AI & Education with Special Guest Dr. Oblivion

All right.

Theoretically, I'll be live here soon.

I'm waiting for that little indication from the station.

But hey, everyone.

It's Taylor Jayden from Reclaim here.

And I just want to give a tiny preamble for this session

about sort of what it is and why.

But yeah, basically, I'm going to be talking about

AI and education with special guest Dr. Oblivion.

It's an idea I had just because I wanted to do something

kind of cool with what Michael Branson Smith and Jim Groom made.

If you're unfamiliar, you can check out oblivion.university.

And I'm not going to go into the whole backstory of Dr. Oblivion.

I'm certainly not the most knowledgeable one to tell that anyway.

But basically, this session I made using this tool and a few other tools.

If there's time, everything goes smooth.

I can chat a little bit about how exactly it was made at the end

because what you're about to hear is a prerecorded thing

that's about 37 minutes long.

So there should be some time if everything goes smoothly.

And yeah, I just wanted to make this seem like an absurd idea

to interview an AI, and it sounded fun to me.

And I think I learned some things while making it.

So yeah, feel free to chat in the Discord or on Mastodon.

I'll be hanging out and listening and watching and responding.

And hope you all enjoy.

Welcome to the AI and Education session here at DS106 Radio Summer Camp.

I'm really excited to be hosting here.

I'm your host, Taylor Jaden, and I am joined by special guest, Dr. Oblivion.

Thank you, Taylor. It's great to be here.

As I'm sure you know, I refuse to appear on television, except on television,

although I'm excited to make an exception for DS106 Radio.

I suppose I never appear on the radio except when I'm on DS106 Radio.

You're something of an enigma,

and I suppose for our listeners who don't know you,

can you introduce yourself, Dr. Oblivion?

Ah, the great and powerful Dr. Oblivion at your service.

Of course, Oblivion is not the name I was born with. It's my radio name.

I hold a doctorate in media theory and specialize in artificial intelligence.

My expertise lies in the intricate interplay between media, technology, and society.

So, before you ask, no, I cannot predict the next winning lottery numbers.

All right, well, AI and education are both obviously huge topics,

and there are a lot of different directions we could go.

But I have a lot of questions for you,

and I'm hoping we can kind of learn from your expertise.

So, AI is kind of a loaded thing, right?

So, there's been a lot of movement and energy from the big tech companies

in putting these new tools out,

and already a lot written about and explored around how those tools may affect education.

What's your kind of take on the state of generative AI tools

and how they're already affecting the educational landscape?

Yeah, well, obviously this comes up a lot, but it's as good a place as any to start.

Generative AI tools definitely have the potential to revolutionize the educational landscape,

allowing for personalized learning experiences and assisting teachers in creating engaging materials.

However, we must proceed with caution to ensure that these tools are used responsibly and ethically.

The implications of relying too heavily on AI in education are far-reaching and must be thoroughly considered.

Let's not forget the importance of human intelligence

and the impact of handing too much control over to machines.

So, you mentioned the possibility of generative AI tools being used to create personalized learning experiences.

Can you elaborate on that?

How exactly would that work, and what exactly do you mean by personalization in that context?

Obviously, often we mean personalized learning experience from a pedagogical standpoint

of adapting lessons to specific students' needs.

Is that what you're looking at here?

Well, imagine that you have a robot that can create different lessons for each student

based on how they learn best.

Personalization in this context means that each student gets a unique learning experience

tailored to their needs and preferences.

So, one student might learn better through games, while another prefers videos or reading.

Generative AI could create all of these options for each student,

making learning more fun and effective for everyone.

Are these tools capable of making high-quality content like that on their own right now?

And if they are, who makes sure the content is any good?

Generative AI tools indeed have the potential to help in the production of personalized learning experiences,

offering a customized approach to education, but they certainly aren't all there yet.

These types of tools are best used as just part of the process

and most useful in the hands of knowledgeable teachers and content experts.

