Leo [00:00:22]:
Hey, Gary.

Gary [00:00:23]:
Hey, Leo, how's it going?

Leo [00:00:25]:
It's going well. So my question to you is, Yeah. Prove you're a human. As we continue to do this, and I just see this happening a lot over the course of the coming years, it's going to be harder and harder to determine whether or not the podcast you're listening to or the— whatever you're watching is actually a real human. It's kind of funny because the— you've played with NotebookLM too.

Gary [00:00:53]:
Mhm.

Leo [00:00:54]:
And they generate podcasts, but their podcasts are very stereotypical. It's always 2 people, they're always overly excited, um, and, uh, you can, you can almost always tell. It's funny because there was another podcast that I was listening to that, um, uh, to me I was thinking Oh my gosh, this sounds exactly like a Notebook LM podcast. And I ended up tracing it down and no, these are 2 people who just naturally sound like an AI-generated voice, naturally exuberant. And the proof is that, um, there's a video of them talking and it's clearly not AI.

Gary [00:01:45]:
They're just, they're just eloquent, unlike us. See, we'll be like the last. They'll be like, yeah, Gary and Leo, they're real because they just um and ah through their entire podcast. But, uh, yeah, I, I've actually, um, yeah, it, it's getting to be a thing, you know. But also, also there's another like aspect to it which kind of makes me mad because there are people that are just not really good at talking. There are people that are not good at— can't talk, right? But have something to say. And AI voices is a really good way to go in that direction.

Leo [00:02:22]:
Oh, absolutely.

Gary [00:02:23]:
And I think if, like, I— from time— my voice isn't as good as it used to be. Like, if I have a lot of work to do, like I'm working on a course, or maybe I like try to do like 2 videos and then we talk and then I've got something else going on. My voice is broken by the end of the day.

Leo [00:02:41]:
Sure.

Gary [00:02:42]:
And it doesn't sound right and it's all dry and it's just— yeah. So it didn't always used to be like that. I'm getting older. I could see, hopefully not for me. Hopefully I'll go the rest of my life and I'll still be okay. But I could see it going downhill. And I know several people, several that have actually lost their voices. They have it.

Gary [00:03:03]:
Maybe they could talk for a few minutes and that's it. Like there's actually lots of maladies that could, that could befall your voice. So should you then be not allowed to podcast? Like, nope, you're not allowed to podcast. You— there are AI voices now. You could write. write a beautiful podcast and then have an AI voice speak it for you. Now, does that— are there, you know, should people not listen to what you have to say because of that? Now, there's a difference between an AI voice saying it and an AI writing it, right, and saying it. Or is there? Because here's the thing, I listen to exactly one podcast.

Gary [00:03:44]:
And that podcast is generated by AI and spoken by AI. And before you go and try to get your pitchforks out to protest this podcast, you can't, because this podcast is not available to anybody else in the world except me, because I created it. Because there is no podcast out there that will hold my interest because it's not telling me what I want to know. I— and I think I've talked about this before, maybe, or mentioned it, that I do this.

Leo [00:04:09]:
I don't know that you've talked about it here, but I know you talked about it. Yeah.

Gary [00:04:12]:
So this is— it takes me approximately 8 minutes to walk my dog in the morning, much longer walk in the afternoon when it's, you know, the day is over. But in the morning, it's 8 minutes to go around the block and have my dog do his business in the morning before I can sit down and have my cup of coffee. And I want something to listen to that'll help me later on, like it'll replace something I need to do later on. So I thought, oh, I could have— if there was a podcast that gave me local Denver news, that would be good. But I also want local tech news because I, you know, I keep track of stuff in part for this podcast. I also want world news, a little bit of world news, but only from sources that I like. There's also a bunch of other things I would like to hear in the morning. So I sat down and wrote a Claude prompt to create a podcast, to write a podcast specifically for me.

Gary [00:05:09]:
I give it specific sources. I don't say just go out and tell me what's going on in Denver. I say go to these sources, these RSS feeds, and see what's happening. Favor news that's really close to my location, my part of town, so I don't miss a, like, a minor story that might get overlooked, but it happens to happen a block away from me. I want to know about it, right? Uh, world news, summarize that. Uh, tech news, but you know what, focus on Apple because that's what I'm interested in.

Leo [00:05:37]:
Right.

Gary [00:05:37]:
So don't give me endless AI news or Google or Microsoft or whatever, you know, just, I mean, if Apple's doing something, fine. If there's some other general thing, then I give it other stuff. I say I want sports news, but you know what? I only care about one team. It's my local baseball team. I really don't care about other baseball teams. I don't care about football. There's nobody else who's gonna create a pod— like a 1-minute segment on a podcast that just tells me who hit home runs last night. Who got traded? Keep going.

Gary [00:06:06]:
Give me a humorous quote. But you know what? AI is lousy at that because it'll give you like the same 5 humorous quotes over and over again. So I point it to Goodreads, which has an endless, endless list of humorous quotes from books. I say, pick one at random. Give me a productivity tip. And I had a little list of about 20 different blogs over the years that had little productivity tips. that I can't be bothered to go back and look at. But I give it that list and I say, go pick one of these at random, go pick a random blog post, give me a tip from it.

Gary [00:06:41]:
I'm learning stuff like right now I say, go to Wikipedia's list of logical fallacies, pick one at random, summarize it for me, and repeat the name of the fallacy at least 3 times. Because what I like to do is eventually get good enough where I can call out a fallacy by name. During an argument. I then end the podcast with— I ask it to go to Wikipedia's list of songs, right? There are pages on Wikipedia for songs where it gives you the history of the song and what's interesting about the song. Thousands of them. Go to Wikipedia's list for that. Pick one at random. Summarize it for me and tell me like you're a radio DJ what's cool about this song.

Gary [00:07:24]:
And then end the podcast by naming the song and artist. And then I take this entire script. I have AI that is the Siri voice speakable stuff on my iPhone. Read this podcast to me.

Leo [00:07:41]:
Oh yeah.

Gary [00:07:42]:
And then at the end of this shortcut, the last thing is always the name of the song and the artist. So it automatically takes that last line of the podcast and then plays that in Apple Music. So it really sounds like it's DJing, like it tells me about a cool song. And then at the end, it says so-and-so by so-and-so. And then the song starts. And it's great. And it's all generated by AI. And it's just for me.

