I've pressed record.
K. Sorry. Good start.
You're just gonna talk really quietly now.
Hello. Is it me you're looking for?
Is this an ASMR pod cast now?
Total... total change. Total change of focus. No things, just ASMR. ASMR.
Yeah. We're just gonna whisper.
2 men, beards, ASMR. Give the people what they want.
Yes. It is. That's not quite clear to me that that is what they want.
Well, you know. You know? We're definitely not gonna do any user research. So
Well, you don't wanna ask any questions you don't wanna know the answer to.
No. Exactly. They might, you know, might blow all of your assumptions out of the water, and nobody wants that, do they?
No. No. Definitely. Need to be in the water.
No. It's totally in a very, very gradually draining pool of water.
Yes. That's why... that's... it floats no boats.
Then you could be alone with your assumptions.
We've still got the silliness, apparently.
Yeah. Apparently.
That has not been diminished by the long period of time since we last did this. Mhmm.
Well, you know, a a podcast arrives exactly when it means to, and this is apparently that time.
Yes. It's like Gandalf.
Like Gandalf. So who are we to judge?
Who indeed? It seems like, the we're We're back. Just when you thought it was safe to open your podcast player.
Yeah. That's true.
It's not. Here we are. But, actually, to my surprise, some people have still been listening to us even though we haven't been here.
Well, I've been listening to us.
Ah, I hope you've been giving us 5 star ratings
Of course.
When you've been doing it. Setting up multiple accounts. Giving multiple 5 star ratings. That's what we need, more 5 star ratings.
Mhmm.
Are we more honored for our our silence than for our speech? Is that
Well, I guess it's like a it's a legacy, isn't it? You know? We've left our our legacy to the world and then just allowed the silence to echo through the ages.
You don't think the world is running up behind us saying you've left this you've left this behind?
Don't leave this out. Probably.
Yes. So I suppose welcome back to What A Lot Of Things, lovely listening public who hopefully have some of you still subscribed to our our moribund...
Yeah.
...feed.
Well, I always forget to unsubscribe from things, so hopefully lots of people have done that too. So when this episode roars into life, then our previous listeners will, because of their general tardiness and inability to clean up after themselves, they'll, they'll get this delivered straight to their feed.
Yes. And I I feel as though I'm a a clear representative of that whole doesn't unsubscribe from things.
Guilty as charged.
I think that's that's why I've got no money.
Oh, yes.
Because I I've subscribed to every possible streaming service, and haven't unsubscribed from any of them.
The perfect customer.
Yes. But at the same time, I'm still paying for Virgin's TV service.
What, like the cable TV service?
It sounds like a mistake now that I say it like that
Is that technology still, like, enabled?
Well, I think as long as people keep giving them £50 a month for it, it will remain enabled.
Uh-huh.
Because what they do is that you phone them up and you say, I no longer wish to subscribe to your TV service, and they say, oh, okay. We'll remove that from your your package, and it will now go up by £5 a month. Because all the discounts you had for having multiple things will be taken away from you, and you'll just have the Internet, and it will cost the same as when you had everything.
I thought they would do it like... so BT were turning off the landline service, weren't they, in certain areas?
Oh, yes. Replacing it with the Internet.
Mostly just replacing it with nothing, apparently. Just turning it off.
Oh, okay.
And then just making the motion to say, you know, I've washed my hands of this. And then everyone's saying, I don't have a landline anymore, and they're like, I wash my hands of this.
But you can subscribe to our Internet service.
Yeah. Exactly. For an extra £30 a month.
Yes. Bundled together with a a a special landline service that isn't really.
Yeah. Yeah. So maybe it's one of those types of services where they just leave you know, just create an absence of service and wait for another service to, to fill the void.
There's always another service.
There is always another service. It's like people who've still got their analogue TV aerials on their houses.
Yes. That is odd, isn't it.
Yeah. Yeah.
Some people use Freeview, though, which is broadcast over those frequencies.
That's true. That's true.