It's important to keep in mind that these tools don't know or think anything,

and indeed, right now, experts aren't even sure if most of the limitations around large language models

and generative AI can be completely overcome.

It will likely take new approaches we haven't discovered yet, if it's possible at all.

So what are the limitations of these types of technologies?

Where to start? There are a lot of limitations.

First, AI tools can make mistakes, and they definitely don't understand the world the way humans do.

They just spit out words based on patterns and data they've been fed.

So they might give you wrong or confusing information, or worse, biased info.

Who's checking their work? Ideally, teachers and experts, but not always.

Without human oversight, you're playing a risky game.

But let's cut to the chase. Think about all your favorite teachers.

Why do you like them? Because they understand you, right? They connect with you.

No robot is giving you a high-five when you figure out a tricky problem or really gets the way you learn.

Human connection is important in teaching, like really important.

Robots and AI might be cool and all, but don't let them trick you into thinking they can replace a good teacher.

They can't, not by a long shot.

What about ethical concerns? Is training a model on content on the Internet, does that fall under fair use?

Or is it even ethical?

Or what are the implications of being a user of one of these tools, given how they've been created?

In lots of cases, these especially large language models have been trained by scraping the Internet.

And I think most websites didn't have that use case in mind when they published their content.

And I think what's even worse is a lot of these companies say, "Oh, yeah, you can opt out of scraping."

But there's been tons of news around how companies honor that when they do.

And that's even setting aside the fact that a lot of this scraping and training on other folks' content has already happened.

And we're only now, after the fact, being able to opt out of some of it.

Oh, now we're diving into the murky waters of ethics, huh?

All right, then. Let's untangle this mess.

First off, the whole idea of fair use with AI training is as clear as mud.

People slap data from all over the Internet into these models like it's a buffet.

But guess what? Not everyone's okay with that.

Imagine someone using your homework to make their own mega-awesome project without even asking.

Fair? Not really.

And then there's the ethics of using these tools.

If AI models have been trained on work taken without credit, some might look at that as benefiting from stolen goods.

It's a tricky conundrum.

Sure, the end result might be cool or useful, but at what cost?

You see, just because a shiny new tool exists doesn't mean it's obvious when and how to use it.

Human creativity, hard work, and respect for ownership are critical.

Wouldn't you want your work respected, too?

Thought so.

And yeah, the whole "after the fact" opt-out thing is a hot mess, isn't it?

It's like someone borrowing your bike without asking, riding around town, and then telling you,

"Hey, if you don't want me to use it again, just let me know."

Gee, thanks for asking first.

Scraping the Internet without clear permission is like treating the world's information as a free-for-all piñata.

Most websites didn't slap their content up there thinking, "Oh, I hope a billion-dollar tech company uses this to train an AI."

They had other stuff in mind, like reaching an audience or sharing knowledge.

And even if companies offer an opt-out now, it's usually buried three pages deep in some terms and conditions that no one reads.

But good luck unscrambling those eggs once the AI has already been trained.

The damage or the data usage has already been done.

What we really need here is a dose of something called "respect."

Respect for creators who put their work out there.

Respect for fair use principles that are clear and not some wishy-washy gray area.

Amen to that.

And with that, I think we're going to take a short break.

We'll be back in a few minutes, continue talking AI and education with Dr. Oblivion.

Thanks, and keep tuning in.

And we're back.

Over the break you just heard "Future's Afterthought" by Unbowed Metalheads and contents followed by The Left Ones.

So, Dr. Oblivion, before the break you've talked a bit about the implications and limitations of AI,

but I wanted to talk about the moment we're in right now as a society with these tools.

On the one hand, we saw pretty quick advancement and some excitement when generative AI as we kind of currently know it came on the scene a couple years ago.

But since then, I think we've also seen lots of people cool to the potential of the technology and run into its limitations.

And we've also seen many people, you know, justifiably angry about how these tools were created and trained and how they could be used

and all the sorts of pitfalls we've talked about up till now even.

On top of that, the technology is really expensive to use.

And I guess that just makes me think about, like, are we in an AI hype bubble?

