Transcript Generated by Easy Cloud AI’s Beluga
All I wanted to do was fix my email. I did not expect to have a minor existential crisis about how much the world is about to change. Normally this is the sort of talk that I’d have given on stage at some conference about the future of the web, but I’m in New Zealand right now and I also have this vague suspicion that those conferences might not be relevant for too much longer.
I was lucky enough to live through the mass adoption of the internet. For anyone younger than about 25 watching this, you won’t remember just how fast and how incredible that transformation was. And I was on the cutting edge of it. I was living it. I was one of the people who understood what was going on. Useful new technologies generally follow a sigmoid curve.
You have a slow takeoff as the technology is invented and the major pain points get sorted out, and then you have this incredible explosion of growth as people find it useful and the technology gets better and better and better and better and loads of competitors enter the market and people find more and more and more and more and more real world practical uses for it.
And there’s this massive race to make newer, better, bigger things that people keep making more stuff with. And then you reach the limit of what’s possible with that technology and the rate of progress flattens out again. Basically, there’s a reason very few people camp outside Apple stores for the new iPhone anymore. I remember Napster from back in 1999 and in hindsight, I think Napster was the first big sign of just how many industries were going to be changed or completely destroyed by the internet.
Not just the music industry, but travel agents, video rental stores, encyclopedias, shopping malls, big box stores, the postal service, journalism, the entire media industry. Napster wasn’t the first sign, but it was the first major one that millions of non technical people were using by choice. The first one to get big mass media attention and legal problems. It was the first big popular disruption, the first warning shot, the first rumble of thunder off in the distance that said the storm was approaching that everything was about to change.
I don’t think many people realised that at the time and even if they did, they couldn’t have predicted the world that would follow, the world that I grew up in, the world that I’ve got really comfortable with, the world of the web, of social media, of smartphones. This video you’re watching now is a very early 2023 video and a very English speaking one.
This is a record of a point in time and space and I might be wrong about this. I don’t know where technology is going to go from here, but I think that world, my world, the one I grew up in, is about to change radically. I think we’re on a new sigmoid curve and I have no idea how far along that curve we are right now and I don’t know if I want to change with it, but I think I’ll have to and all I was trying to do was fix my email.
This is a really specific problem, but it really annoyed me and I have to explain it for the rest of this to make sense, so bear with me. Almost every email system has the idea of folders. You have one copy of an email and once you’re done with it, you file it in a folder like a piece of paper in a filing cabinet because that’s the mid-20th century physical analogy that the inventors went with.
But in 2004, Google launched Gmail, which threw all that out in favour of labels. Email is just in one big pile, you search to find stuff and you add labels to help you sort it. I hate labels because I like the analogy of folders, that’s what I started out with, that’s what I was already using. I didn’t want to change for no reason, but everything else about Gmail was so much better than any of the alternatives at the time that I moved over to it a couple years later.
And I decided, fine, I will just treat labels like folders. I’ll change my workflow as little as possible. Each thread gets one label, that’s like putting it into a folder, which meant that when I backed up my emails, which I do, never trust having only one copy of something, particularly if that copy’s up in the cloud, but when I did that, I could use a regular old-school desktop email program like Microsoft Outlook to make that backup.
You can use programs like that with Gmail accounts, they talk to each other, if they didn’t, stuff like the default mail app on iPhones wouldn’t work and people would complain. But those old-school programs don’t understand labels, so Gmail just lies and says they’re folders. Great, that means my email program can talk to Gmail and request a copy of all those messages one folder at a time and then I know it’s backed up every message.
They’re even stored in actual email folders like I wanted. It took 15 years for me to notice that that wasn’t working properly. I was doing a routine backup last month and I noticed that a thread in that email backup was missing several recent messages and that made me suspicious, the kind of discomfort that you get when you know something is wrong, but you don’t know what it is yet.
So I looked a little bit further and I realized loads of messages were missing my backup, more than a hundred thousand of them over 15 years. They were still there in Gmail on the web, but they hadn’t been lost. It’s just they hadn’t been backed up. And that’s because it turns out in Gmail labels only attach to individual messages and not the full thread.
That is invisible to most people. It’s barely mentioned in the documentation and it’s a really subtle annoying distinction. If you send an email, put a label on that thread, get a reply and then just hit archive. That reply doesn’t get the label and as you go back and forth and add to that thread, none of the other messages there get a label unless you explicitly untag and retag the whole thread, which means more than a hundred thousand unlabeled messages weren’t in my backup.
