It’s that time of year again. We’re heading towards the end with a new beginning around the corner. Something we’ll probably remember about December 2022 is all the interesting tech that’s catching on. On today’s podcast, Ben Popper and Matt Kiernander take a brain-break from their work lives at Stack Overflow to reflect on it all…and whether or not we 👍 it.
Episode notes:
Ben asks Matt to explain Mastodon to him like he’s five. Matt says the experience feels a lot like…LinkedIn?
Matt explains that he took social media apps off his phone for a while…just to chill out. (Ed. note, they're already back on.)
We cover the latest AI to emerge that can write essays, jokes, and yes, some code.
As promised, a link to Disney's amazing new paper on an system that can automatically age (or de-age) actors.
While everyone’s confused about the state of social media and AI chat, physicists have created a wormhole using a quantum computer. (Though it may have been a publicity stunt.)
Shout out to Lifeboat Badge winner ralf htp for their answer to the question ‘how to listen for and react to Ace Editor change events.’ Your answer has helped more than 20,000+ people, so rock on.
TRANSCRIPT
[intro music plays]
Ben Popper How many microservices does it take to confuse an engineer? Not as many as you'd think. Cut through complexity in your applications using Splunk Observability Cloud. Learn more and get a free trial at splk.it/sopodcast. That's splk.it/sopodcast. Make sure you use that link and you'll let them know the podcast sent you.
BP Hello, everybody. Welcome back to the Stack Overflow Podcast, a place to talk all things software and technology. I am Ben Popper, Director of Content here at Stack Overflow, joined as I often am by my colleague and collaborator, Matt Kiernander. What's going on, Matt?
Matt Kiernander Hello. Nice to be here.
BP We are getting to the end of the year. We're going to try to take you out with a few new things that are happening in the world of technology, a few things that are interesting about the world of software tools, tips, tricks. Matt, I wanted to touch on the obvious. All the chatter on Twitter from every sector available, all the chatter on social media is people sharing wonderful examples of the new GPT Chat showing off. It can do a lot of fun things. It can write an episode of Seinfeld about Jerry getting ready to figure out the bubble sort algorithm. That one really made me laugh. Some people are a little cynical about whether or not this is art or comedy or just AI extrusion, but what have you seen? What are some of your favorite things that have come across the way? And then we'll talk a little bit about how Stack Overflow is dealing a bit with some of what this bot can do.
MK To be honest I actually haven't seen anything. I haven't been on Twitter all weekend funny enough
BP Good for you.
MK I know. I feel very zen and calm. So unfortunately I can't be able to chime in with any of the ChatGPT. For those of you who don't know, including myself, what is ChatGPT and how can people use it?
BP Sure, so folks at OpenAI who were behind the system that powers Copilot on GitHub, OpenAI works closely with Microsoft. They've released a version, which is sort of the next turn of the wheel for GPT-3, call it GPT-3.5 or whatever, that's much more conversational. So you say, “I would like you to write me an essay on the history of early Dutch settlement in America.” It can do that. “I'd like you to write that essay but with more detail.” And it does that. “I'd like you to write that essay but in the style of the comedian George Carlin.” It does that. And then also because it's got the same stuff behind it as the Copilot, you can say, “I'd like to write an app in Python that does x,” and it gives you back some code that does that. So people have been showing it off and it's making its best guess as to what you want, what it thinks the next set of characters in the sequence should be based on ingesting all the information on the internet and finding patterns. So the one thing it doesn't do is give you a confidence interval. So a lot of folks have been saying, “Well, this is pretty interesting,” and throwing in some Stack Overflow questions. “What do I do here?” It gives back an answer. It looks pretty good, but if you actually run it all the way through, it doesn't work. So our mods have briefly turned it off –not permanently– just to get a sort of handle on this because it's overwhelming the speed with which it can generate what feels like a good idea, a good attempt, a good answer to x.
MK I mean, the biggest argument I've seen against a lot of this AI generation stuff is that when it comes to nuance in context, the AIs just don't have that capability yet. It's one thing to be like, “Give me an essay in the style of George Carlin,” because it can go through all of George Carlin's scripts and generate something that seemingly is a George Carlin script, but when you start looking at the details, that's where things get a little bit trickier.
BP Yeah. I had asked it to write an essay and then I said to include footnotes, and it did include footnotes. Or I said, “Include links to your sources,” or whatever, and it included those and I was like, “Wow, amazing.” But all the links weren't real links, they were all 404s, but I was like, “This looks like a real link. This looks like the kind of thing where you would've learned this.” So that was an example where it was like, “This is what a link where I might have learned this would look like, not the actual link.” So yeah, there's a bit of a logic gap there. And then for the coding stuff, I mean, it can be small mistakes, it can be big mistakes, but it's just as we've talked about many times, you don't want to copy/paste without looking one layer deeper. If you're copy/pasting from Stack Overflow as we've talked about many times, you might be introducing bugs and security errors and other vulnerabilities. Copy/paste with care.
MK They solve the same issue essentially. A lot of what Stack Overflow does is, somebody has a question and they get an answer, and a lot of the times it's not the exact correct answer that will fit within your code base, it's just a really good point of reference to be like, “This is the foundation of your answer and then you can kind of modify it to integrate with whatever you're doing.”