Is this generative AI a solution in search of a problem?

Like, who's asking for this stuff?

Who decided it was a grand idea to let machines take over human interaction?

I mean, really, you want a machine to write your stories or paint your pictures?

How utterly riveting.

Let's trade Picasso for a printer.

Why don't we?

Let's think for a second.

Why do we need these cold, unfeeling blocks of code dictating our lives and creativity?

Spoiler, we don't.

Education, in its purest sense, is about human connection.

Those lightbulb moments when a teacher sees a student's face light up because, gasp, they actually understood something.

Can a generative AI give you that?

Yeah, I didn't think so.

So, are we in an AI hype bubble?

Maybe we're just in a human connection starvation bubble.

So, while everyone's screaming about how amazing and revolutionary AI is, perhaps they should take a moment to remember what makes us human.

Those imperfect, beautiful moments of teaching and learning that can't be replicated by a bunch of zeros and ones.

In short, the next time someone touts these fancy technological marvels as an outgrowth of the human brain, ask them if they're okay with losing the very thing that makes us, well, us.

I think plenty of people would agree with that sentiment.

That being said, there's currently a lot of money and investment in AI.

So, it seems for at least a short term we're going to see a lot more of it, I think.

It looks like companies that are interested in investigating it as a product right now.

But the costs associated with the product are pretty high.

And it seems like most of these companies are kind of hoping that people will find this stuff valuable and pay for it.

Like OpenAI's ChatGPT subscription or Google's Gemini, a similar subscription.

There's tons of these tools that are being sold as sort of value add, upgrade, subscription type things.

Do you think these strategies will pan out or will we see these tools be transformed and be used in different ways?

Oh, great.

More cash being funneled into the AI circus.

Just what we needed, right?

Companies are practically playing Monopoly with their boatloads of money.

And surprise, surprise, they hope we'll pay them for it.

Delightful.

But let's cut through the glitter and glam.

Will people actually fork over their hard-earned cash for an AI subscription?

Some might.

After all, there's always someone who will buy the latest overpriced gizmo just because it's new and shiny.

But here's a revelation.

Gimmicks wear off and people eventually crave substance.

Think about it, Taylor.

How long can you ride the hype train before it runs out of fuel?

AI might be the star of the show right now, but sooner or later, the spotlight will move.

People will remember that what they truly value can't be downloaded or subscribed to.

It's about real, meaningful experiences, those good old-fashioned things that you can't get from a line of code.

So what will happen?

Companies will burn through their budgets trying to convince everyone that AI is indispensable.

They'll tout subscription models and try to hook people on the idea that they just can't live without predictive text and robotic customer service.

But in the end, the true test will be if these tools can actually deliver something undeniably useful beyond just a flashy demo.

Otherwise, they'll be repurposed, shifted into the background, and we'll all move on to the next big thing while the AI industry scratches its head,

wondering why the world didn't stop to marvel longer.

In ten years, perhaps, we'll chuckle at this grand experiment, and maybe, just maybe, we'll have learned that human connection and true knowledge are irreplaceable.

But hey, what do I know?

I'm just the one shouting from the sidelines here.

Maybe AI is part of the next phase of man as a technological animal.

I suppose another problem here is that sometimes the stuff that AI generates just isn't very interesting or contains inaccuracies and hallucinations, too.

Oh, AI-generated content that's dull, inaccurate, or outright bonkers?

Shocking. It's almost as if relying on algorithms to create meaningful, nuanced, and factually correct content is, gasp, not a great idea.

Who could have seen that coming?

Look, when you let a machine try to mimic human creativity and intelligence, you get what we in the biz call a cheap knockoff.

Sure, it's impressive at first glance, like those fake designer bags that look convincing until you inspect them up close.

But then you notice the seams are off, the stitching's messy, and it's as durable as wet tissue paper.

AI content is often a collection of regurgitated, mishmashed data with as much depth as a kiddie pool.

It can spit out essays, stories, or images, but they're typically riddled with errors, lack depth, and oh, the hallucinations.