Is this a really specific, nerdy problem that only I’m annoyed about? Yes, absolutely. I’m one of the very few people in the world that it’s going to affect. Loads of people don’t sort or label their emails at all and most don’t back them up, but I needed to fix this. I needed my backup to work and I did not want to change my damn workflow.
But how on earth do I fix every email thread over 15 years without painstakingly clicking through all of them? I grew up fixing things with code, so obviously that’s where my brain went. Google has a service called Apps Script. It means that anyone can write simple code to easily change and automate things in Gmail, Calendar, Docs, YouTube, whatever.
It’s a brilliant service. I’ve no idea how it ever got made and it will inevitably one day be shut down with very little warning because that’s just what Google does. But Apps Script meant that to fix this problem, I could just spend an hour or so writing a bit of code and spend a while since I’ve done that.
First thing I do when I’m coding something simple like that is I write out what it’s meant to do in plain English, just the flow of logic, so I can then use that as a guide. So I wrote that down and I thought, hang on, isn’t that an AI thing that’s meant to do this now? Yes, this is a video partly about chat GPT, but I promise I’m not going to go into the details there telling someone about your fascinating AI conversation.
It’s like telling someone about your dreams. They don’t care. It just sounds like a hallucinating nonsense. But I’ve seen posts from people being freaked out about just how good chat GPT is and intellectually, yeah, I looked at it and I agreed with them. It seemed surprisingly good. But it is just a text transformer. All it does is guess what the next word in the phrase is.
I knew we’d been seeing some improvements in that lately. I made a video two years ago about the sentences computers can’t understand but humans can. And at the end of that video, I made a joke about how computers with language skills are 10 years away just as they have been for the last 40 years. I was wrong.
Turns out that’s pretty much solved now. But chat GPT is still just a thing that predicts what the next word is going to be. It produces generic milk toast output that’s confidently wrong about a lot of things. It’s going to help spammers and maybe have a little bit of real world use. Right? But I thought, okay, let’s put my text description in, tell it to translate this into this obscure Google Apps script, see what it does.
And it wrote the code for me in a few seconds. And I felt this pit of the stomach existential horror. And at that moment, I couldn’t explain why. But I figured it out later. It’s just predicting the next word. The code it gave back was a bit wrong. But frankly, my first attempt would have been a bit wrong.
So I asked it in plain English to fix a couple of the errors that had spotted and it did. And that version was nearly right. Not quite, though. It used a really strange approach. I tried to work out why it had done that. And then I realized I could just ask. And it told me in English why it had done that.
And it had made the same mistake I did with my backup. Because turns out Google’s documentation is wrong. Google Apps script itself treats labels as attaching to threads, not messages. Google wrote their own programming language wrong. So I told chat GPT about the error. And it fixed the script. It bodged its way around Google’s bad documentation to build the thing I was actually asking for.
And after all the back and forths, I didn’t actually save much time. But it was so much easier. I never had to pour through documentation to find the exact specific magical incantation that I needed. Also, my coding skills are years out of date. That’s one of the reasons I don’t make computer science videos anymore. The last time I learned a new thing in code was probably 2015.
My old skills were good enough. I didn’t want to change my workflow. I’m not sure if chat GPT is a better code than I am, but it felt about equal. It felt like watching while someone wrote code with me occasionally chiming in and going, um, I think you might have made a mistake there. And sometimes I was the one who made the mistake.
At one point I realized I’d written the wrong thing in that first text description. Chat GPT had just been following my wrong instructions. I know it’s just a text model that predicts what word comes next. But it didn’t feel like it. So at first I thought that pit of the stomach existential dread feeling was because my brain had gone, that’s a human or that’s an alien intelligence.
Or because I was starting to think, well, what if my brain is just a transformer system that’s trying to predict what the next word is? I’m sure I remember some old pop science article about how your brain’s a prediction engine and surprise is just a fancy word for being wrong about what comes next. But then the next day I gave Chat GPT a more difficult problem, one that required not just translating English to code, but actually a little bit of a logical leap.
I think just confidently, completely failed in multiple obvious ways. And that meant that I wasn’t worried about it coming from my job. Not yet, anyway. Sure, no one’s asked for that exact problem to be solved before. The one about the labels and messages. Chat GPT wasn’t just copying and pasting from somewhere. There was arguably creativity involved. But if I’m honest, a lot of the programming I used to do was just copying and pasting bits from somewhere.
At some point, every modern programmer searched on Google or Stack Overflow and adapted something that someone else wrote. It’s basically part of the job at this point. I’m deliberately avoiding the argument about creativity, AI art and copyright, by the way. I don’t know how I feel about it. And I don’t know how the courts will resolve it.
