Assistive AI Automation Tools
47:53 Share

Hello and welcome to App Stories. I'm John Voorhees and with me is Federico Vittici. Hey Federico, good to see you. Hello John. Hello, good to see you. How are you? I'm doing really well. I'm doing really well. You know, we are recording this right before Thanksgiving. It'll come out afterwards. Yes. Happy Thanksgiving, John. Happy Thanksgiving to everybody who celebrated a few days ago as they listened to this. But yeah, I don't know. I mean, I'm getting ready. I've got a house full of kids and

was at the airport last night and getting ready to cook tomorrow and yeah it's busy time busy time are you are you cooking the turkey or are you buying the turkey well I will I will admit that for Thanksgiving we decide we like to enjoy the time we spend with our family more than the cooking part so we are I am tomorrow morning Thanksgiving morning I'm driving to a fancy kind of fancy grocery store and getting all getting the whole thing kind of pre pre-made you know a turkey that we just have to heat up and all the sides and all that stuff

Do you know, John, that I recently came across a word in the English language that I had no idea existed and I saw a picture and I had to Google what it was. I'm sure you're familiar with this, but have you ever tried a, and I'm checking again that I'm pronouncing this correctly. Have you ever tried a spatchcocked turkey? A what? A what?

I've heard you said something and then you said turkey. That's all I know. A patch? No, patch caulking is a specific method for butter, also known as butterflying. It basically means that you're cooking a turkey or a chicken. You cut it open first? You remove the backbone, you cut it open, and you flatten it, basically. Well, that's an English word I've never heard either, Fede.

It was the same for me a few days ago until I saw, I believe it was Jason Schreier on Blue Sky. Follow us on Blue Sky, by the way. Jason Schreier on Blue Sky mentioning that a scotch-cocked turkey is the best way to eat a turkey.

on Thanksgiving. I think that's probably true. It probably cooks a lot more. What is this word? Yeah, it probably cooks a lot more evenly, I would think, if you do that. And you don't have to deal with as many of the bones and all that stuff. But yeah, that's a new one to me, too. I'm not much of a chef, I guess. I didn't know that word.

Okay. Well, moving on from whatever that was, you know, I hope everybody is enjoying a nice Thanksgiving and taking advantage of capitalism with Black Friday deals. Yes. I know I did. I wanted to talk about AI. Just, you know, fun, lightweight Thanksgiving topic. Yes, yes. Because this has obviously been on our minds a lot this year and...

Lately, whereby lately, I mean over the past month, even more so than before, I've been thinking about the different kinds of AI that I think very clearly have appeared in 2024. And, you know, this is obviously a very complex conversation, very politically and emotionally loaded for a lot of people.

And I'm going to try and keep an objective, maybe it's not the right word, but like a bird's eye view, just taking a look at things from above and judging them for what they are. Yeah. And I think it's very clear that we are witnessing at a very macro level two types of AI, you know.

Assuming that we're talking about the AI, large language models, all based on pretty much the same technology. But there are two sort of currents of generative AI and assistive AI.

Where by generative, we mean like the AI that can actually generate, produce content, whether it's text, photos, illustrations, music, movies. Right. Versus the AI that is more meant to assist you in your work. And so help you with repetitive tasks, help you with coding tasks. You're still generating content.

an output, but it's more oriented toward helping you rather than replacing the output of another person. - Yeah, the way I think about this, and obviously it's a continuum, and at some point you cross over from one type of AI to another, but I think of it more like tools like Photoshop, for instance. Photoshop, and just even putting aside the fact that Photoshop has generative AI baked into it, but you know,

Photoshop from five years ago, say. You can use it to mask things. You can use it to alter photographs. You can use it to change the colors. You can do all sorts of things to images with Photoshop. But at the end of the day, the changes you're making are assisting you in a creative goal as the person who is editing that photograph. You're still in control of the tools and the ultimate output.

