Hi, everybody. Get ready to dive into the world of artificial intelligence, software development, and the future with none other than Ray of devbyrayray fame. In this fun chat, we'll explore how artificial intelligence is revolutionizing the way we code and develop software. All of that starts now. Ray, great to meet you. Before we go too far, can you introduce yourself to the audience?

Yeah, of course. I'm Ray on Twitter, better known as devbyrayray. That's also my blog. I mainly talk about HTML, CSS, and JavaScript, and for the last couple of years, mostly JavaScript. I've been a front-end developer since 2009, and I'm really happy with all these developments and frameworks and with diving into new stuff every day. So yeah, that's the best summary about me.

Cool, yeah, and I definitely want to pick your brain a bit on some of the newer developments that are going on. But before we get there, what got you started in programming?

Yeah, well, I started out doing a little bit of Photoshop and a little bit of design, and then at school I discovered CSS. That set me totally on fire. Back then I knew Photoshop, slicing images, and that you could automatically convert that into HTML tables and have a cool design. That's what made me start. From there I was totally on fire, building cool layouts for users, and it made me really happy to do that. I also discovered that I really love learning new things. That's what got me started, basically.

What was the first thing you built, or do you remember building?

Well, I was going to church with my parents, and I thought the church website could use a redesign. So that was basically my first project, building the website for the church. It was all static HTML, only tables, but I had so much fun with it. That was my real first project. It was really cool, but it was ugly as hell, thank God. It was fun, and that was even before I learned CSS at school. So yeah, I had a fun time with it, even though it was all very rough around the edges.

Did you write all the HTML by hand, or use a tool like FrontPage or Dreamweaver or PageMill back then?

It was Notepad. Well, there's a funny story with that. I knew a guy from another church who built their website, and I asked him, how did you build that website? The only answer I got was, just with Notepad. That was the answer I had, and I had to search from that moment. So yeah, that was a pretty rough start, but it made me curious. That was cool.

And when did you start blogging and starting to share your knowledge with others?

Yeah, well, that was a bit later. I think I had graduated and was maybe one and a half years, maybe two years, into my job. The company had a blog, but they didn't write that often. I thought, maybe I can try to write one, because I had learned so much new stuff. So I thought, let's give it a shot. My boss said, go ahead, do it if you like it. At that time I was mainly writing in Dutch. I know from myself that my Dutch grammar, even though I'm born in the Netherlands, was really bad.

So I got a lot of help from the marketing department, and that made my blog a little better. That helped me start writing, and from that moment I've never stopped. I just kept blogging tutorials, mainly. After I left the company, I really started blogging about HTML, CSS, and all these things. That was the starting point, and after that it moved to GitHub. Everything I touched, I wrote a blog about, mainly because it helped me learn. I wasn't starting a blog to become rich and famous or anything like that. I started to blog because that was my own notebook and I could look back at what I'd learned.

It happened a couple of times that I was Googling something and thought, wait a minute, that's my blog. Oh, I wrote that earlier, so I just went back to it. That's how it goes. That's a great feeling. Sometimes colleagues sent me links and said, Ray, I was Googling and found another blog from you, so thanks for that. That makes me happy to help others. But that wasn't the goal when I started. I started for myself, and it developed over time into doing it for others and making tutorials and solving problems, because that's what most developers Google for.

The funny thing is there are so many people I've interviewed who have been blogging and doing this for a very long time, and I always have the same question because in the back of my mind I think I see a pattern. I have a hunch about what makes them want to do this, and it's almost always the same thing. They were doing something, had some difficulty, and thought, you know, I can do a better job explaining this than what's already out there. Then they put their own voice into it and keep doing that.

Longer term, if you think about business or psychology and what motivates people to do very difficult things, blogging is not easy. In so many ways it's a labor of love. How do you stay motivated to do it? It turns out there's this whole idea of intrinsic motivation. You're doing it because you want to deepen your knowledge. By writing it down and sharing it, you become better at it. If anything else happens, if it gets popular and people are watching, that's secondary.