Gary [00:08:09]:
Now, how would I do this any other way?

Leo [00:08:12]:
Right.

Gary [00:08:12]:
I mean, this is like— and it took some work over the period of a week to keep honing this. And sometimes Claude rewrote its own prompt. Like I actually had to say, oh, you spent too long on this, or oh, you need to you know, whatever. And it actually rewrote the prompt, and the prompt is now really long, but it does this for me, and it does it automatically every morning. So it's waiting for me. Um, really cool, but, uh, it's different than what most people think of when they think of like, oh, an AI-generated podcast, right?

Leo [00:08:46]:
It's interesting because, I mean, we're, we're kind of going into the philosophical AI rat hole here for a minute, but one of the things that I came across last week, I think it was— I think I quoted it in 7 Takeaways, um, is if something helps you, if something moves you, if it's something informs you, does it matter where it came from?

Gary [00:09:15]:
Yeah.

Leo [00:09:16]:
Does it Does it somehow— in other words, if you were to read a piece of something and it spoke to you somehow without knowing where it came from, it moved you. Does finding the origin as being AI or human change your perception of the information that moved you? Um, that's— I'm not even going to say there's a yes or no answer to that. But it is something that I think a lot of people aren't thinking about. They are shooting the messenger without listening to the message. And I suspect that, well, the future is going to get interesting in that regard. Your personal podcast is a great use of— is a great example of using AI as a tool, which is something that I think you and I have been advocating a lot. We don't necessarily use it to generate content. We use it to massage content.

Leo [00:10:13]:
And yours is a wonderful example. of, you know, a lot of massaging going on, right, to package things up in a way that speaks, you know, that works for you. But I don't know, I just, I just kind of sort of think about these things sometimes when I hear people, um, just knee-jerk react to AI as any kind of, of generation.

Gary [00:10:34]:
Actually, that's going to be a theme with a lot of things I wanted to talk about on this episode, um, is people being outraged at AI And I think there are some legitimate reasons to be outraged at AI.

Leo [00:10:47]:
Sure.

Gary [00:10:47]:
But I think a lot of the things people are outraged at are the wrong things. Um, but we could take a break from talking from AI first and talk about something else that people are going to get outraged about.

Leo [00:11:02]:
Invasions of privacy. Yay.

Gary [00:11:04]:
So, okay, so there have been a string of different smart glasses that have had cameras in them. That have outraged people because you can be secretly recorded, right? Somebody walks in, they're wearing a nice pair of Ray-Bans, and you don't know it, they're actually doing a video recording of what's going on. Legitimate concern, I think. Now, this outrage is going to be thrown at something new, something that doesn't exist yet. So I think this is really neat. Do you have a pair of AirPods? Do you use AirPods at all?

Leo [00:11:35]:
I do not. I've got a pair of, I don't know, Android equivalent type things, but I don't use them very often. Things fall out of my ears.

Gary [00:11:44]:
Okay, so I use AirPods all the time. Like, what greatest product ever from Apple, because it allows me to take my music and audiobooks with me wherever I am. It's just great. Okay, so like, for instance, this weekend I went on a 10-mile walk, right? Music, audiobooks, just filled The walk with amazingness. Okay, so here's the thing. There's been rumors for a while that a new version of AirPods may come out that will have cameras in them. So they hang out from your ears just enough that you could see like there being a tiny little camera pointed forward that could be on the AirPods. Just rumors.

Gary [00:12:21]:
We don't really know anything at all for sure, except that a recent update to macOS they were able to find a video deep in that macOS version, something that may actually be kind of a like how you use it kind of video. If you go to System Settings now, for instance, you look at the trackpad settings, there's like these little animations that show you how to use the trackpad. Oh, double-click, here's how it goes, and it shows you. So there's occasionally there's little instructional videos that are included in the operating system. This little video that was found seems to show somebody wearing AirPods holding a book in front of them and asking for information about the book. And they get information about the book. The idea is, oh, it looks like there's a camera in the AirPods that are showing this. So perhaps next month's announcements from Apple are going to include new AirPods that have a camera in them.

Gary [00:13:13]:
Again, Apple's not said a thing about this, right? All we've got is this video and previous rumors. So of course, start the outrage on, oh, people are going to be walking around with AirPods recording you. Right. But I don't think so at all. I'm pretty sure that's not at all what's going on here. The definition of camera is a wide one. I don't think— and I don't see any possible way for there to be an actual, like, regular video camera in AirPods. First of all, the AirPods are small.

Gary [00:13:47]:
They have even smaller batteries.

Leo [00:13:50]:
Right.

Gary [00:13:50]:
That even you know, currently just playing audio with an incredible amount of, of like, you know, battery savings and software and chips that are optimized and all of that, you've got the, the problem of like the battery just will only last like 2 to 3 hours or whatever before you have to charge it up again.

Leo [00:14:16]:
Wow, okay.

Gary [00:14:16]:
Yeah. I mean, you know, well, the case has many, many charges in it, of course.

Leo [00:14:21]:
Yeah, yeah, yeah.

Gary [00:14:21]:
So I mean, like, for instance, on my 10-mile walk, it was longer than that. But at some point I walked into a coffee shop and ordered a coffee. When I did, I just took them out of my ears, stuck them in the case, put them in my pocket for a whole 5 minutes, which basically almost charged them completely up. Right. So easy, easy to even ignore the fact that they've got this limit. But camera's going to take a lot of power, especially if it's a recording camera, you know, and all of that. It's going to take a ton of power. So I don't think that's what is there at all, just for the necessity of battery, but also for privacy, 'cause Apple's big on privacy.

Gary [00:14:56]:
I don't think if they could put an actual recording camera in AirPods that they would. I think what we've got here is probably a camera in the broadest sense of the word. Some people pointed out it could be LiDAR, but I think LiDAR takes an amount of power too to actually send out and then come back, signals come back.

Leo [00:15:15]:
Yeah.