And when you and I, in the summer, wrote our letter about how gender of AI needs to be regulated, at that time, I did an extraordinary amount of reading about gender of AI and all the different technologies. And it was that one article that I know you mentioned on Connected that I'll have to find the link to and put in the show notes that I don't remember the author's name or the title off the top of my head. But it was...

Trying to put words around what the different kinds of AI are and what it is about some aspects of AI that really bother people. And one of the concepts in there was the distinction, as you're making, between generative AI and assistive AI. And that really stuck with me from that moment, even in that moment when I was very upset about what was happening with the open web and AI.

it made sense to me that that was qualitatively different, at least in my mind, in terms of what is happening. And that's kind of where I've, I think it's a useful framework to try to think about some of these things.

Yeah. I'm going to give you another buzzword that I've recently seen. You know how all these AI companies are now making agents? Oh, yes. So like little tools that perform actions for you. Another term that I've seen to refer to this type of AI is agentic. Yes, I have too. I've heard that on places like the Verge cast and other places. Agentic AI because you're making agents. But anyway, so I've been thinking about this a lot. Yeah.

In the sense that, if all, there is a bit of a crisis internally within me. Like in the sense that I think, and I think it's fair to have this internal debate. How can you be comfortable with something that at the very core, at the very foundation, even assistive AI is still based on unethical practices

data scraping practices that have lifted off content from people's websites, from people's portfolios, from social media, from Reddit, from all kinds of places. And the thing is, I'm not comfortable with it. Like when I think about it, when I think about it,

I still think that ultimately the training industry should be regulated and more companies should be open in terms of what they're doing with their data sets when it comes to training. They should honor the robots.txt file on website publishers. All of that still holds true.

But this kind of thinking, like, can you appreciate a part of a product? Can you appreciate a part of an industry when if you do your research, you find out that at the very core, there are some really shady and questionable practices that...

if you think about it, you can apply to almost anything in life. That you can apply to the food industry, you can apply to the music industry, you can apply to the fashion industry. Are you going to still drive a car even though the practices of some big oil companies you don't agree with? There's a lot of things. This is something that I've actually been thinking about this a lot. If you dig deep

deep down enough in any field, in any industry, you're going to find the thing. You're going to find the aspect that makes you uncomfortable with saying, oh, you know, I don't want to drive a car anymore. I'm not going to eat food anymore. I'm not going to, you know, and that's that ultimately, unless you are, I don't know, a billionaire or unless you are the president of the United States, like if you're a regular person,

you cannot possibly embark on dozens of battles in your lifetime every single day. It's just, it's impossible. So you need to, that doesn't mean to use unethical products. It means that you can try your best

to understand what it is that you don't like and what it is that you accept and support companies that at the very least try to walk the walk and try to do the right thing, even though sometimes the completely 100% right thing is impossible. Right. So that being said, I've been thinking about this and I realized, well, what if I treated, you know, let's treat AI as any other type of software, right? Right.

I'm always going to think that generative AI is stupid, produces ugly results. It is no replacement for human art. It's no art. In fact, that's not art. Like even the best mid-journey, it can be even the best mid-journey model. It's no art. That is no art. That it's not art. So there's another kind of AI though that sort of piqued my interest. And that is the type of AI

LLM that can assist you with boring tasks. And the reason why it fascinates me is that I see it, and I'm pretty sure that Apple also sees it this way. I see it as the continuation of automation, of workflow and then shortcuts. And then like this tool, I see it as a software tool to let me create things that help me in my daily life and in my work.

They don't replace me. They don't, you know, I'm not using these products to, you know, make a virtual John blogging for Mac stories. I'm not building a John agent, you know, a fake John that blogs on the site. No, I'm just taking advantage of these tools to help me with coding, with renaming files, with basic research. And then I got to do the work.

But these tools, they can speed up my work. Right. Well, they can free up your time to do the creative work because, boy, you know, we both do plenty of boring, repetitive, administrative tasks. It's just part of running the business. And if we can free up some of that time, that's time we can be doing writing or making a podcast or whatever else it is. Exactly. Now, I'm a creative person and...

