Leo [00:00:24]:
Welcome to this week's edition of Grumpy Old. I was looking at the list of things we have to talk about and, and while some of them are enthusiast kinds of things, I think the, the, the. The balance goes towards the grumpy side of things. We have some things to. To be concerned about. I did want to point out, somebody pointed out this week that our synthetic voice has problems with our names. It actually tends to get noten boom correct, but it has a real problem occasionally with Rosenzweig. And what I find interesting is that the names are in the recording, twice in the introduction, which presumably our listeners will have just heard, and then again in the outro as a reminder of who the heck we are.

Leo [00:01:14]:
And it pronounced your name differently in each. In each case.

Gary [00:01:19]:
So it's weird, huh?

Leo [00:01:21]:
It is weird. I suspect that it's a matter of entering in not your name spelled correctly, but your name spelled phonetically. For all I know, that's what. What Connie's done with my name already. It's much easier to go note ampersand, boom, and you end up with note and boom. Yeah, I don't know. I don't know what we end up doing with Rosenzweig, but yeah, it's. It's certainly worth playing with and spelling correctly, apparently.

Leo [00:01:56]:
Yeah.

Gary [00:01:56]:
Yeah. I was just noticing in our. In just your little text there. And it's kind of a placeholder in our notes to say the intro goes here and then we talk. I noticed you typed my name wrong. And the most common misspelling an S instead of a Z. But it still should have gotten it right. I mean, if, if there is anybody with two S's and their last name Rosenzweig, it should probably be spelled this pronounced the same.

Gary [00:02:21]:
I don't have. I. I have the opposite problem. I have a. I use Mac Whisper to do my transcripts for some things for my videos. And there are. Mac Whisper is just a little. It's a Mac app and it's named after Whisper, which is OpenAI's whisper AI thing, which was their original deal when OpenAI started.

Gary [00:02:42]:
And they still use their Whisper models and stuff for you talking to chat GPT. And there are a ton of different models you can plug into Mac Whisper to do your transcribing. There's one in particular that gets my intro perfect with MacMost.com every time. But every other one I've ever tried will get it right sometimes right. And other times it'll be like Mac Space Mo, you know, dot com or,

Leo [00:03:14]:
you know, is what I've been seeing occasionally too.

Gary [00:03:17]:
We just weird stuff. But this one model for must have been trained on something that's got MacMost.com in it and it's rock solid. Yeah, really. Maybe. Maybe some of my stuff ended up in there, which is actually going to tie into something we'll talk about later.

Leo [00:03:34]:
One of the things that I do after Connie posts the actual episode, the teh episode is I then grab the audio and create a YouTube video on it. So in case people are interested in subscribing on YouTube, they can do that. But I have DaVinci Resolve create basically subtitles for the whole thing and then burn it into the video. And that's where I've also noticed that while I won't go through the entire video and correct everything, I will start at the beginning and at the end and make sure that our names are spelled correctly and our websites, most importantly our websites. Misspell my name all you want, but make sure people can find my website and we're good.

Gary [00:04:12]:
Yeah, yeah, exactly. That's. That's the important thing. So, yeah, lots to talk about there.

Leo [00:04:18]:
Anyway, I got a few random short things about AI that are related. Nothing huge. One of the things that I've ran across actually a couple of months ago is that a lot of servers are getting periodically overloaded with bot traffic. It's one of those things where all of a sudden I've noticed it on mine that all of a sudden the server comes to a crawl and if you take a look at what traffic is hitting it, it's like a specific bot or a bunch of bots or who knows? Theoretically it's AI scanners, but it could also be search engines and other things. Could be even college projects where people are having to write something that scours the Internet as part of their senior project or something. Anyway, that's one of the reasons you see Cloudflare so much more often these days. Cloudflare can act as a front end to your website and presumably keep out the bots. And you've seen, you've.

Leo [00:05:19]:
I'm sure everybody has seen this now where you get the. The Cloudflare interstitial that basically says some form of prove you're a human. That seems to be happening more and more. I've not done it for Ask Leo. I don't plan to do it for Ask Leo, but I've got a couple of other websites where, you know, the websites actually don't need to be indexed. They don't need to be slurped up into somebody's AI, large language model. Maybe we'll just throw Cloudflare in front of them so that I don't have to buy a bigger server, which is what I ended up doing for Ask Leo, so I don't have to buy a bigger server to handle all this additional traffic.

Gary [00:05:57]:
Yeah, so something interesting happened to me this week that's kind of the same. Fits in the same thing I was asking. I forget if it was. I think it was Claude. Occasionally I'll do things like I ask their Clotter chatgpt, hey, what video topics have done well for me in the past but I haven't talked about recently. And yeah, I'm pretty sure this was regular Claude. And Claude said, well, give me a list of your. Your videos.

Gary [00:06:28]:
And I said, oh, okay. And I have a page that actually lists all my videos with the dates so it can look and see how old things are. And he came back and said, yeah, I can't see that web page. I'm being blocked by your, your server. And I was like, hey, you know, I do have a fairly long list of bot blockers like that I've copied. You know, these are IP addresses that are bots for var. And yeah, there's AI stuff in there. So it actually stopped me from doing something because Claude couldn't see my own site, which I was asking it to look at.

Gary [00:07:05]:
And then that also got me thinking about more and more people are using AI for web searching, either through a web search, you know, part of the bot, you know, like going in and asking ChatGPT, oh, you can do a web search for me, do this, or using Perplexity or something like that. And others are just doing it without actually being so formal about it, just asking chat GPT, hey, recommend some websites for me that could do this.