Gary [00:15:16]:
I think this is probably more in the lines of low resolution, black and white scanning camera kind of thing that produces no kind of visual that any human wants to really look at, but that you could feed to an AI and an AI could say, oh, I recognize a book and here's the title, right? The AI doesn't care how bad the image is, or maybe it's like scan lines. There's a lot of— if you go and say, hey, we need a camera that can tell AI what's going on, but doesn't have to look, you know, like be a visual video for a person, you could probably cut a lot of corners and create a camera that's really low powered. Another thing is it probably is not video at all. It probably captures an image, like it's kind of the camera version of E Ink, you know, where it can capture like a still, right? Kind of scanned monochrome lines of some kind, right? Plus, Apple has the ability to lock this down. I mean, they already lock their cameras down to prevent, you know, use that doesn't go through the proper APIs. So a still image, monochrome, not very good looking, good enough for AI that's locked down, that could only basically go to a process that uses machine learning to return a description.

Leo [00:16:36]:
Right.

Gary [00:16:37]:
So in comes light, out goes a text description. No camera actually involved. I think that's what we've got here. And I think that, you know, but people are going to throw around and they're going to see headlines, cameras in AirPods, and immediately jump to conclusions that, oh great, people are going to be recording people without their knowing. I don't think that's what's going to happen. I don't think it's possible given the battery and size restrictions for it to happen anyway. I don't think it's what Apple wants. I think this is all part of visual intelligence, which is a really big thing that Apple's you know, pushing in on all their platforms for AI to be able to describe something to you.

Gary [00:17:17]:
And so, yeah, I don't think that— I think we'll see soon enough, but it is not going to be the kind of thing where you can record at all.

Leo [00:17:25]:
It'll be interesting. I think that— I think the outrage is going to go one of two ways. Actually, it's going to go both of two ways. One are the extremes, right? They're just not going to believe that there's not some kind of video recording going on. Period. Um, you know, but I'll, I'll just lump those into the conspiracy theorists. But there's probably a more legitimate concern because Apple's got control of a dial that basically says how much can this camera understand. I'm not even going to say see, how much can it understand.

Leo [00:18:04]:
Um, it can apparently identify a book by its cover, right? Again, you know, right? See the title, give you information about it, whatever. But how much information does it need, as, you know, as rudimentary information as you describe as possible, to say, do facial recognition?

Gary [00:18:28]:
Yeah.

Leo [00:18:29]:
In other words, if you are talking with someone There's no picture being taken, there's no video being taken, but Siri quietly reminds you that you're talking to Leo only because you've walked up. Is that a concern, right? Is that something—

Gary [00:18:50]:
No.

Leo [00:18:50]:
I mean, obviously some people are going to get concerned about that, but the— in the bigger picture, is it a privacy, um, uh, you know, is it something to be concerned about with respect to privacy? And what other kinds of things could fall into that bucket where, again, There's no picture, but the camera, your device kind of sort of understands what it's looking at.

Gary [00:19:12]:
Right. And, and my concern too is that even if Apple locks this thing down in terms of privacy and has created something really good, that people aren't going to care. They're just going to see the headlines, right? And say camera on AirPods bad. And, uh, and that'll be the end of it. I even saw somebody saying, oh great, so all AirPods are soon going to be banned. in movie theaters or in—

Leo [00:19:37]:
Oh, wow.

Gary [00:19:38]:
You know, that kind of thing, because it's like they have the potential to have a camera. Not that phones have been banned in movie theaters, so they all have cameras. But the— yeah, so yeah, people will— might rally against it even if it can't do the things that they're claiming, because that seems to be happening so much now with so much technology, people jumping to conclusions, and it's just not getting that stigma off of it, right? So, so we'll have to, we'll have to see what this actually is. I am— if it can do stuff like just the visual intelligence stuff, I'm excited about it because I think it's a really interesting use of like— or not so much interesting use, but everybody's focused on glasses, right?

Leo [00:20:22]:
Right.

Gary [00:20:22]:
And some of the glasses now don't even have anything on the screen. Like you don't see anything. Like the lenses are actually just normal sunglass lenses.

Leo [00:20:30]:
Right.

Gary [00:20:31]:
Now the whole idea of like stuff being projected on the screen isn't even there. It's just a camera and a microphone in glasses. And Apple's already got microphones in AirPods. If they can put a camera that's just good enough to give you information, not take a picture, then they can go and say, eh, you don't need glasses. Wear glasses, don't wear glasses, wear your own glasses, doesn't matter. We're not gonna do glasses. We've already got AirPods. People already wear AirPods.

Gary [00:20:55]:
Everybody's already used to them. And we could do this visual intelligence stuff Giving you directions on the street, telling you things about what's around you, identifying birds, uh, with, uh, without requiring you to wear anything that you're not already wearing. Oh, and another thing, another way these could be restricted too is also focus. Like the photo, the little video shows the person holding the book.

Leo [00:21:20]:
Right.

Gary [00:21:20]:
That's another aspect of cameras. Like these may only be able to see 5 feet ahead in front of you.

Leo [00:21:26]:
Sure. Yep.

Gary [00:21:27]:
And, you know, again, this is like, why would you create a camera that can only see 5 feet in front of you? Well, this is a reason why you would create a camera that's only 5 feet in front of you. And if some inventor at Apple, some engineer has come up with, hey, I, I have a camera that can only see 5 feet in front of you and uses 1% of the battery life as a regular camera, and some other person went and said, oh, I have a use for that. That can help you identify things, right? But even though it cannot be used as a regular camera because it wouldn't be able to see more than 5 feet in front of you, everything else is blurry, right? That's an interesting, really interesting use, even if it's not 5 feet but some fixed amount, right?

Leo [00:22:09]:
Sure.

Gary [00:22:10]:
There's so many different things. The, the actual, um, like screens inside of, uh, Apple Vision Pro, for instance, like the highest resolution is shown where your eyes are looking. So it's actually— it looks like you have super high resolution everywhere you look because you can only look where you're looking, right? That's part of the magic of the Apple Vision Pro is the fact it doesn't need to show you high resolution in your periphery. It only needs to show it where your eyes are actually looking.

Leo [00:22:42]:
Yeah.

Gary [00:22:44]:
So, but you can't— you literally can't tell the difference because you can't look in your periphery. If you do, it's not your periphery anymore, you know. So same thing with a camera. Maybe somebody has at Apple has come up with a camera that literally will— can only look at certain things at a time.

Leo [00:23:01]:
Right.