If there's one thing I hate is doing the admin work and the repetitive work. I just, if it were up to me, I would just spend all my days from morning to the evening just doing

Being creative, you know, writing, researching, playing around with shortcuts, trying apps. And that is just not the reality, unfortunately, because there's other responsibilities that come with running a business. But you reach a point where you start wondering, well, how many of these responsibilities can I automate? And so this is how I'm looking at this. These tools...

When it comes to the assistive part, they can maybe help me. They can assist me in reducing the time I spend doing the boring work

that people don't know me, people don't know you, John, for our repetitive work. - We do it every day and we do a good job at it, but nobody knows about it. - Right? So nobody knows about it because that's not why people follow you. People follow you because you have a creative output on the internet. And people follow you for your blog posts, for your podcasts, for your posts on social media.

They don't follow you because like, oh, I'm actually following John because he's very good at email and he's very good at renaming files in Finder. No, people don't follow you for that. People follow you for the creative output. So if anything, it's in my audience's interest if I could be more creative, if I could save even more time

you know, spend less time doing those tasks. And I'm going to give you some practical examples in a minute, but this is how I've been approaching it. Yeah. I think that that's a good approach. I do think it's worth mentioning. I think that one counter to what you're saying that I could, I can almost hear people saying as well, you should hire somebody to do those things, to rename your files. And I, I,

I hear it. I, you know, I, I kind of get that on the other hand, it's like when I crop a photo, a screenshot for an article I'm writing in on Mac stories, um,

I don't hire a designer to do that because it's something I can do myself. And it's a smaller task. It's not something that I would ever in a million years even contemplate hiring someone to do. And I think a lot of these tasks that you're talking about fall into the exact same category. These are just things that you and I have done for years. And we have never even given a second thought to hiring someone to do them. And I don't think they rise the level to either the quantity or the type of work

that's worth hiring someone to do. Instead, they're the perfect kind of thing for automation, for finding some sort of tool or shortcut that can allow us to do them more efficiently so we can get back to the stuff that we, you know, both were known for and that we enjoy doing. That is a very good point and I'm glad you raised it. And when I see that kind of opinion, I always think that it's kind of condescending and also unrealistic in the sense that, first of all, I've been creating automations forever. It's not like,

What should I have done over the past 10 years? Instead of building 500 shortcuts that I'm sharing with people, should I have hired 500 people? Instead of building those shortcuts, I've always automated my work. I'm never going to stop automating my work. And also, it's not realistic. Because for a small business, if I started hiring every single person for any trivial task, we would go out of business. That is the simple reality. So...

You know, it's easy to say, oh, you should hire people. I do hire people. When something is really important and needs to be, you know, impactful, you're literally talking to a person who hired an illustrator to illustrate my iOS review. Well, that's the thing is we're...

- We are far more likely to hire creative people to do creative work because that's what makes Mac stories better. It's not having someone rename files faster for us. It's having writers-- - I'm not gonna hire somebody to rename my files. I'm just looking for a way because it's something that I could do manually. I just don't wanna do it manually. And so I think it's helpful for people like me, people like you, people who have been thinking about AI this year.

This thing that is happening around us and, you know, seemingly there's no stopping to it. But I think it's been useful for me to form an opinion and say, I'm never going to like generative AI. I am fine with that. I'm never going to use it. I think it's gross. I think it's not art. That is my opinion. Nothing in that field is probably going to ever change my mind.