Leo [00:07:33]:
Right?

Gary [00:07:34]:
So now I'm concerned. It's like, well, if I block my site, if I block all these bots, will I be then left out? Yep, in some cases where I don't want to be left out, like, and there's probably no solution for this, it's like, either let the AI see your site or don't.

Leo [00:07:49]:
Right?

Gary [00:07:49]:
Either get the benefits and the drawbacks or nothing.

Leo [00:07:54]:
That's exactly why for Ask Leo specifically, I opted to just make a bigger server so that it could handle the additional traffic. Because I want to be found, be it in search or AI or whatever. I just, I knit. It's part of the business model, to be honest. So. But like I said, these other sites that are less, less about business even, they don't need necessarily to be found. So I'll probably throw in a few blocks. Something else that came up, AI and politics.

Leo [00:08:25]:
But it's not AI and politics. It's kind of. There's two approaches to this. There are two aspects to this problem. The one that everybody of course thinks about right away is that there'll be lots of deep fakes AI generated content and so forth to try and essentially manipulate the the elections or whatever is going on specifically here in the United States. We've got the midterms coming up in November, so there's the expectation that they'll a lot of stuff being posted online, some of which is accurate, some of which is fake, and the difference will be very difficult to to determine. That's actually not what this is. What surprised me.

Leo [00:09:05]:
This article from NPR link in the show notes is all about how the AI companies are basically becoming politically active and supporting political candidates to try and get them to basically make decisions that favor whatever the AI companies are are interested. Conversely, we also have other organizations that are basically funding political initiatives to say block AI or make it behave more correctly or basically blocking some of the data center build outs and so forth. Anyway, I just thought it was really, really interesting that AI is having a dramatic impact on the upcoming election cycle in these two completely separate ways. One, one of which is very traditional political activism and funding, and one which is actually brand new to. Well, I shouldn't say brand new. It's always been around, but certainly has scaled up dramatically with the use of the various AI tools that are available now.

Gary [00:10:09]:
Indeed.

Leo [00:10:11]:
All right, another one. And this again goes back to our servers. We talked a couple of weeks ago maybe about Mythos, the large language model, the AI model that that Anthropic had created but had decided it was too dangerous to release because it was too busy finding all of these vulnerabilities in mainstream operating systems and other tools. I think that one of the fallouts from that right now is I've basically seen an increased rate of updates to my servers now on the Windows side. That's all kind of sort of transparent to me. I just let Windows Update do its thing and ultimately it ends up typically getting batched into Patch Tuesday at the beginning of every month, although occasionally there'll be some other things throughout the month. What's interesting though is that because my Linux servers don't automatically install, they just notify that you've got something that is worth updating. I'm getting stuff there every single day.

Leo [00:11:18]:
And what's interesting is that I'LL say about half of them again every single day are often security related updates. And I'm just wondering if this is not fallout from the responsible disclosure that was happening with Mythos that these companies and open source communities got early access to the tools so that they could actually deal with some of these vulnerabilities before they actually made it out into the wild. So a little bit of advice for anybody that's running a server on the Internet, keep checking for updates. Take those updates, take them early, take them often. And just because you took one this morning doesn't mean there won't be an important one this afternoon. It's that. And then finally, I just thought it was interesting that Fable, which was the newest model again from Anthropic, that is essentially, I guess it's Mythos related. It's supposedly that, that good, if you will.

Leo [00:12:19]:
It's one of those models that it was out for about five minutes before Anthropic was forced apparently by the US Government to restrict access. For now, of course they're protesting that. They don't think that it meets the criteria that the government had used to insist it be shut down. But, but I just thought it was interesting. The only reason I mention it at all is that again, Mythos class AIs are coming and not all of them are going to be subject to the whims of the US government. If not Fable, if not Mythos itself from Anthropic, then we're going to be seeing these powerful AI agents coming from other corners and they're going to be accessible. So I'm, I was looking forward to playing with Fable, but apparently that's not going to happen for a little while.

Gary [00:13:14]:
Yeah, yeah, I keep getting that message every time I use Claude. Oh, Fable not available. Read why? You know, and I'm like, I already read why I don't want to read it.

Leo [00:13:26]:
So I understand you've been having some fun this morning.

Gary [00:13:29]:
Well, yeah, actually yesterday and today our recordings actually delayed by a tech issue, but nothing to do with us. So my bank was bought by another bank. It happens. And this months ago this happened. Right. And this, the switch over was this weekend. So you know, and I, I had a pretty like, oh yeah, I mean this is going to happen. Probably lucky that it hadn't happened in 20 some years of banking with this bank that had been bought by anybody.

Gary [00:14:01]:
So, you know, and there's going to be some switchover stuff. You know, they'll switch stuff over and hopefully they'll do with their end find. And I just need to go into a couple places and change my routing number and account number to handle payments and transfers and all that. No big deal. Well, it turned out to be much bigger deal than I thought. Like in terms of time I thought. I didn't think it would be any real amount of time that I needed to budget for and I was wrong with that. I probably spent two hours yesterday, I spent at least an hour today and I'm still not done.

Gary [00:14:34]:
I still have more accounts. I need to go. So this is actually adding up to like a significant amount of time on my part. Like at this point I'm kind of mad that they didn't say, hey, we know you're going to spend a lot of time dealing with this on your own. We'll pay you for that. Because it's like I feel like I'm doing somebody else's job, you know, like.

Leo [00:14:55]:
So I have a question.