BP I think that that's very much more the way it should be looked at. It's like, “This is a sketch and you've got to run this through and test it and see if it works and see what errors it puts out, and then from there you could apply it yourself.” So it's very much giving you a rough draft.
MK That would be so weird if Stack eventually included AI for some kind of default or draft answers. So say for example, you post a question and you get an AI response, which is, “Have you tried this?”
BP It seems inevitable, right? These models are going to be pretty open source. We have an equal amount of great data about coding as anybody else in the world, and we're trying to help people solve problems. So, I'm getting this error, why do you think this is happening? The AI will give you an answer. If it solves your problem, great, let's move on. If it doesn't, let's ask an expert. Yeah, I like that. All right, well next on my list, I had something called Replit, which popped up in my Twitter feed today, and it was about coding on mobile, which I thought was really fun. Do you know anybody who does this or anybody who has ideas for just whatever device I'm on in the moment I'm inspired, or I want to collaborate with my team, pop into the IDE and get going?
MK No, because I feel like that's almost a psychopathic thing to want to do, to be honest. It's really difficult to be on your phone programming because it's kind of like you need a pegboard. You need all your different little documentation and devices and windows and stuff like that open. I can imagine programming on a mobile device to be an incredibly challenging thing to do so I don't know anybody who's done it to be honest, or has expressed a want to do that.
BP I remember Paul Ford talking about this, just sort of saying, “This is something I like to do, to read code on the bus sometimes, check in, and it just sort of gets my brain working.” But you'd have to be a total wild maniac to actually want to do a lot of your work like this. But I don't know. I mean, I think in areas where maybe people don't have access to a very advanced or complicated system this could be a good starting point. Or if you're just feeling inspired and you want to make a small tweak, or somebody throws a PR your way and you're living that remote lifestyle, you can make a small change. So I kind of want to have these people on the show to talk about how they do it. They say they've got their own IDE and package manager and all this stuff that tries to pull together a lot of the more powerful tools into a mobile friendly version. Not sure I believe it, but worth looking into.
MK Because the thing is now, mobile devices are powerful enough where they can run a lot of these things. With a lot of the iPhones, there's more than enough juice there to be able to do the things that we do with our laptops for the most part, unless you're doing anything with containers or anything that's large databases, all that kind of stuff. So for example, I've been looking more into algorithms and stuff like that recently, and sitting on the bus or on the train or whatever and looking at how people implemented certain problems and being able to run that, I have to use CoderPad or Stashpad or one of those kind of web-based online IDEs to be able to mock up a quick algorithm and run it and test it. So something like that would be quite useful. If they, say for example, implemented an algorithm training course where you are just looking at smaller discrete units of functionality, or even things like CSS snippets where you have Pickle Rick in CSS and you can kind of dive into that and tweak it and do that kind thing, I can understand the use case for that.
BP I think, like you were pointing out, some of this stuff is starting to become table stakes, right? Like, this has its own little AI code assistant in here. If you aren't able to use Git at the moment, it's got its own little multiplayer version in here. So they're playing around with a bunch of ideas. It doesn't seem like something that would be adopted by a large enterprise, but it's kind of a grab bag of a bunch of what if’s. What if we just did this? So if anybody's messed around with it and has something to say, we'd love to hear. If folks have other ideas for how they code in mobile I’d be interested.
MK I don't want to poo-poo on this at all because it's always fun seeing interesting things that haven't been done before pop up. So yeah, if you do have a valid use case for this and want to share it, please let us know.
BP Are you on Mastodon? Do you have your own Mastodon server? Can I hop over there? Isn't it basically just sending each other email? Explain Mastodon to me like I'm five.
MK Honestly, I haven't looked into it just yet. I've opened up the page and it looks alarmingly similar to LinkedIn if I'm being honest. That was the vibe that I got and that's the last thing that I need if I'm being honest is another LinkedIn clone. But as things progress with Twitter it seems like there's a lot of people trying to solve the same problem space that Twitter had. There are companies building things from scratch, there are existing platforms like Tumblr and Mastodon which are trying to kind of take some of that market share away and become the next thing. From what I've seen, a lot of people are jumping across to Mastodon as the kind of primary alternative. I'm not sure if it's fueled by that they're just wanting to have a digital footprint there in case something goes wrong or whether or not it's–
BP The life raft, yeah. Exactly. I've seen that vibe, yeah.
MK Yeah. How have you been progressing with your Twitter usage over the last couple of weeks?
BP I took the app off my phone just to chill out for a little bit. I only use it when I'm on desktop. And I don't know. To me, I don't know if it's because of the GPT Chat stuff that dropped or whatever, but for some reason I've been enjoying it a lot more recently. It feels a lot more technical, interesting, just sort of focused on what people are building and learning and a little bit less drama, in my feed in particular. I don't really know why that is but that's been my experience of late. I cut back on it a little and I've been enjoying it a little bit more.