But at the same time, I think I can draw a line and say that kind of AI, not for me. This other kind of AI, if I think of it as a tool, if I think of it as an instrument that is also now at my disposal to make me spend less time doing the trivial work, doing the boring work and

have more time to be creative, then yeah, let's give it a try. Yeah, I think it is important to look at it that way when you're looking at the tools to evaluate the tools for what they are because part of the problem I think with AI right now is that we're in the middle of an extreme hype cycle that is unrealistic and selling a lot of these tools for things that they aren't. And we're also dealing with a lot of

Probably, I would say the worst personalities in all of Silicon Valley in terms of some of these companies that we're dealing with. And that just makes it that much more difficult to deal with because it's very hard to kind of disassociate those things which you might find distasteful from the tools themselves. But I know you've been working on a lot of tools and I've been doing... I literally, Federico, after...

the midsummer AI dust up, I guess I'd call it. I just walked away from all AI for a few months. I just needed to clear my head and reevaluate it and think about more about some of these things we're talking about today. But I've slowly been returning ever since kind of the review season and really digging in deeper with the Apple tools, which I wrote about not too long ago with the release of 18.1 and 15.1 on the Mac.

Because, yeah, these tools are here and they're here to stay. It'll be interesting to see where they all go. I'm a little skeptical about some of these, but there are tools out there that are very useful. And I know you have some practical examples of things you've been able to accomplish that really are kind of astounding. I think particularly when you're talking about things that are adjacent to coding, some of these tools are particularly powerful. Yeah, so my preference when it comes to assistive AI that can...

that is designed to help you while you're working. My preference is for Claude. Yeah.

CloudBianthropic is one of the more popular chatbots and large language models in existence. And their latest model, Cloud 3.5 Sonnet, is what I've been using a lot. I've been using a lot for all kinds of assistive-related work things. And I'm going to give you two examples that I think were sort of remarkable in the sense that...

I was able to create two separate tools for a task that I would have done manually. One of them, in fact, you did do manually, but then just out of curiosity, I also wanted to see if it was something that I could automate with AI in the future. So the first one is I had Claude build me a native tool

Swift Playground app on my iPad to take a folder in the Files app full of old Apple Frame assets in 8 bits. This is something that I wrote about on Mac Stories. Save and convert them to 16-bit files

in another folder of the Files app. Now, this is not the AI doing the conversion. This is Claude building, with a series of native APIs on the iPad, a pure Swift and SwiftUI app playground, a little application that uses the Files framework, uses the, what's it called, CG color space? I believe it's a framework for dealing with color and images on Apple platforms. It built me a reusable application

Swift Playground project that I can use to convert 8-bit images to 16-bit images in the Files app. That was something that, like, you did it for me. Like, the first time I texted you, I was like, "Hey, can you help me here?"

We had some back and forth on iMessage. I had to explain what I was looking for. I don't have access to a terminal on my iPad Pro, but you're using a Mac. So we did that back and forth where I sent you the files. You did some research. You downloaded some open source project. And you set up a terminal command and you send me back the files. And even that, when I did it, was assisted by ChatGPT because the tool was...

And I had not used it before. I was familiar with it, but I hadn't used it. I didn't know the syntax. And so I use ChatGPT to come up with the syntax for doing the conversion in the terminal, which was really helpful. But I mean, I wasn't building an app like you did.

- Yeah, and here's the thing. Is this the kind of task where I would have hired somebody? Absolutely not. Because it was the kind of task where if all else failed, I would have just opened Photoshop and on each file, individual files. - Exported them in a different format. - I would have said export as 16 bit. I would have done it manually. Here's the thing, it would have taken me hours, but I would have done it. And this is the definition of automation.

The only difference is that automation used to be the terminal. Then it also became, for some things, shortcuts. Now there's a different type of automation where that back and forth between me and you, I also did the same back and forth with me and Claude with the same natural language. And I just said, I want you to build a Swift Playground app. Can you do it? And we did some back and forth. And what was really impressive is that Claude had me try multiple versions of the Swift code

and some things were not working, and I could just send Claude some screenshots of the errors that Swift Playground was throwing. And he knew how to optimize that. And at the end of the process, I now have this reusable thing that is like a shortcut, but it's not a shortcut. It's a Swift Playground project that allows me to convert those images. Now, there was never going to be a scenario here in which Mac Stories would have hired a person to convert 16-bit images. Right.