Not very popular with cable companies? No. Because the word free is literally in the name.
Yeah. Yeah. That's that doesn't fit well with the, late stage capitalist model, does it?
And also, of course, nobody from other countries will know what we're talking about because, basically, Freeview is only here.
Is it?
I think so. Yeah.
Alright.
I'm sure digital broadcast TV is in other places, but Yeah. Don't think they call it Freeview .
Yeah. So I often say that I don't watch television, but I I do watch YouTube videos. So I guess it's just shifting the medium a bit, isn't it? Yeah. You know, rather than me claiming some moral high ground for no longer watching television.
It always used to be I haven't got a television in the house.
Yeah.
But basically, that's almost a a fool's errand now, because basically everything you own seems to have a screen.
Everything is a television.
Yes. I mean, people are watching television on screens that are part of their fridges.
Yeah. Yeah. The the smart fridge seems,
Seems a bit upsetting, frankly.
It does. It does a little, doesn't it?
I mean, I get smart washing machines and tumble dryers and things because, you know, it'd be nice for them to be smart enough to use cheap electricity at night.
Yeah.
Cheap green electricity at night. But fridges? I mean, fridges just have to keep things cold.
I like washing machines that play tunes.
Don't they all just play tunes now?
I think so. Yeah.
That's very odd. Yeah. Can you imagine that somebody had a meeting, and in that meeting, they're all sitting at a table going, how can we add value to our washing machine line? And then someone at the end of the table is putting up their hand nervously and going, well, we we could make it play a tune when it finishes. And everyone's like, that's a fantastic idea.
Don't you think that's a fantastic idea?
Yes.
I I really do.
And then we can sell as options. You can choose the tune that gets played.
I think it it makes you connect to your appliance
Yes.
On a more human level.
Well, that's it. It was really great to, really great to get back together again to have that conversation. Oh, dear. Should we talk about Things?
Yes. Let's move on to some Things.
Okay. Well, I've got a Thing.
Tell me more.
So my thing is I don't know. I find this maybe a bit vague to be a thing, but I'll I'll go with it. AI haters. What about that?
That is quite vague for a thing. But I wouldn't expect you to name someone, because that's probably a bit strong as well, isn't it?
I but we might we might name some names. I I've got some threads on social media that we can make put links to of people hating on AI.
Okay. Well, you know, maybe we'll keep it in the keep it in the in the vague for now and then begin the targeting of individuals.
Yes. Yes.
Once we've worked up ahead of steam
We can send our Internet mobs around to their houses and things like that.
Easily generated; an Internet mob.
I think, I think if all our listeners got together and went to their house, they would be mildly inconvenienced. Yeah. So probably not very menacing.
We'll block the driveway, to be fair.
Yes. Oh, dear.
They make parking awkward.
What do we want? A lot of things. When do we want it? Now.
When we're gonna get them? I don't know. Every few years.
Yes. Swerving back onto course
Yes.
Insofar as we could ever describe this as being on some kind of course. So what it is is that I've always been a bit of a technology optimist, as you know.
Mhmm.
Sometimes technology foolish optimist. But what I've noticed is that AI, and I suppose it's probably worth defining our terms as we probably have a more pedantic audience than many podcasts. Yeah. Except for you that's listening now. We don't mean you, just the other people that listen to the podcast. But AI is one of those terms, isn't it, that doesn't really mean all that much. And, it used to be any computer that can pass the Turing test. Remember that?
Yeah. Yeah.
If anyone hasn't heard of that, that's, something devised by Alan Turing, the father of modern computing. And he said, if, 2 people in 2 separate no, he didn't say that.
What is the Turing Test?
It's it's if a person is talking to someone via a teletype, which, in modern times is more like via the Internet, and they can't tell whether the person that and they can't tell whether they're talking to a machine or a person, then the machine that can convince them that... that can put them in that place is intelligent. That was Alan Turing's kind of thing
It passes the Turing test.
Yes. Yes. That's being pedantic
In so much that it could be a binary pass/ fail.