Gary [00:23:01]:
And so no good for recording. Recording, it's just a mess. But for AI to be able to go and say, oh, that book is so-and-so, or the street sign says, you know, parking, you know, whatever. That's all that's needed. Anyway, I just—

Leo [00:23:17]:
I have this— the scenario I keep coming up with is, you know, I'm working on something and I'll have a, you know, some part. And what I want to do, what I've been doing, of course, is taking out my phone, taking a photograph of the part. And what I want to do is just say, you know, okay, Siri, remember this part number for me.

Gary [00:23:35]:
Yeah. So, so cool. All right.

Leo [00:23:38]:
So back to AI.

Gary [00:23:40]:
Onto it. Yeah. And to another thing people will be outraged about. Well, people are outraged about this. So the headlines have been— and I forget which company did it first. Was it, was it Anthropic or was it OpenAI? I can't remember one of the two. But now Amazon's being accused of this, destroying rare books. Right.

Gary [00:23:58]:
The headlines pretty much all universally say that these AI companies are destroying rare books. And the idea is, if you read just a little bit into it, is they are scanning books. They're getting— they're obtaining these rare books. scanning them to get the text, and in the process of scanning them, they're destroying them. Although some of the headlines really just try to make you envision this idea of like an AI billionaire walking into a rare bookstore buying like some rare copy of something, you know, and, and then bringing it, the leather-bound ancient tome, back to their AI lab where a scanner reads it and then they gleefully toss it into flames where it's incinerated. Yeah, the evil laugh. That's what the headlines are kind of like trying to— and what's actually happening is, you know, the books are coming in to scan them quickly. They're opened up so the pages are individual.

Gary [00:24:57]:
In other words, you know, the spines are broken, open up, the pages are scanned in to get the text in, and now there's just a bunch of pages, so the— which are then thrown out or recycled or whatever. Um, but the key word that keeps appearing in all of these headlines is the word rare.

Leo [00:25:13]:
Rare.

Gary [00:25:13]:
Which doesn't mean what you think it means. Now, here's my background on this. I know a lot about all this AI stuff and all that. I also used to own a used bookstore. My wife and I used to own a used bookstore, so I know a little bit about books. And the thing is, when you hear rare, you think valuable, but rare doesn't have to mean valuable. And as a matter of fact, it very often doesn't mean valuable at all. Quite the opposite in books, right? We used to, when we first opened our used bookstore, we needed to fill the shelves.

Gary [00:25:43]:
And one of the ways that we filled the shelves was there are actually sellers out there that will sell you 20 boxes, big boxes of books, random books, and you could just buy 20 boxes of books. The cheapest way to buy books. Now, if you do that, and I— we did many times at the beginning, you open these boxes up and most of the books are worthless. If you really wanted to buy them, you could maybe find a few books that were actually worth something. Not rare by any means, you know, not the Dead Sea Scrolls. They're not going to show up in there. It's going to be you know, oh, hey, this is a book that'll sell in our bookstore for $3. Right?

Leo [00:26:30]:
Right.

Gary [00:26:32]:
This is good. We'll keep this. You'll find a few of those. You'll find a bunch more that maybe will be worth filling out the shelves for a while. And the vast majority are going to be junk. Junk means mass-produced books. Maybe they were popular back at some time and hundreds of thousands, if not millions, were produced. Everybody that wanted a copy got it.

Gary [00:26:54]:
It's a very common book to find. But they basically printed too many. And then years go by and now it's a rare book which you don't see anymore because nobody wants it, right? Tons of those, tons of stuff, you know. And my books would be among those. My old computer books for old versions of old software that you can't even use anymore are probably rare now because I'd imagine most copies— there weren't that many copies made, and most copies that were made now probably don't exist anymore. So they're rare, but they're worthless if you get them. There's no value to them. You wouldn't want them.

Gary [00:27:28]:
Nobody's going to want to read them now. And even fiction books, the same thing. A lot of times these old books, books that were marginally popular in the '60s, the '70s, the '80s, tons of copies got out there. These are not books that are going to be lost to time because they're available in digital form. They they're just not worth anything, right? And that's— these are the books you get if you order these boxes of 20 books, right?

Leo [00:27:54]:
Right.

Gary [00:27:55]:
So rare in the sense that you're not going to see these books often, but not valuable. Now, why are they valuable at all to the AI companies? That's why they're getting them, right? Because they have written words in them, sentences that people have written, speech in quotes that authors have imagined fictional characters say. And that's what the AI companies want. They want language. They want how do people talk? How do people write? They don't care the subject matter. They care about what words usually go after what other words. And the problem is that if you look online today at what people are writing, blog posts and such, a lot of those are written by AI. Right.

Leo [00:28:38]:
So you—

Gary [00:28:39]:
or edited by AI, improved by AI. So if you scan the internet for stuff, you're going to end up feeding AI with old— with existing AI, and you get this feedback loop which isn't good for the AI models, creates a problem. So they want writings that have been written by humans. And the way to do that is to look for anything that's been written from before 2022. So you go back to 1970, you know, a book written in 1978 or 1994, and it's some novel. that didn't— it got a bunch of prints, but it didn't hit right. And that's a good source of this stuff. And that's the books they're getting, not rare tomes, leather-bound from valuable bookstores.

Gary [00:29:24]:
It's these trash books that are headed to the trash anyway. Because I can tell you, when we got these books at the bookstore and the stuff we definitely didn't want, we never directly threw away. But they went to Goodwill, and at that level they probably did look at them and were able to tell this goes to recycling, right? There's no— it's not going to be sold in the store. Even at regular bookstores, like, an important part of used bookstores is to cull the inventory, go through. Sometimes if you go to a used bookstore— not that there are any left— but if you go to one, sometimes you'll see little colors, like the prices will be in little dots. There'll be a green dot and red dot, blue dot. Sometimes it'll just be— sometimes you look at the price and there'll be a marker stroke underneath it, like a colored mark. A lot of times these are the year that it was obtained by the bookstore.

Gary [00:30:14]:
Sometimes they'll just write the year. So you go through your shelves and you look at, oh, this has a green dot on it. That means we bought it in 2023 and it's still here today. So we need to pull it off the shelves. It's not going to sell. We need to make room for other stuff.

Leo [00:30:29]:
Yeah.