There was only a matter of, am I going to waste my time doing this manual conversions? Or do I want to look for an automated solution? So this is example number one. Example number two is always using cloud. I'm working, I'm very excited about this. I'm working on Apple Frames 4. I know that Apple Frames 3.3 just came out, but a couple of nights ago, I had a major breakthrough about

How to considerably slim down Apple Frames and add even more functionalities to it and even more flexibility to it. Because it's something that people have been asking for for a long time. To give you some context, I've been able to remove 500 actions from Apple Frames.

Now, this is based on a new database system. Apple Frames, the original shortcut, is based on a JSON file. It's a little structured dictionary. It's a text file with a bunch of key value pairs that associate the names and the resolutions of Apple devices.

different models and so forth, with instructions for the shortcut on how to frame those screenshots. For Apple Frames 4, I had to change the format of this database. So here's what I did, and this will be more clear once people will be able to see screenshots, when I eventually share the new shortcut.

Manually, it took me about an hour. I went through my entire shortcut. I scrolled the entire shortcut and I created a note in Apple Notes. And I manually had to make a note of like, okay, this resolution belongs to which device? Which coordinates am I using? What's the name of this device? I did it for all the 49 models.

and resolutions supported by Apple Frames. Took me about an hour. Now, this was just a plain text note. It didn't have any specific format. It was not coded in any way. It was just me, line after line, just writing it down. So here's what I did with Cloud. I had to build a new JSON. And I had to basically create a new format with some information from the old one, but formatted in a different way.

So I was not actually merging two different databases. I was merging an old one into a new one that didn't exist yet. So actually, the first step, I used Apple Notes itself. In 18.2, I invoked the chat GPT integration.

And I asked, I created like a long instruction for ChagGPT inside Apple Notes saying, what you're looking at is a list of resolutions. I need you to do this and this. It was like a paragraph long prompt. And ChagGPT understood. And inside Apple Notes, it created a properly formatted JSON file based on that data.

And that was very impressive. But for the next step, I knew that ChaiGPT couldn't do it inside Apple Notes. So I went to ChaiGPT itself in the ChaiGPT app and I asked what I wanted and it couldn't do it. It just, it got stuck on analyzing and it gave me like a series of errors because it couldn't do it. So I was like, well, let's try cloud again.

So what I did was I gave Claude my JSON that I created with chat GPT. And I said, take this JSON and keep it for later. Now I'm going to give you an old file.

And basically I told him to simplify. I told him, you need to understand how to browse the structure of this old file based on the information that I'm giving you here. And you need to create from scratch a new type of database that includes this type of properties. This was all explained in natural language, like with me, almost as if I was explaining it to you. Now, once again,

I could have done it manually. It would have taken me days, probably weeks, and it would have delayed the making of Apple Frames 4 by months. I wouldn't have hired anybody because it was going to be just impossible to explain what I was looking for. I've read your prompts, and I'm not sure I would know what to do. It would have taken me a long time of staring at two sets of data to figure out exactly what it was you were trying to accomplish.

It was something that only I understood because I built it. I don't have the coding skills to do this in Python or JavaScript. I just knew, I knew here's the thing, here's what I'm getting at with this conversation. I knew how to explain it like a person. I didn't know how to explain it in code. I didn't know how to automate this with a shortcut. Even that was beyond my skills.

I knew because I was thinking of, I knew how to explain it in words, in English and nothing else. And Claude did it. It spit out a JSON file with exactly the kind of things that I wanted. Merging an old database file

into the new one. And after that, I was so impressed. I even asked, now I want you to run a series of random checks to make sure that you got this right. And it did get it right. And so as a result, I am almost done with Apple Frames 4. It's a major update that...

I was able to do because this very boring and repetitive task that I could have done manually. - Right. - And it would have, there was no creativity involved here. - No, there wasn't. - It would have involved a lot of cut and paste, cut and paste with the risk

of getting a character wrong, like a comma or a double quote. - Right, or getting everything off by one device so that it's all jumbled up. What I like about this example, Federico, is that the creativity was the breakthrough you had when you realized how you could change the shortcut and eliminate 500 actions. That was the creative part.