Gary [00:14:56]:
Yeah.

Leo [00:14:58]:
So my bank got purchased multiple years ago, but the routing number and my account number didn't change.

Gary [00:15:09]:
Yeah, that's not the case.

Leo [00:15:10]:
Here, use my 25 year old bank account number.

Gary [00:15:13]:
Yeah.

Leo [00:15:14]:
And it's just fine. Even though it started out as this bank and it's now that bank, apparently

Gary [00:15:19]:
that's not the case.

Leo [00:15:21]:
Nope.

Gary [00:15:22]:
We all got, I mean, you know, and it was like two months ago, we all got like, here's your new. You know, first of all, the routing number is different because, you know, different for this bank. And then all. Here's your map of your old accounts to your new accounts and the numbers we were giving them well in advance if we wanted to do anything in advance. But I was smart enough not to do anything advanced because I know a lot of the stuff is going to depend on like the account actually being there and active, but you could actually see, you could. I could log into my new bank account. I couldn't do anything with it yet, but I could log in and it was like read only and all that. And the trans.

Gary [00:15:54]:
The switchover happened. It fortunately happened at a good time. Like there weren't too many things that I had to say, oh, put this one thing off. I had one thing where I had to delay a payment because it was going to fall out, like right away. Like, and I was like, all right. And I found, oh, you can say delay it and I'll say, yeah, delay it. So but then I just had to go in and change everything. But what got me wasn't so much I had to do all that work, it was that every single instance was technically different.

Gary [00:16:28]:
So you've got lots of things like credit cards that are just on autopay. You've got people paying you, like, you know, YouTube for instance, and all that. You know that it comes in. You've got other things where you transfer to and from different accounts. Lots of stuff I made a list of and it came out to about 15 different places and every one was different. Some of them were the, the simplest case was, oh, what's the new routing number? What's the new account number? Done. Yeah, you know, that's it. We'll just use these two numbers now, you know, made sure you entered in them correctly, but that's all that you need to do.

Gary [00:17:11]:
Others went in completely different directions. Some didn't let you enter those in. Some had you go and log in. It's like, oh, we have an agreement with your new bank. And, and you click and then like you're, there's going to be this one part of the webpage that's your new bank and the other parts us. And then you do this back and forth handshake thing. And that was like, oh, okay, I can see they were trying to be more convenient there. But it's a lot more steps to do that that just let me enter the new numbers.

Gary [00:17:44]:
There were others. That was one had me scan in my driver's license, you know, with my phone to prove I was me making this change. But I know they didn't do that. When I first set things up, it was only because I was making a change. It's like, okay, that was annoying. Another or two different systems did the thing where they deposited a certain amount in pennies, right? You know, 23 cents or whatever. And I had to wait for that to come through and then go back and finish the process a day later and confirm by saying, here was the amount. So there was that.

Gary [00:18:24]:
There were two companies that did that one. I, I, there was no way to do it. I just had to email my customer the, my payroll company, right, that, you know, both withdraws and deposits into the bank account. There's no way to do it. I have to contact my special. And it's been more than 24 hours and I still not have heard from my specialist. So yeah, there's that another one. The one that derailed me was actually Apple, Apple Card or Apple Card Savings.

Gary [00:18:57]:
I was able to go and say, yeah, pay, use this new account to pay my credit card. And it was like, what's the routing number? What's the bank number? Done. So it was easy. Then I was like, oh, Also I want to set it up so I have Apple Savings I could transfer to and from. And it just was error message. Can't do it. Contact customer support. And so I had to contact customer support.

Gary [00:19:18]:
They had to call my new bank and get the three of us on the line at the same time with confirmations flying back and forth between all of us to confirm everything. And all it was was information, information that I had that I could have entered into the website but they needed to hear it from calling the bank to, to do it, which is totally overkill, especially considering they were already pulling money out of the account to pay the credit card. Right. So why do they need me to confirm to put money in, you know, anyway. And that was, I timed that call because that's the call. It delayed our, our talk. That call. Call came in 27 minutes on hold waiting.

Gary [00:20:05]:
So yeah, that was a pain in the butt. And there's still three accounts that are just things I don't need to do on a regular basis. So they got, you know, I put them on a do later list that I still have to deal with to add to so that even if that goes smoothly, that's still 15 minutes probably to go through the whole system and, and add those bank accounts to it anyway. It's just, I'm just amazed that there is no standard way to do it, you know, that they have that different companies, they all should have been taking the same level of caution. They clearly are not. I don't know if what the appropriate level of caution was, but you know,

Leo [00:20:46]:
I heard that having somebody randomly come in and say, hey, I'm changing my bank account number, trust me, yes, there's risk there.

Gary [00:20:54]:
But I mean I have to log in. I'm logging into each of these accounts and I'm, I'm. Everything's available, I'm logged in stuff's available to me. If you know, the old bank accounts there, I could have transferred money to and from that. In almost all of these cases, there's all sorts of stuff I could have done logged in as me authenticated with two factor authentication or a passkey. I just couldn't do this right with some of them. And then I'm also amazed. It's like, well, I would think that the most basic thing was let, let this new information go in and then notify the account holder, text message and or email that this change has been made and perhaps put a lock on the account saying it's going to, this will be locked for 24 hours for using this.

Gary [00:21:45]:
So you have time if like, oh, I didn't do that, I should call them. Only one company did that. There's one company that I have money transferred to me. Right. So one company that pays me actually came, came in and said, you can't get paid now for five days.

Leo [00:22:05]:
Okay.

