Episode Transcript
Don (00:11)
Alright, episode 137, we're back, Back. I think the name of the podcast should just be, we're back. Since I guess I say that every single episode. Back, episode 137. So what are we gonna call this one? We're gonna call it, I was gonna call it Robots Fighting Robots.
Rob (00:26)
We're back. We're back. Robots fighting robots, but
Don (00:37)
But there's also like, there's a side note here to like, is technology ruining our lives? So yeah, mean, you know, we don't know what to call it, but we're going to start talking about ⁓ the topic that we can't stop talking about because it's permeating every single orifice of our lives. Well, I went deep. I went deep on that one. Parotting. ⁓
Rob (00:55)
That's right. I was gonna say aspect you went with orifice
Don (01:06)
Take the innuendo as you will. AI, man, we can't get away from it, AI. Yeah, and I- Why don't you start going?
Rob (01:12)
Here's what gonna, here's what gonna, my setup is this. At the risk of sounding like two old curmudgeons and damn technology ruining our lives. Maybe we are kind of becoming those guys in a way, but AI, ⁓ we've talked about it before and we're gonna talk about it again. But here's a real world example ⁓ of good, bad or indifferent, and I'm gonna say bad in this case. ⁓
Don (01:21)
Were those guys not Muppets? Can't stop, won't stop. When is it good?
Rob (01:40)
When is it good? There's instances where it's good. I need a landscape. I need to generate a landscape for an image. That's true. I need help with...
Don (01:51)
Yeah, I say of all the AIs, what Rob's referencing is ⁓ we need to do some alternative crops on some flattened imagery and potentially extending background images. Photoshop AI has actually gotten pretty damn good at that. Like just fill this area with clouds and trees and mountains and whatever, things like that. So yeah, from an image-based perspective, sometimes AI is actually...
Rob (02:13)
And it is, I will say it's a wonderful resource. ⁓ Hey, I have a question about this. It's a better Google. Like it just is a better Google, ⁓ in my opinion. ⁓ But here was the email I received this morning. I'm going to read this. ⁓ Dear Mr. Broadfoot, I hope this message finds you in good spirits.
Don (02:37)
Off to a good start.
Rob (02:38)
Yesterday we were engaged in drafting a creative writing essay as part of our assignment which serves as a substitute for presenting their personal projects to the class Regrettably the essay submitted was partially generated by AI I have confirmed this through my observations the documents version history and AI detection tools
Don (02:54)
Okay.
Rob (03:04)
As this is the first occurrence of such an issue, your daughter is permitted to resubmit the essay for partial credit, so on and so forth. teacher writes me an email saying, ⁓ daughter got busted using AI.
Don (03:15)
Your daughter used a- And we used AI to find the AI.
Rob (03:20)
Yeah, and we have a and everybody was put on notice school wide and parents why that hey, we're using these are the rules surrounding chat GPT and other AI tools and we are going to use software to help combat the rise of the use of AI. So attached to so I naturally was like, ⁓ this is not does not sound good. Yeah. Well,
Don (03:34)
Yeah, absolutely.
Rob (03:43)
Concurrently, my daughter got the email as well and immediately starts texting me. my gosh, I can't, I am so upset and mad right now. I swear. I think the quote was I swear on everything I love. did not use AI for this.
Don (03:56)
all that is only vehement denial. This is not true.
Rob (03:59)
Vehement denial. This is the pickle I find myself in. and the teacher ⁓ submitted a PDF report of the ⁓ essay question with the report from the zero GPT, which is the AI tool they use to detect AI. Yeah, hence robots. Robots fighting robots. Yeah. So.
Don (04:11)
Okay. Okay, zero GPT. Yeah, hence robots fighting.
Rob (04:28)
And looking at the essay and I hope this story doesn't bore people but it's a time of mine for me right now So so I look at the essay the essay is not long. It's it's it's I don't know. It's probably half a page long tight. Okay ⁓ There are three sentences that are identified on this piece of paper that I'm looking at but that says I'm this verbatim. This is from the zero. Okay, the official or Bayon the official report is
Don (04:33)
We'll bring it around. G-B-T-E-S-S-S-S-
Rob (04:57)
highlighted text is suspected to be most likely generated by a
Don (05:02)
suspected to be most likely.