The problem with that solution, that creative solution, is that it required a large amount of manual work that it would have taken you weeks to get this done if you were doing it on your own with all the other stuff you're doing. Instead, you've just been kind of doing this a little bit here and there over the past couple of days, and you're already basically finished with at least that part of what you're doing. Yeah. And here's the thing. With these examples in mind...

Does it, this is sort of thoughts that I've been having. Does it make me a hypocrite when I don't like some aspects of AI? We've certainly been called that. I've heard that accusation thrown at me for this exact thing. It's fine and I love having this conversation because these are the sort of thoughts that I have for myself. Does it make me a hypocrite if I criticize AI for some things but I use it for other things? And the thing is,

With anything complex in life, it's impossible to be, and it's in fact counterproductive to be

100% opinionated, unable to change your mind, unable to be flexible about anything. It's like this in politics. It's like this in personal relationships. I'm not saying that you need to compromise your ethics or compromise your morals. I'm saying that seeing everything as either right or wrong

is very rarely the right approach or the intelligent approach. And the thing is... I think it's an epidemic with social media that's been in large clause by social media over the last decade is that everything is binary in terms of it's either yes or no. It's either moral or immoral. There's no nuance to any conversation.

And the thing that I keep going back to with these examples, and I'm sure our listeners maybe have other examples. How can I criticize a software tool that, in the case of Apple Frames, for example, that helped me put together something that when I put it out, when I release it for free to the public,

I know that Apple Frames is used by hundreds, thousands, because I see the download numbers of designers, developers. So it's like I got the help of Claude to create something. Well, not to create something, to speed up the boring, repetitive aspect of a product. Then when it gets released...

It's going to help other creative people be faster. And this is what I think is fascinating and what I think I am comfortable with in my future. The idea of me having the ideas, me being creative. Claude wouldn't have thought of this approach with Apple Frames. And I didn't even...

Here's the thing. Maybe he would have thought of it, but I didn't even bother asking because this is what I like to do. I like to take a shower and think about my shortcuts. You know, this is what I like to drive and think about my shortcut. Like that aspect is mine. Like thinking about solutions is what I appreciate. It's like building Lego. It's what I like. What I don't like is...

Coding, JSON files, renaming things. So I see this path for me in the future where I get to think about the creative solutions and the boring parts that I don't like, I get to offload to...

to a software tool, just like I always have before with Python and AppleScript, and then with workflow, and then with shortcuts. I've always done this. Now there's a new way to do it that lets me do it in natural language and lets me create even more complex automations. This is how I look at it. I think the generative stuff

And even stuff like the summarization, you know, change the tone of this note. All of that is a facade. All of that is so two years ago. And the fact that Apple is now making a big deal

Out of like, oh, you could change the tone of a note and make it more friendly. Like that stuff is so 2022, 2023. Like it really shows how Apple is behind. I think it's much more fascinating, much more fascinating to use these kinds of tools where they're all about productivity rather than, oh, let me make you a better writer. Let me make you a better...

you know, different kind of creative person. The thing I don't like about summary, I've been thinking about summaries a lot lately because I see contexts where maybe there's some utility to them, but by and large, if it's something for as a research tool, I think it's damaging actually, because I think as someone who's researching a topic that you intend to write about or make a video about or whatever it happens to be,

The act of summarizing and taking the notes yourself is what puts the ideas together and allows you to construct something that's creative and useful. And I don't think just taking a bunch of summaries of articles you've found online and reading summaries is really going to get you very far. I mean, it's the old story with...

kids using cliff notes in high school. Back when I was in high school, you would buy these little yellow books that would explain the plot of the novelty. You could read Charles Dickens in 40 pages instead of 400. And yeah, it'll get you some of the basics, but it's not going to get you the real insights. And that, I think, is a problem with the summarization. I think Apple, too...