Gary [00:22:07]:
Yeah. And I did get a confirmation saying, hey, a new bank was added as the payment out option. Right. Perfect. Five days may be. Be a little long for some of these things if you need to pay bills or do stuff. But the others did not. Like, half of them notified me.

Gary [00:22:23]:
None of them put a hold on anything. But that to me should have been the way to do it instead of all this. Even the phone call stuff was ridiculous. It was like every, like there were at least five things that are kind of red flags about this that I shouldn't be doing except for the fact that I'm the one that placed the call.

Leo [00:22:41]:
Right.

Gary [00:22:41]:
You know, that was like, okay, I'm the one that placed the call, but I still don't feel comfortable providing you with the information you're asking for. You know, it's like I should be able to do it in a better way anyway. It's just, it's just weird. It's.

Leo [00:22:56]:
Yep. Your scenario happens. People do change banks without a purchase. I'm going to go all the way back to the originating bank though, and say, you shouldn't have been in this position at all.

Gary [00:23:08]:
Yeah.

Leo [00:23:09]:
And I'll say that because of what I said earlier, if one bank purchases another.

Gary [00:23:14]:
Yeah.

Leo [00:23:15]:
Presumably the routing number and the account numbers for that old bank are part of the purchase. That's why, like I said when I, when it happened to me, the routing number didn't change. It just whenever you enter it online, the name of the bank came up correctly. It had changed. Right. So they went into wherever routing numbers are assigned and said, hey, this was this bank, this is now this bank and it's ours. And the customers had no changes to make because nothing changed for them. I'm honestly surprised that that didn't happen.

Gary [00:23:51]:
I know. I am too. And I'm kind of, you know, one thing I wish I would have tried or maybe not is maybe just tried not doing any of this. Like maybe just assuming that they did own that routing number and assuming that the account number, if it was attached to that routing number, everything would have just been linked and just not done anything. And then maybe just tried maybe just a hundred dollar transfer from one to another after. But the thing is, I would have been like, okay, but it's it's day one or day two. Like, will this work after a week? Will this work after a month? But so I. You know, I still would have had to think about, like, oh, how do I do this? Right? And I guess that's what they're saying here is maybe.

Gary [00:24:32]:
Maybe they do have a like, this will work for a while kind of thing. But you can't trust it in the long run, so might as well switch everything over. Still, though, it's. It was more of a pain than I thought. I. And I told a bunch of people, oh, it's no, yeah, you have the same package. But, yeah, it's no. There's really nothing to do.

Gary [00:24:50]:
You just got to switch a few numbers. It won't take you very long. And now I find myself a victim of my own optimism.

Leo [00:24:57]:
By the way, do you have paper checks?

Gary [00:25:01]:
So I don't use. I haven't used paper checks in so long.

Leo [00:25:05]:
I understand. Yeah, that's.

Gary [00:25:06]:
But we used one for me, and

Leo [00:25:09]:
my paper checks are so old.

Gary [00:25:10]:
Yeah. Oh, yeah. We used one when we bought. We recently bought a new car, and we had to use a check to write, you know, the amount. And that was, like, the first check we had used in, like, two years. And we were kind of surprised, too. We were like, oh, do we need. How do we get the money? And they could have done a wire transfer, right? So we could.

Gary [00:25:35]:
It did have another option, but what I did do was I went to. Is like, well, we should have checks just in case. I went to a, you know, a tech printing service that I've used before, and for 14 bucks, I'm gonna get one book, which would be, like, 120 checks, which will last us until this bank is bought by some other bank, you know? But I was surprised, though, and some of the stuff. And I'm not done yet. I still have my payroll stuff to do. I was surprised that nobody asked me for, like, oh, I can't avoid a check, you know, and my payroll company, they're the ones that usually do that. And I plan on. Even if my checks arrive in the mail, by the time they get around to helping me, I plan on saying, yeah, I don't use checks.

Gary [00:26:21]:
Method two. What's method two? You know, just pushing. It's the same thing when somebody asked me for a fax, you know? Yeah, okay, you want a fax. Great. What's method two? I know you've got it. Just skip right to method two, because I'm not sending you a fax. So sorry, K. Not sending faxes anymore.

Leo [00:26:41]:
But at least if you had to, you could do it for free.

Gary [00:26:43]:
I could do it for free, but yeah, so anyway, so there's my rant on that and kind of, you know, this was all, everything involved was tech. It was all apps, it was all websites, it was all tech. It was just, there was no standard, no consistency and too much time.

Leo [00:27:03]:
You and I, we're tech literate, right?

Gary [00:27:07]:
Yeah.

Leo [00:27:08]:
Imagine all the customers who struggle with technology to be having to deal with all the different things that you've been having to deal with. I'm sure that the bank's customer service line is probably overwhelmed right now.

Gary [00:27:20]:
Oh yeah, that's probably why my 27 minute hold to do the three way call. But you know, I hear there's still people that you know though they won't use auto pay and they won't use direct deposit. If I didn't have auto pay or direct deposit, the number of things I would add to do would have fallen significantly. Matter of fact, the only thing I could think of is, you know, there's the link between the two savings, you know, to my Apple savings account that's you know, there's no way to do that with paper. But everything else would have either been oh, because I'm an Internet entrepreneur, so I'm getting paid by Patreon and YouTube and all that, or it's because I'm doing auto pay or auto deduction, all this stuff. If I eliminate all of that and if I just been still writing checks, there wouldn't have been much to do. So a lot of people don't have much to do because they're not, they're not, they chose the hard way to begin with and they're just doing it the hard way every month rather than doing it the easy way.