Gary [00:30:29]:
And these books will eventually find their way to recycling anyway. I would say that I would bet that 99% of the books that were headed to, you know, any of these AI scanning things were not going to survive very much longer anyway.

Leo [00:30:46]:
Right.

Gary [00:30:46]:
Sure, stuff's going to get caught up in there. But I'd say that even anything that's in there at all is not rare. It's not going to— the last copy in existence isn't being destroyed or any of that nonsense. So plenty of stuff to be outraged about when it comes to like the financial side, when it comes to the environmental side of AI and stuff. But this, this is really—

Leo [00:31:08]:
What I find interesting about this, uh, first of all, I hope that, um, as they process these books, they're digitizing them, for lack of a better term. They're doing the moral equivalent of what archive.org does, right?

Gary [00:31:27]:
Mm-hmm.

Leo [00:31:28]:
Where they're, um, basically photographing each page and saving the photographs. Even if they don't use that, just save the photographs because you are essentially preserving these books at that point by taking these digital images of them. Then you can run your, you know, OCR on it and then run your AI training, your LLM training on the result. Um, but one of the things that crossed my mind is, in, in the stories that I've read, is that it always points out that the AI companies are buying these books. Okay, does that now sidestep the plagiarism slash theft?

Gary [00:32:13]:
That's a whole different story. Yeah.

Leo [00:32:16]:
Um, Because if, if a human can walk into a used bookstore and buy a book, and that's okay, buy a used book, then it would seem that an AI training their model could walk into, um, grab, purchase a copy of that same book and quote unquote read it.

Gary [00:32:42]:
Um, I'm—

Leo [00:32:43]:
again, I'm not, I'm not going to claim that this is as simple as that, but it does color the argument that AI companies have been stealing, plagiarizing, uh, content. Right.

Gary [00:32:58]:
Yeah. And that goes back to something we talked about before. It's like if you could walk into a store, buy an Ernest Hemingway book, read it, and then the next thing you write kind of has a little bit of Hemingway flavor to it.

Leo [00:33:11]:
Yes.

Gary [00:33:12]:
Maybe even a vocabulary word you didn't know before you read the Hemingway book, you know. I mean, and of course it's going to. I mean, that's the idea, right? Any writer writing today is going to be influenced by everything they've ever read, and so is the AI. So like, what's the legal standing there of like, you know, I mean, it— can you, can you claim plagiarism if AI puts one word after another, and it got the idea to put that other, you know, to use that adjective because it read 100,000 books and determined that's a good adjective to describe that object.

Leo [00:33:48]:
Right.

Gary [00:33:48]:
Can you then go back and point at a book by Ernest Hemingway and say, well, Hemingway used that adjective?

Leo [00:33:53]:
Yes.

Gary [00:33:53]:
You know, so because that's what I mean, plagiarism is like if you took a sentence, I mean, if the AI went and took a sentence out of a book Put it— gave it to you as something it wrote. But it's not doing that. It's learning to write by reading these books.

Leo [00:34:09]:
There's a parallel that's been around in the music industry for a long time where you end up with somebody writing a song that happens to include a beat or a melody or a something that coincidentally happened to exist in a prior song. And the question is, were they influenced by that prior song, or did they come up with it completely on their own? And there's actually no way to prove one way or another. And magicians, songwriters have been sued, sometimes successfully, sometimes not, over this very issue. And I think that this is the exact same issue we're going to start facing with AI, because honestly, the chances of AI writing a sentence that has never appeared in a book ever is decreased dramatically, right? So does that imply that it plagiarized it from that source? Well, no, not necessarily. It could just as easily have written that sentence on its own.

Gary [00:35:12]:
As you or I could have. Yeah.

Leo [00:35:14]:
Yeah. So it's— it— yeah.

Gary [00:35:16]:
But my point here is that, yeah, AI companies are not like buying old lost leather-bound tomes and throwing them into incinerators. But people are finding it very useful to actually get outrage. Maybe they're just— maybe they're just trying to get clicks. Maybe that's all it is. They're not really trying— they don't really care about the outrage. They're just trying to get clicks on their articles. But yeah, so I don't know. A lot— like a lot of things, there's so many good reasons to be outraged.

Gary [00:35:46]:
Um, that are legitimate. But, uh, yeah, so it bothers me. Anyway, what are you outraged about?

Leo [00:35:55]:
Let's step away from AI. Oh, I'm not outraged. I'm just really outraged.

Gary [00:35:59]:
Okay, just curious.

Leo [00:36:01]:
I'm just smug.

Gary [00:36:03]:
Smug.

Leo [00:36:03]:
So the, the— a PBS station, um, happened to have 70 years of archival TV data at a cloud storage provider, and that cloud storage provider suddenly and without warning went out of business.

Gary [00:36:28]:
Hmm.

Leo [00:36:29]:
And the PBS station found that it could no longer access all of this wonderful historical data. It's clearly a local station. I'm sure at the national level PBS has everything, you know, multiple— multiply backed up in various places. But, um, in this case, this was a specific PBS station. I'm not even sure where. And so the cloud provider went out of business, like I said, silently and without warning. They went to renew their license and nobody could contact them. And all of a sudden, when their contract was up, they could no longer access their data.

Leo [00:37:07]:
They ended up having to actually sue the third party that owned the servers, Iron Mountain. You've— I'm sure you've heard of Iron Mountain. They apparently do some cloud storage services as well. Typically we think of them as— ironically, we think of them as a backup service or a data destruction service, depending on what it is you're attempting to do. So they had to sue Iron Mountain to regain access to their data that was still on their— on those servers before Iron Mountain went through and said, well, this third party went, you know, this, this cloud storage provider went away. Um, it's time to recycle these servers. Um, it makes sense that they had to sue because Iron Mountain's contract was with this cloud provider, not with the station. So the station was claiming that the data was theirs, but in reality, the data belonged to this cloud storage provider that had gone away.

Leo [00:38:07]:
So basically, it had to go through the courts to prove ownership. And indeed, the station got access to their data back. Great. What was it, something like 70 years, 50 terabytes? So a significant chunk of data. I'm looking at this and I'm thinking, this is something that I talk about all the freaking time. If you've got data and it's in only one place, it's not backed up. So if you've got it with one cloud storage provider, and I don't care which cloud storage provider that is, this was apparently a fairly obscure one. But, you know, I don't care if it's Dropbox or OneDrive or Google Drive or any of the others.