If I hired a human artist or musician to create a painting or song in a more famous artist style, that would be weird. But it would be legal, as long as they’re not pretending to be the famous artist, as long as the smaller artist is not trading on their name. That’s fine. The style of most of my videos on this channel is just a slightly faster version of the TV shows I’ve watched growing up.
That’s what this intelligence was trained on. And this video is basically in the style of every other video essayist on YouTube. It’s just I tend to go to the beach a bit more often. If someone set up an AI to generate YouTube videos or even YouTube educational videos as a genre, I think I’d be fine with that.
I don’t think I’ve had a problem with my work being a small part of a massive pool of training data. But if someone started ripping off specifically my name and my work, yeah, I can see why artists are furious about that. The British legal term for it is passing off. I’ve said before, please don’t train models on what I make or frankly on anyone specific without their consent.
But okay, I wasn’t worried about being replaced. And I wasn’t worried that I was actually talking to some human equivalent alien artificial intelligence. So why did I still have that feeling of dread? Artificial intelligence, text transformers and diffusion models, everything that we’re currently seeing seems to be on that sigmoid curve of progress. And I don’t know what point on that curve we’ve got to, if we’re already most of the way up the curve, then cool, programmers and artists have brand new tools.
But they can’t create something at a human level with them. It’s not going to take many jobs, it’s going to make people’s work more efficient. The same way that loads of inventions like Photoshop have done before. If we’re at the middle of that curve, then wow, we are going to get some really impressive new tools very soon.
That still needs some humans to work them, maybe Siri and the Google Assistant are going to become the things that they were always promised to be. But that feeling of dread came from the idea that chat GPT and the new AI art systems might be to my world what Napster was to the late 90s, the Herald. The first big warning that this new technology, the thing that was going to change everything, was starting to actually change everything.
Where huge numbers of people, not just the nerds, were actively using it. And it didn’t matter that Napster got sued out of existence, because by then there was Morpheus and Groxter and Lime Wire and Kazaar. And then there was Spotify, the old business model, that idea that you bought a copy of music had been struck a mortal blow.
And no one noticed it for a while. If you hear echoes of the Napster case in the lawsuits against AI art programs, so do I. I was getting a haircut the other week and unprompted, the barber started talking about chat GPT and how he’d actually used it to write a formal email that he couldn’t be bothered to put together himself.
He wasn’t a particularly technical person, he wasn’t a nerd like me. But he’d used it easily, the same way that someone who wanted to get a load of music for free would figure out Napster. It’s not about chat GPT, not specifically. It’s about what it represents. Because if we’re still at the start of the curve for AI, if we’re at the Napster point, then everything is about to change just as fast and just as strangely as it did in the early 2000s, perhaps beyond all recognition.
And this time I’m on the wrong end of it. Unlike the music executive back in 1999, it feels to me like something might have just gone very wrong for the now comfortable world that I grew up with and that I’ve settled down in. That’s where the dread came from. The worry that suddenly I don’t know what comes next.
No one does. I’ve been complaining for years that it feels like nothing has really changed since smartphones came along. And I think that maybe, maybe, I should have been careful what I wished for. At some point I’ll look back on this video and with hindsight I will easily be able to see where we were on the curve.
We all will. Part of me hopes that I am entirely wrong, but in a few years I’ll still be working like this. The email thing is a metaphor. Of course it won’t be email folders for you. It might be something else, something you’re attached to, maybe something minor, or maybe the whole industry you work in. But right now that feeling, that creeping horror, that dread, it turns out that was the worry that after years of being fairly steady and comfortable, my world is about to change.
And despite everything, I will still want my email to be in folders. And now an AI-generated NordVPN ad. Today I want to talk about one of my favorite tools for staying connected while traveling or just browsing the internet. It’s called NordVPN and it’s been a game changer for me. With NordVPN I can access content from anywhere in the world no matter where I am.
Whether I’m in a country with strict internet censorship or I want to access a website that blocks people from a particular country or just want to catch up on my favorite shows from back home, NordVPN has got me covered. Some other amazing features of NordVPN include the ability to connect up to six devices simultaneously and 24-7 customer support.
Whether I’m using the laptop, tablet or phone, NordVPN makes it easy to stay connected and secure. So if you’re looking for a VPN that can help you access content from around the world and keep you safe and secure online, give NordVPN a try. Trust me, you won’t regret it. To get the best deal they’re currently offering, go to NordVPN.com or click the link in the description.
I’m going to be out of a job soon.