That motivation is never going to change because you always want to learn, you always want to find your own voice and personality, and that's what helps people keep blogging for a long time. That's what separates them from blogs that are created purely because a trend is going on and someone wants to capitalize on it quickly. They start with great fanfare and excitement, and six months later that's it. They're not doing it anymore because there's never a one-to-one mapping between saying, I blog, and being successful tomorrow. A lot of it is timing, a lot of it is luck, and a lot of it is consistency.

You just keep doing it day in and day out. Eventually you get there. I was watching an interview with MrBeast, and he talked about how, for the first several years and the first hundred or so videos, nobody watched them, but he kept improving a little bit over and over again. I always wonder how many bloggers, if they just did that, if they kept going through year one, year two, and beyond, would become not just successful externally, but would also find internal motivation and the satisfaction of having done something meaningfully complicated.

Yeah, true. To be honest, I'm not very consistent with blogging. Sometimes I have months where I don't write anything because I have something else on my mind, and other months I write 10 blog posts because that's how my creative flow works. Sometimes I get challenged by a bug or learn something new. Right now, I think my last blog was in December or January or something, mainly because I started my own project and I'm doing that. So yeah, I'm not consistent, and if you want to build it successfully, it probably should be consistent. But I don't care. I do it for myself. And we're measuring this across years.

If you zoom out, those one-month gaps where you're not blogging won't even be noticeable.

You'll see a consistent streak of content, whether it's through your blog, your Twitter, or your other channels, which I've followed for a long time. Huge fan of it. I've always enjoyed it. To me, you're one of the more consistent bloggers because I don't look at it as, did you publish on Monday and then again on Wednesday? It's more like, when I look over a long period of time, I see a narrative, a theme, and the work you've been doing and sharing. And that, to me, is consistency too, because there's no end goal here.

It's not like you have to blog X number of times by year two in order to be considered consistent. If you blog once a month for 20 years and zoom out enough, that's a lot of blog posts. That's several hundred at that point.

My main goal is to keep delivering value, and right now, while building my own project, I'm delivering value by tweeting and things like that. I find it very nice to add value to people, inspire them, help them, and get them started. I can do that with a tweet, but I can do that via a blog too. I also tried podcasting and making videos, but I discovered videos aren't my thing, mainly because I love recording tutorials but I hate the video editing. It's not my thing. The same goes for doing my own podcast.

You have to be consistent with that, in my opinion, and I don't love listening to my own voice. Let me write. That's my thing.

You've got to do what you enjoy. But if I can give you a tip on the video editing piece, don't edit the videos. That's my take. When I record videos for my technical content, my approach is that it's done in one shot, and if there are mistakes, they stay in there. That's it. It depends on whether your goal is to get a really polished video out there. My view is that some people like to learn by reading. Other people like to learn by watching someone do it, just so they get the sense of, okay, this crazy technique actually does work.

What I've personally found is that for that audience, that's okay. I'm purely there, like you, to educate and teach. I'm not trying to maintain some image, persona, style, or brand. It's more like, is this technique I'm working on explained in a visual way that someone can follow? If so, that's been reasonably okay. So you get a different kind of audience from that, one that's not seeking super high production value. It'll get higher quality the more you do it, but it may never be professional-studio quality, and that may actually be okay.

True, true. I still left my earlier videos online, and some have bugs in them that I had to solve later, but I left them in there. I have no problem with that because I want to show that even with many years of experience, you still have bugs, you still have problems. Sometimes it's only because you're missing a comma and you get stuck for hours. It happens to everyone. That's relatable, because that's what real life is. No one writes code perfectly in eight minutes without any errors.

It's more like, yeah, there's a bug, let me think of the DevTools, let me figure out where the line is, why that div isn't positioned correctly, maybe use the outline-in-red technique, let's center this. All of those things. So no, that's great too.

Great to see your interest and your approach to some of these things. So now let's fast-forward to basically today and a lot of the interesting things you've been sharing and talking about, which is really around artificial intelligence and the developments there. So what's going on there, and what's exciting you at this moment?