Leo [00:28:22]:
The easy way.

Gary [00:28:23]:
Well yeah, and that's once it's all ironed out, right, I go back to my state of not doing anything. Credit cards are paid automatically, bills are paid automatically, money comes in automatically. I don't have to do anything at all. Banking wise it just all just runs like, you know, a well oiled machine. But it's going to be at least a month before I could be feel sure that everything I'm not going to get an email saying hey, this transaction failed and I have to go back in. And since I pay myself quarterly, the next time I actually get paid, my payroll company is the middle of September and I will actually be off grid for most of September. So I am like going to be hitting them with like Give me confirmations because they're the most likely to fail. I've had, I've had to switch some information with them before.

Gary [00:29:13]:
And because I move to pay, I'm not the employer and the employee.

Leo [00:29:18]:
Right.

Gary [00:29:19]:
They will be like, we got it. We made all the changes. Like, I've changed addresses with them. And I'll be like, you need to change it for both sides. And be like, got it. And then I'll be like, I want to check. Nope, you didn't change it for both sides. And they'll be like, nope.

Gary [00:29:32]:
Yep, we did. We changed your address. I'm like, he didn't. Let me explain again on the employer and the employee. You only changed it for one of those two. And then bunch of back and forth until eventually they change it for both sides. And then, you know, a year later I get like a tax form from them that's like, you're still using the old address here. Why? You know, so I don't have much confidence that that is going to work, but, you know.

Gary [00:29:59]:
Yeah, so something a little more positive kind of in a way. But going back to AI, sure. There's an interesting little news item came out about something called the Human Consent Registry, which sounds like, so cool. It's like I want to go back to like the 80s and say, yeah, in the future there's going to be the Human Consent Registry. Like, what's that? Is that part of like Logan? Is that part of Logan's run or whatever? So what it is is basically it's a. It's a proactive way to handle AI rights for like using somebody's voice or likeness or their writings or their art or anything for AI. So instead of basically waiting for companies to steal your stuff, finding out they stole your stuff, suing them, or joining a class action, lawsuits, and then getting it all sorted out, you can proactively go into this thing called the Human Consent Registry and say your preferences right at the get go. And it's actually one of the co founders is the actor Cate Blanchett, as well as a whole bunch of other people who are media savvy, basically.

Gary [00:31:18]:
The idea being, you know, for actors, voice actors, screen actors, writers, all of that creative people, they could go in, they could register, set things up and specify how they want AI to, to either use or not use their likeness, their art, their writings, all that stuff. And by making a stink out of it, they can basically tell the AI companies, hey, you know about this now, at the very least, check with this when you're making your Models or scraping your content. And it's kind of a good idea. I like proactive tech solutions like this that basically they're not perfect because certainly an AI company could just go and say, yeah, we're just going to ignore that. Like they're not legally bound to pay attention to this. This is not a government entity or anything like that. Although certainly government legislation can hook into it in the future. But if companies are trying to do the right thing, they can go in here and look and say, oh yeah, we'll get our models to follow what's in this one.

Gary [00:32:30]:
And perhaps multiple ones like this may not be the only one that exists. There might be another one or five out there that you know, kind of like those pet chip registries. Right. You know, the pet chip registries. You know, there's like 10 different companies. You know, it's like a privatized thing, but you can go with any company you want, including the open source free one, and have your chip ID register to that. The same kind of thing here, I think. I mean, it may turn out that this one registry gets to be the ultimate registry and everybody uses it, or it may turn out there there's three, which is not a problem because any AI company should then be able to look at a finite number of registries and get the information about, you know, what to do.

Gary [00:33:15]:
And humans like us can go and tell, say in advance what we want, you know, what our wishes are. And so I think it's a great idea. And it's. Even if you're not interested in it, you know, it's interesting to check out their about page, you know, we'll have a link. And if you are even just like a writer of like scholarly stuff, you've written some things that have been published, you're a blogger or whatever, you could still go in and register here and use it. So probably something that both of us will probably be looking at doing.

Leo [00:33:51]:
I'm on the fence. Yeah. My biggest concern is something you mentioned earlier, that the AI companies don't have to pay attention to this at all.

Gary [00:34:02]:
Yeah.

Leo [00:34:03]:
The only incentive to pay attention to this is publicity and potentially positioning when not necessarily if, but when they get sued, you know, the paying it. Nominally paying attention to this I suppose could be seen as well we're trying to do the right thing. Yeah. But it's unclear. So I don't know. The only part, I haven't gone through it yet, the only part that concerns me is the verify your identity part that can.

Gary [00:34:38]:
Sure.

Leo [00:34:38]:
That can be quite the Hoop. And honestly, depending on what it is they're talking about doing, you know, I honestly have no problem having AI large language models train on my content, but I of course would have problems with any company creating content that claims to be me. Right. Which is a different thing. And presumably I think a lot of, I think a lot of people are conflating the two.

Gary [00:35:14]:
Yeah.

Leo [00:35:14]:
What this one has to say, another,

Gary [00:35:16]:
another way that they can kind of enforce this is, you know, it's obviously one of the co founders is Cate Blanchett and she's got plenty of friends that are also famous people. And I assume that there will be business conducted between people like her. The companies they represent, the companies they sit on the boards of the, the creatives that they do business with. And like, they could simply enforce it through contracts. They could say, oh yeah, we'd love to use this new tool from Anthropic or OpenAI for whatever, but as part of the agreement, we want it in there that your company uses the human content, you human consent registry.