I'm really skeptical, Federico, about where Apple's going with some of the AI stuff because I do feel like they're behind. And I do feel like some of the... As much as I'm excited about App Intense and having Siri be able to take actions across multiple apps on my behalf, that's kind of an agent model, which I think is... It's intriguing, but I also...

I don't know. I mean, shortcuts, I just don't feel like it has the track record to make that kind of enormous leap. That's a huge leap. I mean, Siri shortcuts are great. App shortcuts are great, but they're always just like single action, one-off things. Putting those together across multiple apps, that's a whole nother story. I'm going to, it'll be, I mean, I'll be really glad to see if they can, they can pull it off, but I, I do remain pretty skeptical about it at this point.

Yeah, that's a good point because something that I appreciate about ChatGPT or Cloud is obviously the fact that there's a large language model with the capability of persistence, right? You can hold a persistent conversation. In fact, you can return to an old chat days after it took place and continue from there. So this idea of siloed conversations

persistent memory is what's really fascinating to me. And obviously, like, sure, Apple is going to work on this quasi-agent model, but without a large language model theory behind it, without a proper, like, what Apple needs to have and what it seems like they will only have, again,

leading credibility to the idea that they are two years behind, what they will likely have in the spring of 2026, 18 months from now, is a shipping version of a large language model Siri, where realistically you're going to have a little Siri app on your phone where you can see all of your conversations and you can continue your chats with Siri. Basically, chat GPT, but Siri. Yep.

Because right now, sure, they can add these integrations with these actions based on shortcuts. But it's always going to... My concern is that it's always going to be very one-off. Like, very self-contained...

Well, that's what the chat... This disposable interaction is what I was trying to say. That's exactly what the chat GPT integration is right now in 18.2, right? Exactly. It's just a replacement. I should actually make a note. Disposable. Yeah, it's just a replacement for...

the part of siri that goes out to the web and does a web search for you when it doesn't have the answer itself because it's not persistent in any way and it can't really do anything with the local file system or the local apps it's just doing what is effectively a replacement for a google search for you yeah and like it doesn't remember anything even if you sign in with your chat gpt account uh

In ChaiGPT, you have the ability to create memories, to store data in memory. You can say things like, keep this in memory or remember this, and it'll keep... It's like a little database that you manually create. But even if you're logged in in the ChaiGPT integration in iOS 18.2, you don't have access to your memory, right? So you can't say to ChaiGPT via Siri, hey, access this clipboard that I have in my memory. You can't do that. So it's all very disposable. And my concern is that

without a large language model, and without the ability for students to remember things, you're going to have, starting with 18.4, I'm guessing, in March, you're going to have this maybe

more flexible interactions where you can do things like, hey, grab my passport from that folder and convert it from a JPEG to a PDF and send it off to John. And it does that for you. Very impressive. But you're basically just building these one-off shortcuts in a disposable fashion. Does that get Apple through to the next 18 months?

just that sort of thing? Is that enough to sort of hold Apple's efforts in AI above water while, and we've got to think about this, what will ChaiGPT and Cloud look like in 18 months? Right. That's a long time in AI land. Very long time. So, but at the same time, you know,

I can work on my iPad and I can keep shortcuts on one side and cloud on the other, chat GPT, use split view like that, and I get the best of both worlds. I get better assistive AI tools and I still get to use the native apps that I like to use. And I think there's a really also fascinating angle here in terms of using shortcuts and sort of combining shortcuts

the classic traditional style automation of shortcuts with AI features. I am very keen to see if Apple will create any Apple intelligence actions.

for shortcuts, I could see a scenario in which maybe I can create something simple like a gradient using Apple intelligence and use it as a variable in an image shortcut, like something like that. I think it's really fascinating. Again, another type of assistive AI, like create an empty canvas for me now that they own Pixelmator and 4G.