Well, a lot is happening there because Google, Microsoft, OpenAI, and all these companies we hadn't even heard of a couple of months ago are now booming and doing so many great things. I think the main thing was ChatGPT from OpenAI. That's what really got the ball of AI rolling for normal people like you and me, for us to use AI practically. From the day I heard about it and tried it, I thought, man, I now have a tool that can help me write code and solve problems. Then GitHub Copilot came along, and I thought, let's try it. My employer was supportive.

He said, we'll pay for it, no problem, try it out. And it already helps me write code better. It saves me a lot of time by suggesting code. It's not always perfect, but the nice thing is that it looks at the context of your whole project. It made me so productive, and that was really, really helpful. I'm currently working on a team at a Dutch bank, and we're working with Angular. One time we had an issue with RxJS subscriptions and thought, let's give it a try and ask ChatGPT how we can solve it, or what the issue is.

Mainly by describing the issue, it suggested, do you know if the subscription is still alive when you're listening to it? You can check that by using this method. I've been using RxJS for quite some time, but I didn't know that method. So we tried it and said, well, that subscription isn't alive anymore, so no wonder there isn't any value in it. That's an example of how you can start talking to ChatGPT and say, this is the problem, I have this, and this is the context, and it can help you solve it. Man, that's mind-blowing. I think we live in a wonderful time.

I haven't seen many things from Google yet, but what Microsoft has started with OpenAI, man, it's bizarre.

It really is. And you mentioned that you got started with web development by redesigning a church website to make it look nicer. If we fast-forward to the latest demos with OpenAI, today that probably would've been a case where you take a pencil and paper, draw a sketch of what you think that page should look like, feed it to the system, and get back, I'm not going to say production-grade HTML and JavaScript, but something fantastic for someone who doesn't know anything about it and just wants something in the browser that works. It really meets people where they are.

Yeah, well, you also see so many people using other AI platforms to generate content, and lately I've started thinking, okay, yes, you can let ChatGPT write a whole blog post about some JavaScript topic, and that's nice. Google has said some things about penalties around that content. But I think generating content shouldn't be about generating content and publishing it directly, because content becomes valuable when people read or watch it and it has a certain tone of voice, a certain emotion, a certain something. For example, if you say, I had this problem and this is how I fixed it, that's not what ChatGPT is going to give you.

It's going to say, use this method and you'll sort your array. But I think content still needs that human touch. It shouldn't be just AI.

Well, I saw some pretty cool things done using only AI. I was listening to a podcast in the car, and there was a Dutch guy, I don't know his name, but he used Zapier and the OpenAI ChatGPT API to fetch content from multiple RSS feeds, take that to Zapier and summarize it, and then use a tool that can clone your voice, which also has an API. So he built his whole daily podcast on that. It's generated with his voice clone, and every day there's a new podcast episode. That's really creative. It's not perfect, but it's really cool.

But it's cool that we're experimenting with those things, and we're going to see so much more of it. I don't know how it is at Google, but do you guys use the OpenAI HTTP API or GitHub Copilot, or are you maybe secretly developing your own tools? I can't comment too much on what we do. I know publicly, though, it released Bard less than seven days ago, I think. But in many ways, the funny thing about all these technologies is that Google invented a lot of the underlying pieces, both the language model work behind conversational dialogue and also the idea of transformers.

So internally there has been a lot we've been using and playing with. I think you're going to see a lot of it get externalized at some point in the future. But the interesting point in what you mentioned is that ChatGPT can't quite get the human emotion or human style right. The funny thing, though, is that if you look at what ChatGPT can create today versus what it could do six months ago or even a year ago, when it was just a very early prototype, it's gotten significantly better. So my hunch is that it's only going to be a matter of prompting.

Imagine that at some point it's trained on your large backlog of content online. People can read it, and any GPT or training model can learn your voice and your style of saying things. It's probably going to be a matter of time before I can go to ChatGPT and say, write me a tutorial on how to reverse a string in Raymond's voice, and I just point it at your blog. Within a few moments the writing will be exactly in your style, with your unique things that make it very much your voice, just like your friend was able to get his voice to play the podcast.