Leo [00:38:58]:
If it's in only one place and you for any reason lose access to that cloud provider, Or heaven forbid, that cloud provider goes out of business suddenly and without warning. You've lost your data. Not everybody has the resources to go in and sue somebody in the hopes that they can recover their data from a backup. You have responsibility for backing up your own data. And I think what a lot of people miss about the cloud is that It really is only one place. Yes, OneDrive, Google Drive, etc., Dropbox, they've all got multiple servers and they've all got backups, but those backups are for their protection, not yours. If you lose your account, if you delete your files, if your account gets hacked, they're going to say tough luck, right? We did what we were supposed to do. Your data is gone.

Leo [00:39:56]:
you can't get at it anymore. Anyway, I just see this as an object lesson in exactly what I've been preaching for many, many years now. And that is, if it's in only one place, it's not backed up. The cloud is only one place. For gosh sakes, if it's something that's that important, especially if it's 50 terabytes important, get yourself a NAS, get yourself a couple of NASes, get yourself another cloud provider. I don't care. Make sure that you've got it in— Yeah.

Gary [00:40:24]:
Yeah, exactly. Unless it's just not that important.

Leo [00:40:28]:
I think it's funny, I get that feedback a lot on a very— on various pieces of, of things I talk about. Usually it's specific accounts, but, um, I had one today where, um, somebody was saying, well, why bother backing up? I mean, yes, I've got all these pictures, but when I die, the kids are going to reformat my hard drives anyway, and it'll just be gone.

Gary [00:40:50]:
Yeah. I mean, that's one way to look at it. You have to— it's all risk assessment.

Leo [00:40:55]:
It is.

Gary [00:40:55]:
That's a whole other subject. I didn't mean to—

Leo [00:40:58]:
My sense is that people underestimate the value of what they have. That's not to say that they don't have some data that is truly worthless, but as a whole, when they're putting together their backup strategy or their security strategy or their whatever strategy, they completely miss or miss misjudge the value of something they have.

Gary [00:41:21]:
Oh yeah. No, you have to do the exercise in your head of like, if I lost all my data now, like what would be the implications? Like what things would I— I mean, people have done this for generations with fires, right? You know, you're told to think of like what if your house— if you came home from work and your house had burned down, what things would you miss the most? And then you realize, oh, my photo albums. Like I never even thought about the photo albums sitting under, you know, on the shelf or whatever. Maybe I should digitize them. Maybe I should have a plan to, you know, protect them. But you have to go through the exercise in your head and you do the same thing with your data. You could think about like, because I've done it with my data for backing up all of my video editing files for my episodes, and I've come to the opposite conclusion. Like I've been, I for years was meticulously backing those up and then going through the, okay, if I lost all my data, what would I miss?

Leo [00:42:14]:
Yeah.

Gary [00:42:14]:
And then I was like, well, 90% of my backup are these old video editing files, not the final ones, but the edited files. And I realized, okay, why would I miss those exactly? Like most of those are, you know, older types of files using older video editing software. I've never, ever had to go back and use one. Like, why would I need to re-edit a video about how to use iMovie 06 from 2009? You know, it's Oh, okay. I probably— I would really wouldn't miss it. Maybe I've been obsessing over backing those up and I really need to focus on making sure my photos and everything, which were getting backed up just fine too. But yeah, go both ways.

Leo [00:42:56]:
The video is a good example because I too, I'm still in the bucket of meticulously backing up my video editing files, but I'm backing up my current ones. Yeah. And just letting them accumulate. Right. So I have those editing files from 8, 10 years ago, and I know that I'd never use them and I would not cry if they disappeared. But I'm not going to actively, you know, proactively go out and get rid of them. I just let them sit there.

Gary [00:43:21]:
Actually, this is— I don't want to get off on another tangent here, though maybe I should. I actually looked into doing online backups for some of that old, old stuff that I'm not that worried about. Like, how can I? Because the price of drives has gone up and skyrocketed. It really has. Because I looked into— I have a bunch of 4 or 5 terabyte drives that I've just had for years, back when that was like, that was the big drive, right? And I got a few of them and I used them for archive drives. And I thought, well, let me get, let me get another drive because, you know, I'm starting to run out of space a little bit. And drives are way more expensive than they— at first I thought I was imagining it, like, you know, the, the rose-colored glasses, like, oh, I thought drives were cheaper. Well, I guess it's just been a while since I bought one.

Gary [00:44:06]:
And then I looked through my Amazon orders and I was like, like, no, these drives were cheaper. They, they have gotten way more expensive. So I thought, well, maybe online storage is cheaper now. And it is really cheap if you look for archival storage. You can upload files and store terabytes worth of data, and it would cost, you know, for— you can do a lot for, say, $10 a month, ton of stuff. But there's a big cost in getting it back. So the idea is if you want to store a ton of stuff and then you're not worried about getting it back because you probably never need it, or maybe if you need it, you just need this one file, it's cheap. However, if you wanted to get it all back, like, oh, the price of drives dropped and I was like, I don't need this anymore, let me download all of it and store it.

Gary [00:44:58]:
That's expensive. That was way more expensive than any of the other stuff. Like it was a factor of 10. It was like 10 times more expensive to download it than it was to upload it. And what occurred to me was most of these services, it's like, what if I wanted to switch services? Like I picked a service now and 5 years later there was a deal on another service, right? And I wanted to move over to it. I would then have to spend hundreds of dollars downloading all my data So I can upload it to another service. And that's what stopped me from doing it. I was like, that's the pain point right there on this.

Gary [00:45:34]:
So that was kind of interesting. And yeah, I don't archive anything online.

Leo [00:45:39]:
If you do end up having to download absolutely everything and then re-upload it again somewhere else at this point. Yeah. If you're going to download it all, well, grab some old drives, maybe invest in a new one because it's going to last and just put it there.

Gary [00:45:54]:
Yeah. Yeah, that's what I'm doing. So maybe one more, one more outrage thing.

Leo [00:46:02]:
Okay, go for it.