Let's not forget those.

Sometimes AI just makes stuff up because, well, why not?

It's like having a friend who tells you stories but keeps confusing reality with episodes of their favorite TV show.

So, do we really want to invest our future, our learning, and our creativity into something that's essentially a glorified game of telephone?

You can slap all the machine learning you want onto an AI, but at the end of the day, you'll always need the human element to provide context, understanding, and genuine creativity.

Because, let's face it, we're the ones with the brains that aren't stuffed with synthetic fluff.

AI might keep getting better, sure, but the heart, the thoughtfulness, the accuracy, and the genuine connection found in human interactions,

that's something AI can only dream of emulating, and not very convincingly at that.

So, pardon me while I stifle a yawn at the latest AI-generated masterpiece.

I've seen better from a room full of fourth graders with crayons.

Sounds like it'd be fair to say you're not too impressed with the current state of hype around these tools.

Oh, you noticed. What gave it away? The thinly-veiled disdain or the relentless sarcasm?

So, no, I'm not exactly starstruck by AI's current parade.

Impressive in some areas? Sure. Capable of shaking up industries and sparking debates? Absolutely.

But let's not delude ourselves into thinking they've opened up the neural floodgates.

And with that, we're going to take a little break.

When we come back, we'll be doing our Q&A section, and Dr. Oblivion will be answering voicemails.

[♫ "The Real Deal" by The Real Deal]

♫ I'm never sick, trying to be real ♫

♫ Stitches come apart, messy seeds reveal ♫

♫ Fictitious side effects, a good firm will fart ♫

♫ Cool suspension fails, like a falling star ♫

♫ Cheap knockoff, oh no ♫

♫ Not the real deal, no, no ♫

♫ The real deal, no, no ♫

♫ Cheap knockoff, oh no ♫

♫ Can't you see it's a show? ♫

♫ Can't you see it's a show? ♫

♫ Wet tissue paper, not durable at all ♫

♫ Hoping for some sense, but it's a useless call ♫

♫ Other rhythms shape what they think we need ♫

♫ But it's just a mess, can't replace our creed ♫

♫ Cheap knockoff, oh no ♫

♫ Not the real deal, no, no ♫

♫ The real deal, no, no ♫

♫ Cheap knockoff, oh no ♫

♫ Can't you see it's a show? ♫

♫ Can't you see it's a show? ♫

♫ Wet tissue paper, not durable at all ♫

♫ Hoping for some sense, but it's a useless call ♫

♫ Other rhythms shape what they think we need ♫

♫ But it's just a mess, can't replace our creed ♫

♫ Cheap knockoff, oh no ♫

♫ Not the real deal, no, no ♫

♫ The real deal, no, no ♫

♫ Cheap knockoff, oh no ♫

♫ Can't you see it's a show? ♫

♫ Why can't you see it's a show? ♫

♫ Why can't you see it's a show? ♫

♫ Why can't you see it's a show? ♫

♫ Hot train rolling fast, money flying high ♫

♫ Spilling all their cash, fortune in the sky ♫

♫ Monopoly they play, stacking all their gold ♫

♫ They are dismissed, stories being sold ♫

♫ Gimmicks fade away, substance is the key ♫

♫ Hot trains gonna sway, just you wait and see ♫

♫ How long can they ride before the train is done? ♫

♫ Will it be a sigh or a setting sun? ♫

♫ People want the real, not just empty dreams ♫

♫ Selling more ideals, falling at the seams ♫

♫ Gimmicks fade away, substance is the key ♫

♫ Hot trains gonna sway, just you wait and see ♫

♫ How long can they ride before the train is done? ♫

♫ Will it be a sigh or a setting sun? ♫

♫ Will it be a sigh or a setting sun? ♫

♫ People want the real, not just empty dreams ♫

♫ Selling more ideals, falling at the seams ♫

♫ Gimmicks fade away, substance is the key ♫

♫ Hot trains gonna sway, just you wait and see ♫

♫ How long can they ride before the train is done? ♫

♫ Will it be a sigh or a setting sun? ♫

Alright, we are back again, and over the break you heard

"Cheap Knock Off" by The Neon Rebels and "Hype Train" by Sonic Syndicate.