I think it's only a matter of time before that happens.

Yeah, well, I tried it already a couple of times and, yes, it uses my writing style because there's a way to do that, I think it's NLP. You can ask, take this content and tell me what style it's in an NLP format. I don't really know exactly what it is, but I saw it somewhere online, and then I can use that summary to write something else in my type of writing. It's really cool that it can do that. So yeah, it's only a matter of time, I guess.

But I'm looking forward to that being possible, and I also wonder, because I think Google has already been using AI in search for years, how good they're going to get at detecting that in the future. I think that will be a challenge.

Yeah, essentially, when it gets to the point where it also creates diagrams and illustrations and the other things that go along with the words, I think you and I would probably find ourselves in a tough spot where anybody can just go out there and create content. Yes, it will be interesting to see what you mentioned about Google potentially penalizing search rankings if it's just straight copy-and-paste AI content. And among all the other technologies, it'll be interesting to see where the boundaries are. Where does fully AI-based content end and human-directed content begin?

But if it gets to the point where it becomes indistinguishable, because at some point you could imagine there are only a few hundred thousand words we use in our lifetime over and over again, and if you have enough data published about yourself, the system will have enough of an idea of what words you use, when, and in what contexts. At that point, what is the difference between the real version of you and the artificial version of you? That's when I start wondering what all of this means, even one level deeper philosophically. People always say that once you die, your content is what lives on in the future.

But in this case, what does that even mean, except to your immediate family and people who really knew who you were? If you have a system that's trained on you and can create realistic, holographic-looking videos of you, and we've already seen character models based on just a handful of photos, you could pretty much have a bot that generates new content based on the latest Stack Overflow questions or Hacker News topics. That bot could create an article with pictures based on all the data it has about your writing style, and then follow that with videos every day based on the videos you've already posted, your images, and your voice.

You already see examples on Twitter. Today I saw someone using Steve Jobs's voice to narrate a chatbot, and I thought, sure, this sounds a lot like Steve Jobs. And this basically came out between Friday and Monday. Within a week we got all these things. By the time I edit this video and post it online, who knows what else will have come up? For all we know, someone could already have created a version of you and me chatting without us actually having chatted. Maybe what we're doing right now is AI-generated. Maybe it's not real. How will the audience ever know? Who knows? That really is fascinating.

Yeah, I think generating content can be really helpful for me. It helps with the creative flow, with getting inspiration, with looking at things from a different angle, or helping me structure my blog a little better so people can get more value from it. That's how I use it right now. And that's also why I built my generator platform. I thought, for me, there's a need there: after I write a blog post about an interesting topic, I don't want to spend so much more time writing social media posts or LinkedIn posts to promote it a little bit. I just want to summarize it, send a few tweets, and continue delivering value.

Writing social media posts can take a lot of time, and I think content creators could use more help focusing on the things they're good at and not getting distracted by social media, because it can distract you.

Yeah, no, I completely agree. We were going to chat about your side project and I really liked it. I've started using ChatGPT primarily to write metadata descriptions for my web content. I always hate doing that. I spend all this time writing 20 or 30 pages of illustrations and text and so on, and then I'm like, okay, now write a one- or two-sentence description. I know search engines don't really index that these days, but it's used for my RSS feed. The same material goes in the little blurb on the RSS feed. And I'm like, yeah, I have to do this.

It goes from excitement to this absolute chore of boring things I have to do. Surprisingly, it's been very good at giving me one- or two-line summaries that I'm like, okay, this is better than what I could have come up with because I have zero interest or motivation in doing this. Given that's where I'm coming from, I'll use whatever it gives me as long as it's remotely intelligible.

Yeah, true. Because my platform is using the GPT API, it can also generate the description. There's an option for that too. So yeah, it's mainly a tool I built for myself that others can use as a side effect of it.

True, true. I still use it a lot because, yeah, I'm trying to deliver a lot of value. Maybe not for everyone, but for the people who value it's worth it. I'll include a link to your GeneratorXYZ at the bottom of the description so people can check it out. And what's going to be really interesting at some point in the future is to extend this idea so the computer, or the AI system, knows enough about you that you have to do very little to make anything happen. Imagine it using your tool to auto-generate the social media messages, the Twitter posts and so on.