MK I've actually taken off all notifications from Instagram on my phone. I've also deleted the Twitter app and just use it on desktop when I'm working, and I've found my phone not going off every 15 seconds to 30 minutes has been quite nice actually.
BP I smell a New Year’s resolution coming up.
MK Yeah, there we go. The digital detox. I am going to have a look into Mastodon after this though and just see whether or not there's any content on there that I especially vibe with. But my Twitter feed over the last week has been predominantly people complaining about Twitter, which has been frustrating. And then people calling out people complaining about Twitter, which then creates more discourse about Twitter.
BP Gets a little one-note. Yeah.
MK Too much.
BP All right. This is one that Cassidy dropped into the newsletter which is coming out next week, so it'll have been out. Physicists have created a wormhole using a quantum computer. Sounds like sci-fi. Sounds kind of terrifying. But luckily it's only a holographic wormhole, so not the real deal. And also this is how we learn. The unprecedented experiment explores the possibility that space-time somehow emerges from quantum information, even as the work's interpretation remains disputed. I mean, all I need to know is that they've made a wormhole of some kind and I feel satisfied with that.
MK I think ever since we did that episode, I think he was working at Quantum Computing at Amazon.
BP Yes, from AWS. Yeah.
MK Yeah. That was probably one of the episodes that we had that really kind of shook my perception around our future. Because the quantum computing stuff, it's solving really difficult problems for us, and whenever I see anything that comes up with quantum computing, I take a lot more interest in it now knowing about the kind of problem space that a lot of these highly intelligent people are working on.
BP Yeah, I believe some of this may have come out of the Caltech lab, which is where the Amazon Quantum Solutions program, that's where their sort of partnership with academia is based. So I think there may be something to that.
MK When they say create a hologram, is that like a physical hologram that we can actually see in the real world or is this a digital hologram? I'm not quite sure of the nomenclature there.
BP Let me look here.
MK Unlike an ordinary hologram, the wormhole isn't something we can see, while it can be considered a filament of real space-time, according to co-author Daniel Jafferis of Harvard University.
BP Yeah, that's what I was going to say– a filament of real space-time.
MK Oh yeah. You just had that kicking around in the back of your brain.
BP It was right on the tip of my tongue. I mean, I think basically it's like, in concept, in theory, in digital space, in maybe even a projection of digital space, they made a wormhole. In our seeable, touchable, physical reality, no, they have not. But that's how quantum works. It makes fools of all the smartest people who are trying their best to make physics make sense, and then it sort of remains just outside our grasp.
MK I do kind of wonder, just with how technology has been progressing over take the last 50 years for instance, where we were solving problems like, say for example, developing radio transmitters and then Bluetooth, and then we had things like Wi-Fi and iPhones, computers, all that kind of stuff. And the problem space seems to be getting more and more difficult as we move into quantum computing, AI, all that kind of stuff. It seems like the envelope between where people can contribute to technology is being pushed further and further and further out where people are having to study for longer periods of time to be able to generate value.
BP That's true, but you don't have to study at all to ask the AI to do something miraculous and it'll just do it for you in a second.
MK Very true. Maybe this is the tipping point.
BP Yeah. All right, everybody. I had one more thing I just wanted to shout out. We can't really share it with you on the podcast, but I'll put it in the show notes. Disney shared their latest AI tool that can either make you look older or younger, and in the past they used to achieve this sort of effect in Hollywood. They would film and then they'd go in frame by frame and somebody would retouch it, or you'd wear a rubber mask or whatever. This one is really astonishing. It's like if you had the greatest Snapchat filter of all time in 4K, but as I'm moving around it's sticking to my face. So go check out the note in the show links. I feel like pretty soon you're going to be able to stare in the mirror and really get a pretty concrete picture of what you're going to look like. For better or worse, I don't know. It's a really awesome demo. The folks at the Disney AI Imagineering stuff are always releasing cool demos of what they've been working on.
MK I really hope they don't come out with smart mirrors that can kind of incorporate filters as part of that. That would be pretty brutal.
BP Oof. Yeah, “This is what you will look like if you don't do x,” and then there's a link to the product.
MK Yeah, it syncs up with all your nutrition and fitness goals and will be like, “This is your projected outcome in two years time.”
BP Right, right. Exactly. Oof. Our new Black Mirror episode, coming soon.
MK Yeah.
[music plays]
BP All right, everybody. It is that time of the show. Let's give a shout out to Ralf Htp for winning a lifeboat badge, coming on Stack Overflow and saving some knowledge from the dustbin of history. “How to listen for and react to Ace Editor change events.” Appreciate it, Ralf. You've helped almost 20,000 people. It's always great to see folks come on and spread a little knowledge around the community. I am Ben Popper. I'm the Director of Content here at Stack Overflow. You can always find me on Twitter @BenPopper. Email us with questions or suggestions, podcast@stackoverflow.com. And if you like the show, leave us a rating and a review. It really helps.
MK And I’m Matt Kiernander. I'm a Developer Advocate here at Stack Overflow. You can find me online on Twitter and YouTube @MattKander.
BP All right, y'all. Thanks for listening and we will talk to you soon.
MK Thanks, everyone. Bye!
[outro music plays]