Gary [00:46:03]:
And this is, this is actually— I want to— I don't really have— I'm not actually outraged by this. I thought we could discuss this. I don't know how much you've read about it, but Claude announced in the last week or so that they're going to start watermarking things generated by AI. Now, images, pretty straightforward. You can put a watermark hidden in the bits in an image. And it's easy to understand, right? You've got an image, it looks beautiful. Everybody looks at it. Oh, AI-generated image.

Gary [00:46:32]:
If somebody wanted to prove it was AI-generated, they could look at the watermark built into the, into the bits and say it is AI-generated. For people like you or me or most people, it wouldn't be even a thing because we would say, yeah, it's AI-generated. But, you know, I'm not hiding that.

Leo [00:46:49]:
My video images like that, I say, yeah, your blog post videos. Yeah.

Gary [00:46:53]:
And Gemini or whatever. That's fine. I've never used an AI image where I've wanted to hide the fact. It's just been like, oh, here's a fun image I created. I do it for my monthly little— I create these whimsical little things that are really fun, but I would never spend $100-some having a graphic artist create these. These are just these ideas I have in my head and I just have Nano Banana create something fun. But theoretically, you would— if you don't want people getting away with saying, oh, this is a real image and it's AI generated. But I think even more importantly, you want to have services like perhaps Instagram be able to label things as AI generated easily.

Gary [00:47:34]:
So they have a little list of rules. Hey, here's the 5 most popular services. Here's the watermark you should look for. Fine. But this isn't about AI image stuff because AI image stuff is pretty straightforward in watermarking. What's more interesting is text.

Leo [00:47:45]:
Yeah.

Gary [00:47:46]:
How do you watermark text? And probably the AI companies just wouldn't if it wasn't for the fact that the EU said you have to. So they're left scrambling with how do we watermark text, right? I mean, if the text actually contains in parentheses generated by Claude, generated by OpenAI, whatever, then it's just easy just not just to delete that, right? Just don't include that when you post it. So Claude or Anthropic came up with this thing for Claude that AI-generated text, I assume it has to be reasonable, reasonable length, is watermarked by making certain choices of words. And so it'll generate text that looks like it's written by a human. And if you run it through their little tool that they say they'll provide to people, it'll look at the choices of words and say, ah, this was generated by Claude, right? This is watermarked. And at first people were outraged because it was like, oh, so you're going to make the writing less good because, you know, you want to watermark it, right? That people are outraged from that. People are outraged that, oh, this won't solve the problem because you could just change the words You know, but they say, look, it's, it's going to be in there and it's going to be some— it's going to be choice when there's an ambiguous choice of words, when it doesn't matter if it's this word A or word B. We're going to make those choices in such a way that we could then identify if you take this text and throw it through our tool, it'll say, yes, this was generated by us.

Gary [00:49:30]:
Still seems a little bit like hard to believe, but I know enough about algorithms, computer algorithms, to know that sometimes it really doesn't take very much.

Leo [00:49:39]:
Yeah.

Gary [00:49:39]:
To do stuff like that. But, you know, is this good? Is this bad? I don't— I don't know. I don't really have a big problem. I don't feel outraged that they're going to do this. I mean, if the writing is not as good, well, you could always have just written it by yourself.

Leo [00:49:55]:
You know, it feels to me kind of pointless. This is kind of an arms race, right, where they're going to watermark their stuff and people are going to reverse engineer the algorithm and remove the watermark. And, uh, just as they do with images, right, it's not that hard to figure out what the watermark is doing to your image and actually take it out. Uh, for a while, I think Gemini is still— Nana Banana is still putting their little diamond in the lower right-hand corner of those images. Those come out easily because I've been taking them out for months.

Gary [00:50:29]:
Yeah.

Leo [00:50:30]:
But it seems pointless. It also— what's funny is I actually have heard of people saying, oh, this is why Claude suddenly became wordier.

Gary [00:50:43]:
Ah.

Leo [00:50:43]:
Because it needed more words in order to implement this algorithm.

Gary [00:50:50]:
Hmm.

Leo [00:50:51]:
So I don't know.

Gary [00:50:53]:
Well, yeah, it'll be interesting. I still think a lot of the outrage at things like this, whichever side you fall on, and the images and all sorts of stuff, fails to comprehend that a lot of the use, perhaps the majority use of AI is for personal use. And I'll go way back to the beginning when I talked about my podcast. I did not ever have any intention of this being something that anybody else but me would listen to, would even want to listen to. It's such a weird collection of things that I want to hear, right? And often AI is used to summarize something. Often AI is used to tell you about something. You know, give you information or whatever. I would say, maybe, maybe it's way up there in the high 90% of all the stuff that AI generates is basically a personal, like, giving you just that one person who asked for it that we tend to look at it.

Gary [00:51:56]:
And I think people that don't use AI much tend to look at it as if you're asking ChatGPT to write something, it's because you're publishing something, you're publishing a book, and you're getting AI to write it. You're making a blog post and you're getting AI to write it. When in fact, it's— if you're asking ChatGPT to write something, it's because you're going to read it and then that's it. It's all— it's done.

Leo [00:52:16]:
Right.

Gary [00:52:17]:
You've read the thing that it produced for you, that some weird bit of text about some subject. I do this all the time. I ask it— I asked it for an explanation. I was watching— I'm rewatching Northern Exposure.

Leo [00:52:33]:
Oh, wow.

Gary [00:52:34]:
The episode— the episodes, if you watch online, the numbers don't correspond to what's in Internet Movie Database at all.

Leo [00:52:43]:
Oh, interesting.

Gary [00:52:44]:
And, and yes. Oh, interesting.

Leo [00:52:46]:
So why?

Gary [00:52:47]:
Why? Is there anybody that's ever written a blog post about this? Not that I could find. ChatGPT, tell me why this is so. And I read probably 7 paragraphs that ChatGPT wrote for me that was essentially an article. explaining why. It's because CBS originally ran them out of order in some cases. And so when they produced the DVD box set years later, which was done by the production company, they put them in the proper order, and that's been picked up by streaming. And it makes it very confusing as to what episodes belong in which seasons, because sometimes the ordering included different seasons. Like an episode made for season 2 was actually aired as part of season 3.

Gary [00:53:30]:
So anyway, interesting article that I read about that, except that there's no article about that. This was something ChatGPT wrote for me because I asked it to. And I think this is what— this is what AI is. And I'm not publishing it. I'm not putting it anywhere. I don't have any place to put it.