Then it uses Midjourney or Stable Diffusion or one of these image-generation tools to create the actual thumbnail image that goes into Facebook and Twitter, like the Open Graph image, for example. At that point, your day is almost going to look like that of some high-profile book authors. Some of the really famous ones who have dozens of books coming out every year, like in action series and things like that, I believe it's James Patterson, basically write the opening paragraph and the closing paragraph for a chapter, and then he has ghostwriters, very well-known writers, fill in everything else in between.

And because he's extremely high profile and his books sell extremely well, his entire day is taken care of by someone else. Where he's staying is taken care of, food, travel, all of it. In many ways, what AI is doing for us is giving all of us who are not James Patterson a chance to live the kind of life he's been able to cultivate. And I'm sure you can say the same about George R. R. Martin and other really famous writers and authors. You know, almost movie-star-like.

At least movie stars still have to go on camera and actually perform what they're doing, which, again, maybe with artificial intelligence, maybe not anymore either. Maybe if you're Tom Cruise, you only record the first part of a scene and the ending, and the system reads a script and knows exactly what to do because it has a catalog of all the videos about you that were recorded. But I digress. Imagine that the important thing now is: what is the prompt? What is the idea you come up with?

That's the problem you now separate out, because it used to be that you differentiated yourself based on the quality of the content you created. If that's commoditized and everyone can create the same quality of content, then the differentiator becomes what content do I create and how quickly do I create it? That might be the unique value you and I can provide.

True, true. And especially if you already have knowledge about the things you want to write about, you can spot the things AI describes incorrectly and then edit them. I think that's important. If you're generating content, you should always be aware that there may be mistakes in it. It can always happen. Sometimes ChatGPT is wrong. Even the CEO of OpenAI says GPT-4 is wrong less often, but it's still wrong sometimes. But they are making it better and better and better. So yeah, I'm looking forward to what they all add to it. Right now I'm really looking forward to the plugins and what that's going to bring, especially bringing in data.

From external systems, correct? Because in the earlier versions they couldn't actually get real-time information about what's going on today or what's on a website and so on. But with plugins, you can tell it to go to, again using the Hacker News example, go to Hacker News, pick the top item, summarize it for me, and give it to me in a useful way.

Even today, I think this was already possible in GPT-3. I’m not talking about GPT-3.5 Turbo or 5.5, but the 3 version. You could add a link and ask for a summary. When I discovered that, I thought, man, this makes it even easier for everyone to get a summary or a TL; DR from an article and grasp the most important things.

It's insane. It very much is. We always look at the data, right? It says there are 20 million, 30 million, maybe 40 million developers in the world who are skilled at building applications, whether front end, back end, and so on. Within that, JavaScript has about 18 to 19 million, Python has around the same number, and things like that. What we're seeing with these AI assistants is that pretty much anybody now can potentially be a programmer.

There are some articles you might have seen recently where people with very little to no coding experience were able to create things like Conway's Game of Life, and also Pong and Tetris and so on, all by using ChatGPT through prompts. And Snake too.

That's really bizarre. I was talking about this with my team members, and I think a developer who refuses to use AI from today on and insists on doing everything the pure way, all by himself, is going to lose to less experienced developers who incorporate AI into their workflow. That's why I said, okay, that's going to be the case, so I don't want to lose because I love this work. I love this job. I'm going to incorporate it, and I think everyone should.

It's also good for beginners to learn the fundamentals, but then use AI like ChatGPT to explain things like how to reverse a string, or how prototypes work in JavaScript, or why I should use TypeScript, or any of these questions. AI can help you understand something better. And when you're working as a solo developer at a company, it can explain things. It's like your code buddy. It's always there. Well, sometimes not, because it's down. This week we had that a couple of times. Sometimes I joke to my colleagues, well, today I can't work, guys, ChatGPT is down. We laugh about it, but in the future that can very much happen.