Rob (05:04)
suspected to be most likely generated by...
Don (05:07)
Three sentences out of an essay have been highlighted.
Rob (05:10)
It's important to note that I fancy myself a writer. Yeah. Okay. So I read said essay and the highlighted text, ⁓ the text in question, the three sentences, in my opinion, some might say professional opinion. Sure. don't know. Thank you. In no way, or form does it stand out as being out of that tonally off. Okay.
Don (05:16)
Sure, this makes sense. Call your professor.
Rob (05:36)
or from a contextual standpoint, it reads, it flows. It reads just like the rest of the thing.
Don (05:45)
Is there any outlandish words in those three sentences? there a weird? ⁓ There's nothing that stands out as to why these would be ⁓ flat.
Rob (05:53)
Nope, are no words that I can say are out of her vocabulary or...
Don (06:01)
It sounds like the rest of the essay. It doesn't feel like a weird oddball shoved in. That's right. Okay, got it. So where do we go from here? So what do do?
Rob (06:03)
It like a- So what do do? So that's where I'm like, what do do? I've got my daughter claiming like, I absolutely did not do this. I believe my daughter. ⁓ And then I have this teacher accusing her based on an AI tool used to detect AI. So, you know, my immediate response was to email the teacher to say, hey, ⁓ I need to understand in greater detail. ⁓
Don (06:22)
based on a Yeah, it's crazy.
Rob (06:38)
how you came to this conclusion outside of just, if this is your only evidence, I can't accept this. It would never hold up in a court of law. Highlighted text is suspected to be most likely generated. And furthermore, when you look at these AI detection tools, it's important to note ⁓ short text, doesn't do well with short text because...
Don (06:51)
expected to be most likely.
Rob (07:03)
part of what it's doing is analyzing patterns, writing patterns, so the longer the text, the more accurate the tool actually becomes. We're talking about a half a page here. We're not talking about a smaller sample size.
Don (07:10)
Interesting. smaller sample size. And what's the batting average of these AI detection tools that that are AI themselves whose only mission by the way is to try to find a reason to exist It's kind of like asking someone for their opinion on something while they are now forced to solicit feedback whether they really have feedback or not or qualified to provide such feedback, right? ⁓ So it's like hey AI like their only mission is to find
Rob (07:29)
Yeah, right.
Don (07:46)
to justify their existence. Yeah, exactly.
Rob (07:48)
So, you know, back to the original email, ⁓ it says...
Don (07:53)
But I what's the batting average of these AI detection? You think it's good, mean, that's it.
Rob (07:56)
I don't know, anything. Well, so in a fun little twist, I of course immediately went to chat GPT to ask about zero GPT, right? I mean, you got it like, yeah. And it basically was like, said it's never a hundred percent accurate. And that the only way to truly, you know, and effectively, uh, dissuade, um, you have to have more evidence. You can't just lean a hundred percent on the.
Don (08:05)
⁓ zero GPT Fantastic.
Rob (08:25)
AI tool. there has to be other methods of detecting this, right? So that was my point was she the teacher in question says I have confirmed this through my observation, the documents version history and the AI detection. Sounds like an AI generated.
Don (08:27)
Focus group of one. Sounds like a copy and paste of the that goes to every, yeah, exactly. Are you going into the metadata of every student submitted, you know, Google Classroom, whatever, to check versioning? So boring.
Rob (08:56)
Come on. Barring any additional evidence, I can't accept this ruling.
Don (09:01)
Yeah, yeah, I don't disagree with you.
Rob (09:04)
So now I've got a, I potentially have a little bit of a fight on my hands. So it just kind of goes back to, we can't completely make decisions based on AI. It's not there yet. And you have to have other information outside of that.
Don (09:23)
It's like, ⁓ remember when, I don't think it was Garmin, but I want to say maybe it was Apple Maps or something. We had Garmin and TomTom for our GPS, right? And we had the little extra device. Nothing was built into your car. And then all of a sudden, Apple Maps launched and what a disaster. Apple Maps is, remember how much, and it was, my God, people are driving into lakes.