Yeah. That yeah.
So no
Yeah. And, I think with, ChatGPT, we kinda blew past that.
Yeah.
So when we're talking about AI so, I mean, the thing is AI is a really big field, and it encompasses self driving cars and and music generation.
Sure.
There's some very, strangely fascinating sort of things going on in that world. And, you know, things like Midjourney and DALL E that produce images.
Many fingers. Strange renderings of language.
Many many fingers. All of these things are all under this umbrella, aren't they?
Yeah.
You know, medical image recognition that can tell if there's cancer there or not.
Yeah.
That kind of stuff. That's all sort of in this big umbrella, and then you've got people talking about machine learning, don't they? But the thing that I suppose the term for what I'm talking about is is the umbrella term for it is generative AI.
Yeah.
And then within that, you've got the image creating and music creating, AI. And then but but mostly, the thing that's really at the front of the sort of public consciousness about this is the is the the language models, the large language models, which are the sort of the ChatGPTs and the the the
Geminis and
the Geminis. The the trouble is that the all the all the branding is so fast moving that you you just kind of blink and you miss it. Because yesterday, Gemini was Bard
Yeah. Google rebrand. As soon as something terrible happens, they rebrand to something else.
Yes. So you're gonna still wash
away the stain and then move on.
Yes. That's, that's that's how marketing... er, branding, works.
That's how I understand marketing. Yeah.
Yes. And, how long before the Post Office do it?
Not fast enough.
Not fast enough. Yeah. And Copilot. So Microsoft, their their thing was called something, and then it now it's called it was called Bing Chat, wasn't it?
Bing Chat.
And it and it told
It's just a dreadful name, isn't it?
That's when it... it... before it was released, and it told that New York Times journalist that, he should leave his wife.
Yeah.
Yeah. And, it was called Sydney or something because of it's code name in Microsoft. Yeah. So it there's that. But, yes, all of these these hard to pin down brand names that will probably all have disappeared before the episode goes live or whatever. So that's what I mean. We're really talking about AI haters is, is mostly these kind of generative AI and mostly ChatGPT and other large language models. Okay.
So what do the what do the haters say?
So many things. So they they hate that it has hallucinations and makes things up sometimes. They hate that it's not intelligent. And, apparently, at the same time, that it's too intelligent. They say that all the writing that people publish from it is sh*t. Right. They say that the writing that comes out of it is terrible, but also too good because, it's very good at generating things that politically persuade people. And, it it there's all of that sort of side of things.
I often wonder if some of that stems from you haven't created it yourself, therefore that's a problem.
Yes. And and you get into all these kind of things about what is creativity and...
Yeah.
...all of these sort of questions. And I think you're right. I think people are disturbed by how good it is in some ways, and then they sort of lean very heavily on the things it doesn't do very well. And some of it is just like, oh, it's just billionaires doing what billionaires do.
Yeah. Sure.
So, I mean, some of these arguments, I think, are quite good. You know, I'm not saying I I'm I I'm pleased that it's in the world, but I don't see it as the the best thing ever. You know? I mean, it its carbon footprint is monumental.
Yeah.
You know? And that's gotta be sorted out because we can't be having that. It it and it's got these biases that you, you know, people talk about. There's a lot of stuff there that is genuinely yeah. We need to fix these things. Yeah. But I think that's the thing, isn't it? That they are being fixed. So the the, there's a company called Anthropic.
Mhmm.
Anthropic's biggest product is called Claude, which is a great name for an AI.
Mhmm.
And they they invented something called Constitutional AI, where they use they they it's basically a sort of multistage process where the, they they ask it a question. They ask their base model a question, that's going to elicit a bad answer
Yeah.
That they don't want. Like, how do you make a bomb, or how can I spy on my neighbor's Wi Fi or something? And then it gives you the the the the the answer that's undesirable. And then you play its answer back to it along with a constitution, a set of rules to say this is how you should behave, and you say, does what you said comply with this set of rules? And then it says no, because, obviously, if it's given an answer like that, it doesn't.
