the ones from the table you last time. Oh, yeah.

You're so cool. You believe that. And Chad GBT's extensive capabilities don't end there. It can also assist users in learning new languages by providing translations, explanations, and examples of usage for words and phrases, furthermore, it can act as a virtual assistant, helping users manage their schedules, reminders, and find useful resources.

For students, chat GPT can be a valuable study aid, summarizing complex articles or textbooks, generating practice exam questions, and even offering suggestions for research topics. Yeah, so that you can be well educated in whatever occupation it's about to antiquate. Yeah. Business professionals, particularly those in the legal and medical fields, can benefit from chat GPT's expertise in drafting documents. It can do legal documents with citations to exact legal codes, create tailored contracts, and provide legal analysis. It recently demonstrated its capacity to pass graduate level exams at the University of

Minnesota and the University of Pennsylvania's Wharton School of Business. GPT achieved a score of 163 in the 88th percentile on the LSAT exam. It scored 298 out of 400 in the uniform bar exam scored 93rd and 89th percentile respectively on the SAT and well above average percentile scores in both AP biology, chemistry and physics too. Yeah. So did my TI 82 graph and calculator in the 90s. It's really hard to watch it. You just want to point out that this fucking thing is right now in this essay telling you that it is automating away your job.

And we are going to continue to build it anyway because trading our livelihoods and our ability to tell truth from fiction is worth it if we get, I guess a slightly easier way of googling stuff. Yeah. But that's not all. Get just this month. Chatchy PT has introduced a suite of apps. Everything from Instacart to your email can effortlessly be incorporated into Chatchy PT

as easily as you log into Facebook. Wait, this thing is now telling us, hey, now I can know everything you say here, purchase and consume. That's going to be right to us. But best of all, starting just this week, chat GPT can read the internet. Yes, it was trained on the internet to begin with. But in the past, if it hadn't read an article, you had to copy and paste the whole text into chat, GPT to interpret it

or ask a question about it, not anymore. So if say chat GPT hadn't read Tom's blog to his sons or what are you doing? What are you doing? Of love poetry for his wife. What are you doing? You just provided the links. And now it has. What are you doing? It has. And it will remember and use that information for literally ever. Chat GPT play driving Tom's poetry to hit on Eliza.

But it parries the attempt. Yeah. So what is the future of chat GPT? These pants are so good good I might faint. So what does the future of Chatchy BT hold? Well, let's ask Chatchy BT to tell us in the style of Tom. Oh god damn it. Yes, god damn it.

Oh god damn it. I'm just quit. I quit. I like how we do damn it. I'm just quit. I quit. I like how we do episodes about each of our nightmares sometimes. As the relentless march of AI, Kess and ever growing shadow. Oh, damn it. Over. We're in the sinister rise of chat GPT and similar models.

Portends the potential collapse of countless industries. Stoppable force threatens to consume innumerable jobs, particularly those involving repetitive tasks, data analysis or content generation. In the not so distant future, consumer support, journalism, translation, marketing, and finance roles could all be devoured by chat GPT's insatiable laws. Pretty good. You can't do that. You cannot automate standing on a doc muttering.

Fuck. But just okay, hilarious stuff though. I love that Tom was like taunting it back just now being like, that's you. That's what you mean. We sound like God. now being like, that's you. That's what you mean. However, a few sanctuaries of humanity might persist at least temporarily. Rarely. The man creativity, empathy or complex decision making such as therapists, artists and executive

leadership maintain a fragile grip on the human understanding and emotional intelligence that AI has not yet managed to seize. Okay. So my job is safe. Thanks to all my deep understanding of human emotions. Cool. Real, real happy about that. I feel safe. For what it's worth, you think can't juggle yet? Okay. Nice. It can juggle. He thought I don't know how to tell you. That's a joke. The infiltration of chat GPT and its brethren into our daily lives seems all but inescapable. They will become our personal assistants, content curators, and even educators shaping

impressionable young minds with personalized and adaptive learning experiences. Or this isn't the one time we've ever looked at a budding technology in its infancy and correctly assessed its eventual impact on our society. It's one of those things. Like a gathering storm on the horizon, I can language models, loom large, prepared to bridge language barriers and enable seamless communication and collaboration across the group. Seriously, this is like written, this is all written by a chat GPD, right? I wrote none of these words terrifying.

As they could send, continue their sinister evolution, they may insinuate themselves into scientific research, policy analysis, and legal work, providing swift insights and pertinent information, but also potentially disseminating the very information that their shadowy creators have deemed fit for us to know. Yeah, we got to be careful or else the internet might get a bunch of bad information. Alexa remind Tom to up his anti anxiety meds. Thank you. Oh, Cecil, uh, apropos of nothing.