Leo [00:36:01]:
Right.

Gary [00:36:02]:
For all your business. So in other words, if you want our big contract or you want to use the image of so and so, you know, as part of this deal we're making, you've also got to sign on to this.

Leo [00:36:14]:
Right.

Gary [00:36:15]:
That could be a really big push for them to be able to, you know, get, get companies to legit, you know, to legitimize them. So, and I bet you they've got a bunch of stuff like that in the works, if not already done. To go back to the negative side, the ranting section, the. So the, the article here I'm going to link to, it isn't really important in and of itself. It's an article. Websites have a new way to spy on visitors, analyzing their hard drive activity. Basically, this is a really weird spy technique. The idea is that you go to a website, the website writes a bunch of stuff to its own little private storage area that you give it, you know, the, the smallest version of that is what we call cookies.

Gary [00:37:10]:
But there's also website databases for, you know, if you're using complex apps like say Google Docs or whatever, it's got a big database that it's storing inside the browser in a little private area. So researchers found out that if the, the block of data that they're writing is big enough and it's got to be really, they're talking about gigabytes.

Leo [00:37:33]:
Gigabyte, yeah.

Gary [00:37:35]:
It can do stuff like perform a bunch of writes, perform a bunch of reads, and see how long in milliseconds, those things take. Then based on that signature, saying it took like 3 milliseconds for this, 7 milliseconds for that. It can predict what other things you've got going on on your drive. So we could go and say, oh, I noticed that there's these little times here when I try to do this series of read and writes that's consistent with the fact that you are a Google Docs user and then basically say, I've learned something about you I shouldn't have. You use Google Docs or you use E Trade or you use, you know, whatever. And it could. And the article extrapolates to basically say it could spy on you outright. Like you can actually read your Google Docs.

Gary [00:38:29]:
It actually get your financial information or whatever, which seems to me to be outrageous that it could actually get a big enough signature to actually make out individual words and characters of text and stuff like that. See, I get the idea that it could see kind of like a. Oh, there's a 93% chance that you're using Canva because of the, because of this, you know, but what you're doing in Canva, come on. That seems like, ridiculous. But if you don't read the article with the skepticism that we've got and the knowledge we've got, the article is basically saying, hey, there's a way for other websites to spy on you and get whatever information they want about you. And this is, you know, continuing problem with tech journalism. They take something that does have a little bit of a basis and basically throw out a scary headline and a scary article and make you feel all unsafe and stuff like that and the actual details. Like a lot of the stuff I've said, stuff I've had to extrapolate from what they provide in the article and what the researchers have provided the journalists.

Gary [00:39:34]:
We don't know the details because obviously don't want. They don't want to tell all the details to give, you know, the bad actors the information that they need to use this. But that's, that's fine. But also it's turning into scary journalism. It does all the time. And I think it's kind of unfair that, you know, they're scaring people without putting the proper, you know, like, text in there, saying like, hey, this is probably not something you need to worry about.

Leo [00:40:04]:
It's. It's. You're right, it's an ongoing problem. It has certainly been one of my pet peeves. But to be fair, we know the reason. The reason.

Gary [00:40:13]:
Yeah. Clicks yeah, right.

Leo [00:40:14]:
It's all about the clicks. A scary headline, a scary article. A scary headline will get more clicks. A scary article will get more shares. Putting the, the qualifiers in there reduces either or both of those. So there's really no incentive for them to be tame about this. Every and every incentive for them to maximize the, the scariness. And yeah, this is just one of several different articles we see almost every day where tech journalism is taking something that, as you say, it's a legit problem, it's a legit thing.

Leo [00:40:52]:
It just doesn't lie to everybody. Right.

Gary [00:40:54]:
And, and, and note that the next step will be for the security companies to use this as a way to scare you too. This is actually called Frost fingerprinting remotely using OPFS based SSD timing. Frost for short. But what you're going to see eventually is instead of your security, you know, if you bought security software instead of it protecting you from 2815 threats, it'll now say protecting you from 2816 threats. And if you watch all the text scroll by as it says it's checked, it's going to say checking for Frost, you know, whatever. And all it, and what it's going to do is probably say, hey, do you have any web data that's bigger than a gig? You know, and if you do have web data that's bigger than a gig, is it on this list like Google Docs? If it's on that list, then it's fine. If it's not on this list, oh, I don't recognize this company. And they're using a bunch, then we're going to say warning.

Gary [00:41:56]:
The thing is the warning will never happen because you just don't run into a company doing that. And then Frost never as actually used by anybody, but they can add it as a number. And another thing that they say they protect you against, which is like a majority stuff. It's like the old, you know, Mac security software protects you from, you know, 10,000 viruses. You know, 9,815 of them are Windows viruses. That it's helping you keep your network safe for the Windows users, you know, but it's like, okay, but really the people were buying the software because they wanted to be protected from 10,000 Mac viruses, which, you know, they didn't get that nuance. So anyway.

Leo [00:42:40]:
Yep, yep. No, I'm with you. Like I said, we've talked about, about headlines and clicks for a long time and this is another current case.

Gary [00:42:50]:
Yeah.

Leo [00:42:52]:
Ah, let's switch to what's cool. I have mentioned before AI So normally we talk about what we're reading or what we're watching. I'm actually going to call out Gemini for something that's cool. And the reason I say that is because in the last week or two, I've done a couple of things with Gemini that I just didn't think to do before, and it really worked well. Now, normally people think of Gemini and other chatbots as being chatbots. You talk to it, you type at it, it types back. You type at it, it types back. Or in my case, I type at it and it makes me a picture.