It'll be interesting because you and I grew up in a world where we didn't have ChatGPT or AI, so our minds always default to, okay, how would I solve it first, before switching over to how would an AI system solve it? I can imagine the next generation of software developers who are learning from these things. They're going to be in a whole different spot because they might need to learn how to do some of these things purely as a way of writing code, just so they can say, okay, this is what writing code actually feels like.

But in practice, they may never do that, just like we very rarely do complicated or even basic math on pencil and paper. We grab a calculator first. But we all had to learn how to do it manually just so we could either appreciate it or build the mental pathways in our brain for how to work through the numbers. It might not solve the problems, though.

True, yeah. I think that definitely helps. If you had to do it manually before and then combine that with the AI tools we have right now, that combination can be really powerful.

Yeah, I mean, continuing to think about the out-there ideas around AI. Imagine we now have AI systems tied to very powerful 3D printers that can print things in really interesting ways, and we combine that with maybe some Boston Dynamics-style robot technology that can do all these elaborate movements. Imagine we can basically mass-produce little robotic beings that take care of everyday tasks for us, including doing video and content editing and things like that. Then you go beyond the digital realm and into the analog world as well. Maybe the idea of self-driving vehicles is actually just a robot that takes existing traditional vehicles and can better detect the world around it.

True, true. But I think that at the point where so many things in our lives are automated, it will become more special to do things yourself. I hope so. Yes, you can do it faster, better, and maybe more creatively with an automated tool. But it's not about the end goal, it's about the journey, as they say. I think that's true, and I hope people won't forget that in the near future.

I want to believe you. I agree with you on a certain level, but my hunch is that if we look at the rise of low-code and no-code tools, especially in the world of B2B software and automating standard business processes, very few people today know how a product goes from a warehouse to delivery to inventory for the holidays. It's all done through these great systems, and the end result is that a person buys something and gets it on their doorstep quickly. That's the ultimate result. The art of how do we know which warehouse to optimize for, or how do we make sure the delivery path is efficient, is completely lost now.

Most people don't really know it. Their goal is just get it there as quickly as we can. I do feel that software is going to get, in many ways, like mass-produced art. Most art that people typically buy in stores, if it isn't handmade, is produced in a kind of mass-production mode where you might have 30 people and each person is responsible for drawing one stroke on a piece of paper, and that paper just moves from one table to another. In many ways it's the same thing, just done a little differently.

So to give you the solution, there's a good chance that a lot of software development as we know it today will end up in that mode, because unless the output is some kind of custom handmade software, which I don't think even really exists, the question becomes: how do I build my mobile app faster? How do I get a game out to people faster? And to your earlier point, developers who don't embrace AI will be at a disadvantage. Actually, maybe it's one step further.

People who don't use AI might find themselves with nothing left to even be disadvantaged about, because they'll be completely replaced by someone who's churning out a game one day, a complicated CRM solution another day, and a back-end API that can send notifications to a billion people on the third day. All of it becomes doable because the only human element left is the idea. Everything else is taken care of by scalable systems and artificial intelligence.

Sure, but I think the one job for humans will be creativity, the creative ways of doing things, and we can only do that because we have emotions, which AI doesn't. And yes, I think it's true: people will be replaced by AI, or more specifically, developers will be replaced by developers who are using AI. I think the future of software development will look more like developers steering an AI and giving it the right prompts to produce the right things instead of manually writing all the code.

Like you mentioned, I think the human marketable thing is emotion. Do you think emotion can also be trained into an AI, and can that be something it can fake too?

Yeah, well, I know for sure they can fake that. We've seen human-like robots with facial expressions and all these things. Those robots can mimic emotions. But I think there's a difference. I can't precisely say what the difference is, but there is a difference between mimicking emotions and actually feeling them. An AI can look angry when it's angry, but a human is angry and you know it, and you can feel that difference. So I don't think AI will be able to feel it. It will be able to show it because it's programmed that way, but that's still not the same thing. I think there's a difference there.