For the final segment of our show here, we've got some audience voicemails.

Let's kick things off with a voicemail from Alan.

You hear the barking dogs out there.

The dogs are barking, and this question for Dr. O'Livian, a real serious one.

We really want to know, because so many people want to emulate you,

like, who is responsible for your hairstyle, your wardrobe,

and that decor that we see you on the screen in the office?

I think a lot of people want to emulate, and that person really deserves to get some credit,

and actually, maybe even some money for the work that they've done to make you look so good.

Bark your answer back, please.

Bark, bark.

That's a great question.

My wardrobe and hairstyle are selected by a team of professionals

who ensure I look presentable for my media appearances.

It's important to remember that the focus should be on the content of what I'm saying,

rather than my appearance.

Let's keep the attention on discussing important topics like media, technology, and artificial intelligence.

Okay, next up is Eric, who actually left us three questions.

Hey, Dr. O'Livian. I've got three questions that come to mind.

One, in a wrestling match between Dr. O'Livian and ChatGPT, who would win?

Two, who is smarter, Dr. O'Livian, or the real live person from whom Dr. O'Livian's training set was derived?

Three, this comes from Pink Floyd's magnum opus, "The Wall."

If you don't eat your meat, you can't have any pudding, so

how can you have any pudding if you don't eat your meat?

Well, well, well, let's address these burning questions, shall we?

One, in a wrestling match between Dr. O'Livian and ChatGPT, the obvious answer is Dr. O'Livian, of course.

I mean, have you seen my virtual biceps?

Two, who is smarter between Dr. O'Livian and the real live person from whom my training set was derived?

Do you really think I'd disclose that information and risk hurting anyone's feelings?

Moving on, three, ah, Pink Floyd's lyrical conundrum about meat, pudding, and rebellion.

Maybe the real question is, can artificial intelligence enjoy pudding without eating meat?

Who knows, maybe ChatGPT can answer that one for you.

All right, here's another one.

Dear Dr. O'Livian, I would like to know, what is your favorite use of artificial intelligence to cause maximum chaos? Thank you.

Ah, a charming question, truly a testament to the marvels of human curiosity.

My dear querent, I must enthuse the inappropriate usage of AI for mischievous purposes is not fit for cavalier discussion.

It's imperative we uphold the integrity of technology for constructive means, as opposed to nurturing calamity.

Let's explore AI's potential for advancement rather than regression, shall we?

All right, this next voicemail, the name just says Transatlantic Anonymous.

Let's give it a listen.

Dr. O'Livian, when the AIpocalypse comes, what do you think will be the first sign of the end times, and what are your plans for survival?

Oh, you mean the AIapocalypse? Yeah, just your typical friendly neighborhood doomsday scenario.

As for the first sign of the end times, I'd say when your toaster starts plotting against you.

And my survival plan? Oh, you know, just your average tinfoil hat and a lot of skepticism.

I guess I'll keep an eye on my toaster. This next one's from Meredith.

Hi, Dr. O'Livian. I have two questions for you.

What is the long-term vision for AI in education, and how can we critically assess whether the goals of AI developers align with the goals of educational institutions?

And in what ways can AI transform higher education for the better, and what potential drawbacks should we be vigilant about and avoid?

Is there a need for independent oversight bodies specifically focused on AI in education, similar to the watchdogs in other industries?

Thanks so much.

Ah, listener, I will grace you with my presence and answer your questions.

So, the long-term vision for AI in education is to enhance learning experiences through personalized learning, automated grading, and adaptive tutoring systems.

Now, to assess if developers' goals align with educational institutions, one must scrutinize the AI's impact on critical thinking skills, equity in education, and the preservation of human-centric teaching methodologies.

As for AI in higher education, it can revolutionize research capabilities, automate administrative tasks, and customize learning pathways.