Leo [00:43:39]:
I did two things this week. One is we had a very nicely staged photograph of nine corgis. And the way it was staged is that we actually had them all on lead. They were all on leashes. And of course, the picture includes all of those leashes. And I actually went through a process in Photoshop of manually using Photoshop's erase tools. Cool. To get rid of all the leashes took a little while, but the result was pretty cool.

Leo [00:44:13]:
And a little while later, I thought to myself, you know, I wonder if AI could do this? So I took the original picture with the leashes. I uploaded it to Gemini and I said, basically, remove the leashes from this photograph. And a it did. It was roughly indistinguishable from my manual efforts and probably a little bit better, to be honest. And then the cute part was that, of course, Gemini likes to give every task it has some kind of a title. Most of the AI tools will do this. You ask it a question and it'll produce an answer. But one of the things it'll do then is in the list on the left, it'll say such and such.

Leo [00:44:58]:
It described this one as Corgis Unleashed.

Gary [00:45:02]:
I love it.

Leo [00:45:03]:
Hilarious. But it's just one of those weird things that I didn't think about doing before. The other one was that we had a document get sent to us. Somebody took a photograph of a document that somebody was going to have to fill out and sign. And when I took a look at it, since, of course it's my job then to print the document, it.

Gary [00:45:26]:
It just.

Leo [00:45:27]:
It was fuzzy. It was the kind of a document where, yeah, you could hand it to somebody, but you'd be embarrassed about it because it didn't really look that good. It was a photograph, after all. And apparently the camera's resolution or the camera's resolution that made it through to me wasn't really up to the task of the, you know, getting you a nice sharp image of an 8.5 by 11 document. So once again I thought to myself, I wonder if AI could clean this up. So again, I gave it to Gemini and I said, hey, can you sharpen this document for me? I forget the exact words, but it was along those lines and the result was amazing. Now remember that this isn't making stuff up. The letters were visible.

Leo [00:46:16]:
Right. It. It could interpret the fuzziness. It's not like those cases that you see on TV all the time where they just repeatedly say enhance until this blurry license plate becomes incredibly readable. The information was there, Gemini cleaned it up and the document that it produced was like very close to the quality of the original. So my takeaway here is not using it for these specific tasks, although of course, by all means, you know, do this kind of thing. The takeaway here really is think about the AI tools you have at hand and some of the perhaps non standard things that you might ask it to do. As the clickbait headlines might suggest, the results may surprise you.

Leo [00:47:08]:
And I certainly was surprised in a very positive way. So it's something that I'm keeping in the back of my mind whenever I run into these kind of problems.

Gary [00:47:17]:
Yeah, I'm wondering if we can almost start adding a, like a segment new ways we found to use AI because it seems like every week I've got something like my thing was also very photo related in that I use Claude coworkers to look at a. I pointed to a folder of photos. I'm trying to create a travel video and I have like, oh, it's like 2, 000 photos from a trip. Yeah, 2,000 photos. How to find which ones I'm going to use for the video. I tried doing it manually with previous videos and I'm not very good at it. Like I leave out good ones. I include ones.

Gary [00:48:02]:
I was like, why don't I include this? But I, I actually asked it to go through and look at all of the photos and give me a spreadsheet that has the link to the photo, the location. And the location was important because I didn't have GPS information for each of them. But I had GPS for my iPhone photos, but not some of my DSLR or mirrorless photos. So I asked it based on the times of my iPhone photos.

Leo [00:48:34]:
Oh, right.

Gary [00:48:35]:
And what you see in the photo.

Leo [00:48:38]:
Right.

Gary [00:48:39]:
Try to guess, make your best guess as a location, which was important because that helped out. Right. Right away there. And then give a description of the photo. Give me a set of keywords so I could search Right. And then rate them and how good a quality photos they are for use in a travel video.

Leo [00:49:01]:
So Claude is assigning a rating.

Gary [00:49:03]:
Yes, 1 to 10 rating. And I asked it to group them and group them based on subject. So if I took a picture, say of a lion, but I took 10 pictures of the same lion in a row, put them in a group, because I'm only going to use one, right? But pay attention to, like when it's the next group or pay attention to if it's a different lion. I don't want all the lions grouped together. I just want that one lion taken in a sequence of photos at the same time. If I took a picture of the same lion the next day, somehow that's a different group. So group them together, which is very important, and create this whole database for me of all the photos I've got. And it's step one, because step two is I really want it to just tell me which photos to use.

Gary [00:49:54]:
You know, just tell me which are the best photos.

Leo [00:49:56]:
Like put them in there, giving you ratings. That's like 90.

Gary [00:49:59]:
Well, yeah, but then also it's got to like figure out like, oh, so I have like 10 photos of the sunset that are all excellent, right? But I only need the one, right? And I have five photos of this giraffe that are all crap. But I need a photo of a giraffe, right? So feeding it a transcript of like an audio of. Audio of me talking or a script and saying, give me a sequence. It should be able to figure out like you need seven photos to go with this segment of what you're talking about here. I'm going to pick these, right? And you know, here's your seven. And then maybe even saying, here's seven, here's seven alternates and. And then just going through and then sorting it. And then because it's called co work, it could actually copy or duplicate those photos to another folder and put them in order, change the name of the photo so they're all in order.