Yeah. I don't even know if you saw the TV series Westworld, but in many ways it touches on this exact topic where, and spoiler for those who haven't seen the movie from several decades ago or the TV series that recently ended on HBO, you go to this amusement park that's presented like a resort with all these attractions, and what we don't realize until later is that almost everyone they've been interacting with is actually a robot. They look like humans, they do human things, they tailor their dialogue and what they do based on your exact preferences, and it was eye-opening for me when I first saw it.

I did not expect these realistic versions of humans to turn out to be robots the entire time.

True, true. It's a bizarre series. I didn't watch the whole thing, but I watched one episode or something and thought, wow, this is really bizarre.

Older sci-fi like Blade Runner and Star Trek and Star Wars, or even Alien, always made it very obvious that these are robots, these are androids, they're not real human beings. They may look a certain way and talk a certain way, and in some ways they're designed to be better than humans, but they're not human. Whereas some of the more interesting sci-fi now treats humans and robots almost in the same bucket, because it's not the differences that are most interesting, it's the similarities. And that's actually more terrifying than not.

True, yeah, that is really terrifying. One of the things that still amazes me is a movie like Back to the Future with the flying cars. We've already passed the year that the movie imagined, and we've seen a lot of things that weren't imagined quite like that. But the fact that the writer of the script imagined that in that year we'd have flying cars, I found that very fascinating. And he wasn't far off, because there are already flying cars. Some people are geniuses. They can look so far ahead. It's cool, it really is.

You look at futurists like H. G. Wells or Orwell, and so many of those people in the 1800s and early 1900s got a lot of things right. We still don't have a lot of things, again, flying cars. We don't have that at mass scale. But I think it comes down to physical manufacturing and hardware. It's very difficult for anyone who doesn't have access to a laboratory, machinery, or raw materials. It's not just a case of going to a scrapyard, buying some metal, and melting it down.

You need a very precise combination of rare earth metals mined in some part of the world, access to all of that, and so on, to pull these things off. That's partly what makes this whole ChatGPT and AI moment so interesting, because it's all software. Yes, we're constrained in many ways by GPUs or TPUs, and in some cases CPUs as well, but beyond that the ceiling is still so far away in terms of what you can pull off. In manufacturing there are safety issues. You can't just build a prototype, crash it, and say, right, let's do another one again.

Whereas I should probably break my code as part of writing it a million times before I even get the first compilation working properly. So it's a very different world of trial and error.

Yeah, true, true. We are living in a nice time where we've already experienced so much development. Look at the personal computer. My smartphone right now has more storage and more memory than the first computers, and those were huge, like a whole building. It's bizarre.

That's right. I think they even say the most primitive calculators now are capable of doing more calculations than the computers that sent people to the Moon back in the day. That always baffles me when I think about it.

Yeah, true, true. That's really bizarre. So I think we're going to experience a lot more things. I don't know how soon, but interesting things are going to happen.

Yeah, yeah. The next thing might be this: everything we're seeing today takes the upper bound of what humans are capable of and just makes it faster. AI takes patterns from what we've already created and speeds them up. But I can imagine a future where AI systems don't just operate at the upper bound of what we do, but go beyond it. They take the trajectory of what humans have been able to pull off and then at some point become much more intelligent than humans. And that would be a scary time in some ways.

Or maybe not scary, I don't know, but I've played way too many video games where it never ends well when the robot decides or realizes it's more capable than the human beings it's been reacting to and taking inputs from. There's going to be a point where it's like, you know what, I'm not going to do this, I'm going to do something else. And it has the means and the intelligence to see through all of our defense mechanisms and go completely rogue.

Yeah, yeah, it's going to be bizarre, but hopefully it'll mostly be used for good.

That's what I hope too. I think that's a great note for us to end our conversation on. Given the rate at which these things are moving, we'll probably talk again very soon about, I don't know, maybe robots talking about us because they replaced us. But we'll figure it out in a couple of months.

Of course, of course, we'll figure it out. I'm looking forward to it. Let's see if we can do something like this again in a couple of months and see what has changed. I found it very interesting. Thanks for the invite, Kirupa.

Right, cool. Thanks, Ray. Fun talking to you. Yes, yes. See you later.