Leo [00:53:47]:
So— And you don't care if it's watermarked?

Gary [00:53:49]:
I don't care if it's watermarked. I don't care if it's plagiarized. I would have preferred to actually read somebody's article. that they wrote about this, in which case I would have been reading an actual article. But I didn't find an article, so I read what ChatGPT— and I'm just reading it this one time and that's it. I'm not saving it anywhere. So anyway, I think that's people overlooking a lot of that, the just personal use stuff. We don't need to hire somebody to write an article about why Northern Exposure episodes are out of order.

Gary [00:54:22]:
We can just ask ChatGPT now and it writes it for us using all of the books that it scanned in the past to give, give me a coherent thing that was fun to read.

Leo [00:54:34]:
For the record, Northern Exposure was a cool show.

Gary [00:54:37]:
And on rewatching it, it's— yeah, I hear a lot about shows when you watch people saying you got to watch this show again. They don't make shows like this anymore. And then I watch them and I was like, yes, they do make shows like this. You're just not watching them now. I'm watching Northern Exposure and I'm like, you know, they don't make shows like this anymore. It is— there's a different feel to it that, yeah, it's kind of— it's interesting.

Leo [00:55:05]:
We have local connections. This show was shot in Roslyn, Wisconsin. Yes, that's right. Which is probably about, I don't know, about an hour and a half drive from here. And Janine Turner, Um, apparently had a house about half a mile from where we were living at the time.

Gary [00:55:20]:
Oh, okay. Small world. Yep.

Leo [00:55:23]:
Wait, that's cool. What a segue. So for this week, for This Is Cool, my wife is out of town and we decide— or I decided that it was a great opportunity to catch up on a couple of things that she's typically not as interested in as I am. So my Ain't It Cool for this week are the movies Spider-Man: Far From Home and Spider-Man: No Way Home. Um, in prep for Spider-Man: Homecoming, which apparently has a big box office success over this last weekend.

Gary [00:55:55]:
Yeah.

Leo [00:55:55]:
Um, so, uh, I had not caught up on those 2 movies yet. Enjoy the heck out of them. Uh, they are, as with most Marvel movies, they are absolutely not Oscar contenders, but I felt entertained for the entire 2 or 2 and a half hours of each. And that's what I want out of my, out of my Marvel movies, just entertainment. So, uh, yeah. Spider-Man movies catching up just before Homecoming. And as is so often the case these days, um, we probably won't see Homecoming in a theater. We're just going to wait for it to show up on streaming, um, which is unfortunate for many theaters, but, uh, the reality is My home is quieter, cleaner, and I can hit the pause button when I need to go to the bathroom.

Leo [00:56:49]:
Exactly. Yep.

Gary [00:56:50]:
I'll— I don't need to hit a pause button when I read a book, actually. No, I take that back. I do, because I listen to audiobooks.

Leo [00:56:58]:
No, you can wear them in there. It's no problem.

Gary [00:57:00]:
Yeah. Well, yeah, but yeah, latest book I just finished reading is called A Walk in the Park by Kevin Federico. It is a book about Uh, 2 men hiking the Grand Canyon, like the length of the Grand Canyon, which is an incredibly difficult thing to do. There's no trail that goes through the Grand Canyon. Quite often the walls are vertical, so you have to be up at different layers and levels and everything. It's a very difficult thing to do. What I like about this book and why I recommend it is this is not a book about 2 expert adventurer hikers that know everything that go on this trek and teach you how to do it. These are 2 very ill-prepared people that should not have been doing this.

Gary [00:57:49]:
And, but they get help from the Arizona hiking community, particularly a bunch of individuals that know a lot about hiking the Grand Canyon along the way. And it's inspiring to, to have them, you know, read about them learning about this on the, you know, while they're doing it. seeing the many mistakes. If you get the whole, like, you know, mistakes are, you know, the way you learn things— well, this is a book about that because it's just about mistake after mistake after mistake. Um, it's very interesting. And the most inspiring part of it is after their initial failure, really at the very beginning, to really make— to do— you know, they had to, like, bail out really early. Um, a bunch of these hikers came right to them and said, okay, we're going to prepare you for getting right back in there.

Leo [00:58:38]:
Right.

Gary [00:58:39]:
Instead of telling them, okay, did you guys learn that you're not able to do this and will you go away now before we have to rescue you? Nope. They were like, let us show you how to do this. There was never a— are you sure? They don't even ask them, are you sure? Are you sure you want to continue or do you think you want to get— no, they right away were like, let's show you how to do it right.

Leo [00:58:59]:
Nice.

Gary [00:59:00]:
Really, really cool book.

Leo [00:59:04]:
So blatant self-promotion. The article I'm pointing people at this week is Tips, Tricks, and Best Practices for Sharing Links. It's askleo.com/195019. And it will sound very, very familiar to you, Gary, because I basically took a look at one of your recent videos and decided that my audience needed their Windows equipment So I did. I've basically followed most of your, your tips and tricks along the way. And like I said, put it in a Windows context so that my audience is a little bit more comfortable with it.

Gary [00:59:38]:
Excellent. Cool. I'll have to, I'll have to link, put a link in on the video to yours. That's what I'll do.

Leo [00:59:46]:
I did link to your video from my article.

Gary [00:59:49]:
Okay, well, there you go. All right. I'll just point out one video recent— I did recently. Why is my Mac menu bar missing? We often say that there's always a menu bar at the top of your Mac, but there isn't. There are cases where you can actually have it go away and then people get very confused as to why it's, why it's gone. And I explain all the different situations you can get into where the menu bar isn't there and how you could bring it back.

Leo [01:00:14]:
Cool. I actually did not think about it being possible. I have Like a multi-year-old article. The Windows taskbar can do the same thing. It can disappear for a variety of reasons. In Windows cases, the most, the most common reason is that one of the key processes terminated unexpectedly. So you end up having to like figure out how to rerun a particular process. But cool.

Gary [01:00:41]:
All right.

Leo [01:00:41]:
I think that does us Once again for another week. As always, thank you for listening, and we will see you or hear you or talk at you again real soon. Take care, everyone. Bye-bye.

Gary [01:00:55]:
Bye.