Rob (09:37)
What's on your down?
Don (09:49)
You know what mean? People are driving off of roads into whatever because they're not paying. It was one of those like, are they? I mean, I guess they are. But at some point, you have to take control of the car that you are driving. You have to go. And you have to look out. Yeah, you have to look out the windshield and go, I know what's telling me to turn left. That doesn't seem right. You know what mean? Or maybe stop the car. So hey, you know.
Rob (10:03)
right
Don (10:16)
It's a brave new frontier for teachers as well. And hey, give students an inch and they're going to take that inch. I get it, man. I get it. We all get it, right? But to your point, none of it's foolproof. So what do we do? Where are you going with this? Or is this the next part, episode 138?
Rob (10:26)
It alone is not like you. I mean, yeah, I guess we'll I guess the next episode will come back and but where my you know where it stands now is you know, parenting 101 is a I'm gonna I'm gonna back my daughter. Yeah, of course, of course be I'm going to make her rework. She's got to rework it regardless of what happened or didn't happen. She's got to put the work in. Yeah, resolve the thing but I'm absolutely going to hold the teacher accountable and I need an explanation outside of this.
Don (10:51)
He's got a reward. Yeah, maybe kind of sort of would.
Rob (11:04)
Maybe kind of sort of would be maybe ⁓ most likely scenario. Yeah. In order to go around accusing students of plagiarism. Yeah. You just need more than that. And did she do it? I don't know. Yeah. I wasn't there. I don't know.
Don (11:18)
Well, seeing the essay from across the table here, it doesn't even look like it's worth using AI to do like, like, it wasn't like we submitted. Yeah, it wasn't like we submitted this, you know, ⁓ 300 page manuscript as a draft, you know what I mean? And like, you know, like it's, it would take more work to use AI than to just write it.
Rob (11:39)
I know, so call it an essay. I know. It's a few paragraphs, but...
Don (11:44)
Well, utilizing AI is more work. guess I can pivot really quick. Speaking of AI to real world in terms of advertising and clients and whatnot, is we now have a few clients where they have a level of expectation that AI is just going to solve all the magical problems of the world from image creation perspective, right? Yeah. So like, hey, Photoshop, yeah, extend this cloud, add a little background here and there. Like, yeah, it is definitely helpful. I mean, could we do that without AI? Sure.
But now it's gotten to the point where it's like client requests sometimes come in asking for, just AI this. Like, hey, I want three people in a bar cheering, celebrating, held high, doing this, that, and whatever, and all these things like that.
Rob (12:31)
These different ethnicities, I need all of this stuff.
Don (12:34)
And it's like we know that image creation doesn't do humans or animals well. It just doesn't happen. You know what mean? That type of thing, right? So it's one of those like, man, we've been working on this thing for three weeks now, just beating this image to death. When like, dude, we could have just done a photo shoot for this and made like, we've over engineered how difficult things are when they should be easy. It's like, all right, hey. We had the time. You have the budget. Just do a shoot for this. Get exactly what you want. Instead of cobbling together. And own it. I'm not even joking. Like 80 different exports to massage in together to get something that's like, You know, it's kind of passable that no one's super happy. I mean, at the end of the day, I guess they're happy with it. it's like, ugh, it shouldn't be this difficult. We've now, in this instance, or this.
Rob (13:19)
It's okay. Yeah.
Don (13:31)
tipping point of trying to simplify our lives, we've just over engineered things and made it far more difficult. ⁓ My real world example of that was my birthday dinner. I told you about going out for brunch and that whole thing or whatever. It was QR codes for parking, QR codes to get in the elevator. Gone are the days where a server will take your order and just write it down on a notepad, which takes, I don't know. Five seconds. OK, you want the veal, you want the this, whatever. I don't even know why I thought about veal. You want the eggs, you want the things. I rip the ticket, I hand it to someone who does it. No, no, no, no, no. I got to spend five minutes typing everything into this little computer while I'm awkwardly standing at your table. It's like, I get it. It's the digital existence, the digital kitchen. I input it here, it goes there, the tracking. you know, the back kitchen, the way the bills work, the payment. I mean, I understand. But it's like we've taken things that should be simple and should be easy to deal with and now made it five times harder in the auspice of trying to make it simplified. Makes no sense. It's crazy.