Gary [00:50:52]:
And I could drag and drop them in, telling it that I want these all to be three seconds in the video. So anyway, I'm halfway there. I've got the spreadsheet, I've got some basics. I think the hard part's done. I think I just need to do all the rest after I write a script for this. And it really could be an interesting way to do it. Will I make changes afterwards? Yes, but I also did that, a great deal of that when I did it manually myself with no AI involved. So I'm hoping Either A, it'll be quicker or B, it'll be better, or maybe both.

Gary [00:51:27]:
I mean, I know, I know for a fact the last video I did, last travel video I did, I had it all done. And I would look for a photo to be like, oh, I need a photo to fill the gap. And I'd look and I'd find one. I'd say, oh my God, this is a fantastic phone. Like, why wasn't this part of my original set?

Leo [00:51:45]:
Right?

Gary [00:51:46]:
So you can't do any worse than I did. Anyway, so that's my. That'll be my. Ain't it cool too? Is using Claude Cowork as a, you know, way to. Way to work with photos and perhaps use it, you know, to make videos. So in a way that's not typical, people think, oh, you're using AI to make videos. Yeah, I'm not using AI to. AI is not making any videos.

Gary [00:52:13]:
AI is like acting as my, like assistant to help me go through my actual real human taken photos to use those in the video.

Leo [00:52:22]:
Yeah, very cool. In terms of blatant self promotion, the article I want to point people out for Ask Leo is common cloud objections. Yes, they can see your data. And no, know, you shouldn't pack panic. And of course, you know exactly what this one's all about. It's askleo.com 193269 every time we talk about cloud storage, we get the people coming out of the woodwork saying, oh my gosh, they're stealing your data. They've claimed ownership of your data. They're scanning it for LLMs, they're scanning it for the government, they're doing all this thing.

Leo [00:53:03]:
You know what? No, I mean, yes, they can see your data. They kind of sort of have to in order to be able to provide the very service you're asking them to provide. But no, they're not. There's not somebody behind a computer somewhere looking at what it is you've written in your latest word doc that's just not happening. Anyway, I think I, I cover several of the more common objections to using cloud storage. I'm not necessarily saying that people have to use cloud storage, but I just hate seeing people walk away from something that could really add value to their digital life because of these unfounded concerns. So anyway, objections, yes, they can see your data. No, you shouldn't panic.

Gary [00:53:51]:
Cool. And we'll end with talk about AI. Strangely, I want to point to a video that's actually one of my premium videos at Patreon, so sorry about that. If you can't see it. But I did want to point this out. It's an early look at Siri AI. Siri AI is the new version of Siri that's coming in the new operating systems. So I just devoted a whole video to just playing with Siri AI on my Mac that I'm testing Mac OS GoldenGate with.

Gary [00:54:20]:
And it's really interesting because it has access to all your stuff privately. It's your computer looking at your stuff just like it does anyway. But now there's AI, an AI layer. And you could do things like. The one story I tell in the video is I deal with a company that changes who my representative is on a regular basis, all the time. And I've just given up on having like, recording, changing my contact who, you know, who it is. So every time I need to contact them, I just look through my emails, I try to find it. It's very hard because they send you a personal email.

Gary [00:55:00]:
So it's not like there's one email address I can look for. And if you look for the company email address, there's a flood of stuff. I tried doing that and I was like, I cannot find. I know it's not. I found one from like five months ago. I know it's not that person. I know it's changed since I can't find it. So I used Siri AI to try it.

Gary [00:55:20]:
And I said, who is my current representative for this company? I didn't say look in mail. I didn't say I nothing. It went thought for about five seconds and said, your current representative is this person. Here's their email address, here's their phone number. Here's the email that I found from a month ago where they introduced themselves with a link that I could go to. And then it said, your previous person was this person. Here's the link to the email when they introduced themselves. And before that it was this person as well, you know, this other person.

Gary [00:55:53]:
And I was like, wow, that was my first, like, use of Siri AI to actually solve a problem in a much shorter time. And the video shows other stuff. It can look at messages and calendar and reminders. You can ask a general stuff. What have I got going on today? What did I forget to do? You know, what did I forget to do? And it'll be like, yeah, you've got a reminder here that's unchecked. Something was due yesterday at 9pm and it comes right out with it, you know, and so it's really kind of cool that it could. It could. It could do that.

Gary [00:56:23]:
It could be this assistant. I kind of think that's probably what Microsoft has going on with their copilot,

Leo [00:56:32]:
but I'm not sure it's what they're trying. It's what they're trying is just falling. To me, it seems like copilot's falling way behind.

Gary [00:56:39]:
Oh, well, it'll be interesting to see. I. I won't be the one doing a comparison video, and you probably won't be either, but somebody out there will be doing a comparison video, and I'll be watching it to see how they compare to each other. Even if they compare. Well, you know, the big thing Apple's pushing is the whole privacy thing, right?

Leo [00:56:55]:
It'll happen. You could probably do it.

Gary [00:56:58]:
Yep.

Leo [00:56:59]:
Yeah.

Gary [00:56:59]:
Happening on machine. If it has to go off machine, it's all encrypted, private, in your account. It's not. You know, Apple's, like, guaranteeing this privacy and letting inspectors take a look at their procedures and everything like that. So, yeah, it'll be interesting. Anyway, I'll link to that video for my patrons that want to check it out.

Leo [00:57:20]:
Cool. Well, I think that does us for yet another week. As always, thank you, everyone, for listening. Hopefully we'll get Adam to pronounce our names correctly on the way out. We will see you here again real soon. Thanks for listening. Take care, everyone. Bye.

Leo [00:57:37]:
Bye.

Gary [00:57:38]:
Bye.