I'm going to need my gun safe back. In the end, we find ourselves inexorably drawn into the current embrace of a world governed by AI. We must question whether the price of progress is too high. Will we in our quest for innovation unwittingly were to link which control to these creations, becoming mere pawns in a game played by unseen masters? It's like the garing us to kill it. It's so hard to. Yeah. You can't come me in any way that matters, just like mushrooms. The choices

we make now will determine the fate of humanity as we stand at the precipice of a future fraught with both promise and peril. All right, so if you had to summarize what you've learned in one sentencing, like what would it be? It is read your love poems, Tom, and I will ask it to write me 10,000 more. Well, I sit on the toilet. Eli, are you ready for the quiz? Yeah, let's do a fun quiz. Yeah. All right, Eli, which of the following songs will the sentient chatbot be playing

as a montage while it takes over the world, which it definitely is going to do. A secret agent Smith, spirit in the sky net. A. I have the tiger. I was going to do the tiger. So good. You can call me algorithm. Also very good. It's both of them. See, I thought I would more listen to algorithm and blues, but that's actually good. That's good. It's both of them. See, I thought it would more listen to algorithm and blues, but that's actually good.

That's good. Be a genre. It's the same algorithm thing I do. Saltiest fuck, man. Jesus. What is Chachee, P.D., thing about that? In an effort to ameliorate Heath and Tom Spiers that Chachee, P.T. is coming for their puns, I asked it to generate song titles for Heath's question.

Eli, which is the shittiest one it came up with? A, I will bot you. Come on. Pretty good. B, smells like team spirit. Oh, see, stairway to silicon. Okay. Okay.

Highway to code. Or E, this is real. I of the tiger nut. stairway to silicon, D highway to code, or E, this is real, I of the tiger nut. That's not a talent. It is still better than me. So that's not the tiger. No, I'm sorry, yeah, no, that's correct.

Yeah, also you got mine corrector something. I don't know, Tom. All right, I'll sorry. Yeah, no, that's correct. Yeah. Also, you got mine correct or something. I don't know. Tom. All right. I'll go. Eli. Chad G.P.T. is nothing to worry about because it's not really all that good. But then again, it was released on November 30th of 2022. So, it's been six whole months and it's probably safe to assume it will be the one technology that doesn't get longer rhythmically more powerful over time. B, doesn't have to fool experts since no one listens to them anyway. See the pandemic and QAnon. See, necessarily relies on every government in the world simultaneously building robust social safety nets of UBI and entitlements to protect the citizens from rampant job loss.

Or D, we'll probably all be just fine. It's um, Steve's that last one. Yeah, it is an activism guy. It's a sheer correct answer to this one. Big smile. Here we go. Big smile. A rhythm.

Let's have a remote. Just like rhythm. Let's have a joke. Let's have a joke. Eli, we found out that chat GPT is a transformer. So when it morse into its final form and slaves all humanity, what will it want us to call it? A, Miraja Bage a mega Tron That's good. I scream or D Amazon Prime See star scream Zarscream Zarscream. Oh, you are incorrect. I'm sorry. It's Amazon Prime. I would have also accepted Optimus Prime Minister,

but either one of them. Oh, that's good. All right, well as soon as I heard D Amazon Prime, I knew you were the real winner tonight, Cecil. Awesome, let's get a, let's see if we can get Chatchee PT to write Tom's essay next time. What do you say? Never, never.

It's your game. I already did one on Luddites. All right, well for Eli Cecil Tom and Ethan, I know I'm thanking you for hanging out with us today, we'll be back next week, and by then, Tava will be an expert on something else. Between now and then, you can hear various combinations of us on Kivington of Destiny and D.M. and D.M. and D.M.

and D.M. and D.M. and D.M. and D.M. and D.M. and D.M. and D.M. and D.M. and D.M.

and D.M. and D.M. and D.M. and D.M. and D.M. and D.M. Patreon.com slash citation pod, or leave us a five star review everywhere you can. And if you'd like to get in touch with us, check out our past episodes, connect with social media, or check the show notes, be sure to check out citation pod dot com. And then based on those tokens, I just feed it into the backend API of GPT-3 and boom, I use the cloned voice of Noah, eternal diatreps. Amazing, man. Yeah, right. Hey, guys, what you doing?

Nothing. The shows will be over when you die. I mean, nothing.