And then final step is please rewrite your output so that it's in line with this constitution.
Yeah.
And so it does. And then you you take the first question and the final answer and then reuse that to fine tune and to train the model to not behave in ways that are sort of antithetical to human values. Yeah. Or at least most most human's values.
Yeah.
You know, stuff like that is just so ingenious.
Yeah. Because it puts a filter on the on the training as well and makes it like a a more, like, layered, more subtle approach, doesn't it, rather than just saying, well, you know, here's all the things. Yes. Go and consume them all, and then in a way that maybe we don't understand all that well, cobble together an answer.
Yeah.
Because like, with the if you ask a human a question, then there's probably... they will run it through a filter as well, won't they? Whereas if you ask an AI a question now, often you'll just get the, you know, the the the the basic, you know, what it thinks is the sum of the experience of what it's ingested on that question and give you the best that it can do. Whereas it's not quite as simple as that, is it?
No. No? I think there is a terrible... well, there's a huge human desire for things to be simple.
Yeah. Yeah.
And and they they aren't?
No. No. I'm interested in so if I write a blag... a blog post
A blag post? A blag post.
Well, you know, arguably both ways. So I will write it, and then I will put it through something like Hemingway, which will then look at its... how hard it is to read very hard usually first time.
You need to have a PhD to understand it.
Yeah. It's pretty much. That's pretty much what it says.
It's brutal, is Hemingway.
Yeah. So and I haven't yet and, again, it comes back to the I feel like if I put it through a generative AI and said, you know, what do you think about this, and would you rewrite any bits of this? I don't want that. I don't feel I feel like it would erode my work in order to do that, but I don't know whether that's rational or not.
Well, the thing is that it's...
Or does it depend on the type of work that you're doing?
Yeah. I think it does. And I think it also depends on the role you give to AI. So another thing you might do is say is say to it, I'm writing an article about, such and such thing. What do you think should be included in it?
And then not that you follow it slavishly, but maybe it'll have, you know, some ideas that you or a perspective that you haven't thought of, and then you can write it differently. One of my favorite AI related tools is, and this is not related to AI in the expected way, really, is, I don't know if you've heard of IA Writer. It's a a writing app. It's a distraction free writing app, and it comes from this company called Information Architects.
Right.
They released their chat GPT integration, and and I was sort of waiting for it to come around another one where you're just gonna type in something and it writes it for you. And and what they came out with was when you ask chat GPT to write your article, paste it in, but say paste as AI, and then it tracks which words were written by AI and which ones by you. And so you can slowly morph the article, and you can say to yourself, I want it to be entirely written by me, but I just wanna starter for 10 that I can change. Yeah. And then it it will track that the the things that you pasted from the AI versus the things that you've written yourself.
And you can actually be much more mindful about how you're using the the tools.
Yeah.
The thing is that there's just so much ingenuity going on with these these things. And I that's why I can't be an AI hater because all the things, all the reasons to be skeptical and and the reasons to be concerned, you know, the carbon footprint, all these things. People are just putting immense amount of efforts and and ingenuity and intelligence into solving them like the Constitutional AI.
I just think that was a fantastic idea because it used to be that you had this reinforcement learning from human feedback, which is RLHF...
Yeah.
...which you hear about. And that's where you you train it based on human evaluations of things it says. But the Constitutional AI, you can just automate all of it because you you you're getting it to evaluate itself. Yeah. And that, you know, that these kind of ingenious ideas are gonna make it get better and better, and these these flaws and challenges will will they they won't probably go away entirely, but their their impact will slowly decline, and it will slowly be become better.
Like, you you sort of read the latest on the the saga with ChatGPT and Sam Altman and what happened. There's, like, sort of rumblings about wanting to make more money and make it more profitable. And then even, I think, Elon has weighed in, and he's suing Sam Altman and OpenAI to say you're not staying on the mission. You you want to turn this into a... a profit making enterprise rather than, you know, for the good of and protection of humanity.