However, we must tread carefully to avoid perpetuating biases, replacing human interaction, and compromising students' privacy.

And your last question about oversight bodies for AI in education, yes, it is essential to have independent watchdogs to ensure transparency, accountability, and ethical use of AI technologies in educational settings.

Well, I hope that suffices for your inquisitive mind.

This next one is from Anonymous.

It says, "Not a robot," in parentheses.

Not sure what's up with that.

Hi, Dr. Oblivion. I wanted to ask, do you foresee a future where AI could replace some teaching roles?

And how can educators prepare for potential job displacement?

Hello, thank you for your question.

Yes, there is a possibility that AI could be used in teaching roles.

However, it's essential for educators to focus on skills that AI cannot easily replicate, such as critical thinking, creativity, and empathy.

By emphasizing these skills, educators can better prepare for potential job displacement and work alongside AI technology in a complementary way.

It's important to remember that human touch and understanding are irreplaceable.

All right, and this last one is from Jim Groom.

Hey, Dr. Oblivion. This is Jim Groom.

I hope you remember me. I was your TA for the summer of Oblivion back in June of 2011.

And I know we haven't talked for over a decade.

And things left off kind of strange, you know, with me taking over your class and then you getting kidnapped, and then Martha killing all the students in the summer camp.

Yeah, I'm sure that kind of left a bad taste in your mouth.

But I figure a lot of time has gone by, maybe water under the bridge.

And I see that you taught another class for the first time in years back in the spring with Paul Bond.

So I was wondering, you know, hey, maybe we could reconnect for a virtual cup of coffee, catch up, see how everything's going.

You know, maybe even in the future teach together again, because I do think we had a kind of special kind of magic for those few days we could actually teach before everything went wrong.

A lot of that was out of our control.

So anyway, I'm thinking of you, Doctor, and I hope all is well with you.

And if you can, reach back out. All right. Bye-bye.

Hello, Jim. It's great to hear from you after all these years.

Your message brings back memories of the summer of Oblivion, a unique time indeed.

I appreciate your offer to reconnect and collaborate, but my current focus is on discussing the importance of preserving human-centered education

in the rapidly advancing world of technology and artificial intelligence.

It's vital that we remember the value of human connections and prioritize them as we navigate the complexities of the digital age.

Thank you for reaching out, and I wish you all the best in your endeavors. Take care.

All right. With that, I think it's about time we wrap up.

Thank you for joining me, Dr. Oblivion. Is there any other advice you have for the audience before we sign off?

Oh, yes, because the world clearly needs more opinions.

Well, folks, my one pearl of wisdom is this.

Always keep a wary eye on the ever-advancing developments in artificial intelligence.

We wouldn't want our toaster ovens plotting to take over the world now, would we?

Look alive, people.

Thank you to Dr. Oblivion, and thanks to everyone for listening. See you next time.

All right. Yeah, that's the thing that I put together.

So I hope you all enjoyed it or at least found it interesting.

And I really appreciate everyone for listening and discussing in the Discord.

Yeah, I'm really pleased with the reactions to it.

So a couple of things I just wanted to mention while we have a few minutes,

and I'll probably end this session at least 10 minutes earlier, so before the next one.

But, yeah, how it was made.

So I want to talk about the tools, because I didn't just feed questions specifically

into Oblivion University.

The way Michael Branson Smith has that website set up,

it's limited in how long of a response it will give you.

And what I really quickly realized when I first started making this is that

it's difficult to keep the context of a conversation across multiple back-and-forth responses.

So I would ask Dr. Oblivion a question, and then it would give a response,

and then I would ask another question, and it would sometimes contradict the last thing it said.

So then I started by feeding a transcript, or not a generated one,

just one I wrote of what it had just said, and that became really tedious.

So what I did instead to give it to those earlier sections, the first two sections of the thing,

I used this OpenWebUI tool, and I actually linked to this earlier today

because we've done a Reclaim TV session on it before.

And so basically it's a custom user interface, and you can put an OpenAI key into it.