Rob (14:47)
Go try and park somewhere in downtown Decatur and then try and figure out which app you have to download so that you can then enter in your license plate number so you can put in your zone so that you can then enter in your credit card. mean, I just want to run it in the store for like 10 minutes and ⁓ grab whatever.
Don (15:06)
Yeah, just give me a ticket and I'll give the guy five bucks or whatever. Yeah, it's strange. There are a lot of great things from an image perspective. mean, I don't know what's going on with the video side of life yet. I know it's getting better. I guess the new thing now is your voice, speaking of our podcast and our whatever, is now nothing can be voice authenticated anymore because all the robots can copy your voice.
Rob (15:08)
This is crazy. is now the time when we tell the people that we're not actually.
Don (15:38)
that we're robots.
Rob (15:39)
We're not actually giving this podcast?
Don (15:41)
But that's going to be a significant problem. Forget about all the fake political ads or this or that. Anytime anybody, anything audio, it's all fake, That's crazy, man.
Rob (15:55)
For all our listeners out there and all our readers out there. There's a book called the coming wave That I highly recommend if you want to be terrified about it It is ⁓ it is It's very academic, ⁓ but it's a really good it's ⁓ I'm not I'm not gonna go too deep into the weeds But if you if you look back to Ted Kaczynski the bomber, yeah and to his manifesto
Don (16:04)
of the future if you don't want to sleep at night.
Rob (16:23)
outside of the horrible, awful, awful things he did. And he was a horrible human being. That manifesto was pretty accurate, almost frighteningly accurate, in of the machines taking over and what was going to happen. So the coming wave subscribes to a lot of this. I mean, it doesn't align itself in any way, shape or form with the manifesto or anything like that. But if you've read both, there are points that are very, very similar. And it's almost an academic take on
Don (16:42)
Thank you.
Rob (16:50)
just the future and technology and how we're at that point where we've kind of gone too far. And you can't, once the robots are learning from each other, ⁓ bad things can happen.
Don (17:04)
I mean, is learning the right word because it's not a fact-based intelligence?
Rob (17:11)
It matter. can learn things that aren't fact, but like.
Don (17:13)
Yeah, I mean, well, I mean, yes, they are interacting with each other and it is machine.
Rob (17:17)
Can use the word manipulating. could use, you know, I mean, there's a lot of words you can use, it's, yeah, it's scary stuff, man. ⁓ So that's your Uplifting Podcast for the day. ⁓
Don (17:28)
In the meantime, we'll use the robots to make some beautiful advertising and graphic design. ⁓ But yeah, the power of persuasion, right? The fun sort of tagline, unofficial tagline of advertising in general. It's like, well, now robots can synthesize and fake everything in terms of the power of persuasion, right? Any celebrity can endorse anything or anybody can contradict anything. It's really, really weird to think of. Yeah. And now we're using
Rob (17:53)
Yep. It's very strange.
Don (17:58)
AI authenticators to double check for AI, but their batting average is not good. So that's crazy.
Rob (18:05)
Is So anyway, on next week's episode, ⁓ we will follow up and I'll let you know how the case unfolds.
Don (18:19)
I'm curious. I'm very curious. ⁓ Yeah, unfortunately, I think she probably, just in the time being, probably has to resubmit a little bit. You know what I mean? Just to show the effort. Just to show the effort. That's right. But not admit that ⁓ there was deceitfulness in there. So craziness. All right, where can the robots find us, Mr.
Rob (18:41)
Well, they already know okay, they can find us yeah, but for those of you are For those of you who aren't robots so can find us online We're not easy or hard to find rather mock the agency comm and of course, I'm to socials. I'm hard to handle I'm hard to handle
Don (18:45)
BLEEP BLEEP BLEEP hiding in a bunker. What I'm doing. ⁓ Too hot to handle. All right. Thanks, everybody. See you next
Rob (19:02)
Exactly.