Automator, they could make some actions for those apps. So I think it's really fascinating to see that evolution as well. But to sum up, I think we need to be realistic in the sense of if you and I complain about generative AI, we should do it because we believe strongly in that idea. We're not going to change the world. We are not

But we're not going to ever be okay with it either, which I think... And that's fine. And sometimes in life, it's okay to lose a battle and maybe not even actually participate in the battle, but hold your opinion strongly. And for generative AI, that's what I think. At the same time,

And this actually applies, I guess, to our listeners. You know, you work in an office, you know, you work in, you know, maybe you're part of a big company. If you're not going to use these assistive tools to help you in your work, the scary thought that I had a few days ago is that somebody else will. Right. And somebody else will appear to

to be more productive, not a peer, they will be considered more productive and faster than you because they're using a new type of software that can do the sort of thing that I described a few minutes ago. So I think there's a balance here in terms of recognizing some of the values for personal productivity while also thinking about

This industry, most of this industry, when it comes to generative output, is kind of unethical and gross. Right, right. Well, I think it also is important to be smart about how you use these tools because I've already seen lots of stories about people using them in office environments and things as a way to try to replace...

critical thought and create in a way that just makes people look bad. And, you know, you don't want to, you don't want to be writing that important memo to somebody about some topic, have them read it and say, this makes no sense whatsoever. And it's because it was just, you know, created by chat GPT or Claude or something else.

So, I mean, it's great for the kind of repetitive tasks that Federico explained, whether that's in coding or in a more traditional office environment. But I think we all need to be kind of careful about how we use them because they're not a good replacement for actual thinking and thinking.

and reasoning and being creative. So yeah, no, I'm right with you. And always, and always verify, you know, what you're, even when you're using those two, when I was using cloud, for example, with these things, I always verify. Oh yeah. Even though it seems,

plausible enough, I always verify. Because you don't want to use files in production that you didn't... And also, I'm a control freak. So even though the AI is impressive, I'm never going to fully trust it. But yeah, I think this is what I've been at for the past month or so. Like, is there a space in my life for these sort of tools, for this

growing industry of assistive AI, which I'm glad that it seems to be more what these companies are focusing on rather than, hey, we can create photos for you now. We can create illustrations for you. And the fact that Apple is now

advertising that once again it really shows it really shows how they are basically where most of these companies were two years ago yeah um but yeah i think this this flavor makes me more comfortable and ultimately we'll see what happens yeah yeah it

I'm sitting here thinking about the image playgrounds and its inability to generate hot dogs, which I discovered the other day. I was thinking about this. I think I'm going to work on an Apple intelligence story for the site. I don't think I'm going to cover image playgrounds at all. I don't even want to talk about it. It's...

so gross and useless and silly. I'm not going to cover it. I know. I've used it a handful of times just screwing around and I just don't like it. It doesn't make me feel good to use it. And I've never shared them except in very small circles with anybody. But, well, Federico, I think that this...

was a good conversation. I think it's a healthy outlook on, on AI. And I think I, I hope so. I share your views on this stuff. I think you and I are kind of on the same wavelength when it comes to this stuff, but, uh, which is good, which is good. I'm, I'm glad that we are, uh, but we'll see. I mean, there's a lot of testing to do. I mean, I'm, I have been dipping my toe back into all this stuff as well. And I need to spend some more time over the holidays working on things like Claude. Claude's one that I haven't used a lot. So, uh,

seeing what you were able to do with it this past week. I really want to spend some more time with it myself too. All right, everybody, you can find us over at MacStories.net. Federico and I are on Blue Sky now, as so is the entire Mac Stories team. All you have to do is search for any of us there. We even have a starter pack that includes all the people at Mac Stories. You can find that. I'll put a link to it in the show notes.

We are also on Mastodon. Go to vatici.maxstories.net and johnvorhees.maxstories.net to follow us there. And we're on threads and Instagram. Federico is at vatici. That's V-I-T-I-C-C-I. And I'm at John Vorhees. J-O-H-N-V-O-O-R-H-W-E-S. Talk to you next week, Federico. Ciao, John.