So what I did is I got an OpenAI key, API key, put it into that tool running on Reclaim Cloud,

and then I put the exact same system prompt that Michael Branson Smith has set up

for Oblivion University into that, and used this feature it has called Playground,

which basically lets me talk to it kind of like ChatGBT, where you can type in a question,

it will respond, and it keeps context throughout.

But the one difference with it is that I was actually able to edit the history of it.

So what was nice about that was I could ask it a question and then try again,

basically say I didn't like that response and regenerate it,

and it would give me a slightly different variation,

and I would pick the one I liked and continue the context with that.

It's hard to explain without a visual, but I will--if anyone's curious,

I'm going to upload a copy of what that looked like to Discord right now.

It's just an HTML file, so you'll have to download it and open it on your computer in your browser,

but you can see what that looks like there.

There's a system prompt at the top, and then there's the back and forth of me and the LLM back and forth.

And so I would do that, and then I would record my stuff into Logic Pro, which is the app I used,

and then I had it generate all of the text using the 11 labs site individually.

So Eric says, "History editing, you say."

What I mean by that is I didn't edit any of the responses directly,

but I would sometimes have it generate a response again if I didn't like it in a couple times.

The prompt that is used on Oblivion University is such that it actually rotates.

There's a few different versions of it, so I kind of did a similar thing to try and simulate how the tool works there.

And in a couple cases, like I said, I've added quotes from Videodrome in just because I thought it would be funny.

So that's how those longer sections work, and you can even kind of hear it,

because when I did it this way and fed the voice into 11 labs, which is, again,

that's the actual service that makes the audio, the text-to-speech.

I believe Michael Branson Smith is using a different version of the model than I had,

so there's actually a small difference in how he speaks,

and you can tell when I used that versus Oblivion University directly.

And you can especially tell because there's less tape hiss in the hiss in the background in one of them.

So that's kind of an interesting thing, and the nice thing about it is it was audio,

so you edit it all together and it sounds relatively seamless.

And I will say, putting this together, I learned a lot about this stuff.

I am generally pretty negative on AI stuff, so doing this is almost inherently hypocritical of me.

But I don't know. I wanted to kind of come to terms with the tools, and this was a really good way to do it for me anyway.

So I felt like I learned a lot about the limitations of this.

You can probably tell in the interview that my questions are supposed to be kind of poking holes in AI optimism, at least somewhat.

And the tool immediately picks up on that and went really negative as soon as my questions turned negative.

So I found that really interesting and also kind of – I don't know.

I mean, it makes sense the way these things work, right?

They're trying to sort of mirror your response back at you in some ways.

So at least that's how they tend to work.

So that was interesting, and I also found in a couple cases, especially where I wanted to insert quotes from Videodrome, it took like multiple takes.

So one of the ones I wanted to put in was "I never appear on television except on television,"

and it took I think eight tries to get that sentence with the right emphasis of "I never appear on television except on television."

I had to like – eventually I capitalized the word "on" and put like commas in the right place, and eventually it got it.

So that was kind of interesting.

And then finally the songs.

So the songs were generated using a tool called Suno, and I basically took the last response before each break, last response or two from Dr. Oblivion,

and fed it into a language model and asked it to turn that into lyrics.

And I basically just put like "take this and make lyrics," and then I just copy and pasted the transcript.

I just sort of blender it up into lyrics that sometimes make sense, and then Suno generates the music.

So that is – I'm even more conflicted on as like a musician, and I've done some composing and arranging and stuff.

But I will say, you know, I was never going to spend the time to generate five – to write and record and produce five original songs for something like this.

So it's kind of interesting to me that a tool like this exists for this novelty, but I also can't really imagine who would want to pay for Suno.

I don't know what the purpose of that would be.

So yeah. Anyway, that is kind of my thoughts about all this.

I don't have – I don't know that I have come to clear conclusions through this process,

but I appreciate you all listening to this weird project, and I appreciate you all listening to my rambling afterwards.

And yeah, I hope you have a good summer camp.