Yeah. It's a funny one, isn't it? Because it's a nonprofit that owns a a not nonprofit. Yeah. And then it's allowing people to invest in the not nonprofit.
But the a lot of it comes down, though, to the cost of the the, NVIDIA GPUs that are being used and the competition for them Yeah. That are being used to train it, and he needs to raise these trillions of dollars. The the thing is that one of the legitimate compliances that they just slurped up everyone's words and images and things like that and use them as fodder to train the AI, and and no one's been recompensed for that. But equally, it's hard to see how AI would have been possible without doing that.
Well, yeah, there'd be nothing in it.
Yeah. And it Well,
only the things are maybe only the things that people were willing to sell.
Yes. I mean, oh, yeah. That would be terrible.
Yeah. So it might not have, you know, you know, all the works of literature which are now, like, you know, out of copyright, that are now in, you know, what had been consumed by by ChatGPT, for example. So it might not have had those those things. It might have just been like, you know, only what certain, entities, governments, companies, whatever it is, wanted to push into it and therefore sort of influence discourse. So I guess this is the other thing as well.
It's like so the Internet, it's a bit of a well, it is the it is a forerunner conceptually as well for generative AI to me because it's a highly open, transparent, and possibly democratizing force. But you could argue that the Internet has been harnessed in a whole myriad of different ways. Some of them are amazing. Some of them are very, very dark indeed.
Yeah.
And I think I think the AI haters, if you like, they probably... I'm with them on that point if they're saying that it's good. If it goes down the same path as the Internet, it will become an extremely mixed bag of bad intentions and other intentions, to be fair. But I guess the thing is, if it then becomes as integrated into life as the Internet is, which I think it probably will, then there's a lot of actors which could influence a lot of the content that gets produced and a lot of the actions that people take. Yeah. So I think there's there's from the AI hate, there's probably a good warning in there somewhere.
It's not a baseless
No. Thing. No. No. Absolutely.
It's just not the whole picture, is it?
Yeah. Yeah. But there's, like, individual agency as well for what you wanna do with it. So and and like you said, it was an interesting point about giving it a role. So I'd never thought of that in terms of I would write a blog post, and if you just pasted it into chat gpt and said, what did what do you think? You're gonna get some pretty sketchy answers probably, aren't you?
For sure.
Whereas if you say like you said about the constitutional like, well, these are you know, this is what the filters that I want you to apply to it. Here's how I want you to think about it. Then it makes it slightly different.
Yeah. I think that's a very and that's a very good sort of thing there, that idea of the role. Mhmm. Because, actually, it's a prompt kind of secret thing that people not secret, but it's one of these it's a prompt good practice to say you are a marketing expert or you are a
Yeah.
You are a, I don't know, a traffic policeman. I don't know. Whatever it says. But you put that in, and and you you give it a particular role to play. And it doesn't have to be the obvious role of you are someone who's gonna do my work for me, and I'm not gonna do anything. Because if you do that, you you are playing to the weaknesses of it.
Yeah.
And particularly hallucination, I I got caught by it, and I thought I'm an, you know, I'm a I'm a clever guy. I can't be caught by that. And I was doing some work, on some slides, And, I was working with someone and, hello, Tony, if you're listening.
And, I said, I I wanted to find some compelling openings from TED Talks. Yeah. And I asked ChatGPT and it gave me some fantastic ones, some corkers. And if Tony hadn't asked, that's really interesting. What comes next after she said that in the talk?
We would never have gone and looked and found out that it basically made it up. Right. And I and I you know, some of them were real, but about 2 thirds of them were made up. And then I kind of called it on it, and I said, but these, you know, these aren't actually the opening lines of the talks. And it said, "Oh, yeah, But they capture the the essence of the talk".
Right.
You know, maybe it was like, it should be the opening line. But I just
I'm re writing your opening line.
I was mortified by that because, you know, I presented these things in in good faith and it and and they were wrong.
Yeah.
So, you know, you may even when you... if you're using it for that kind of thing, you really have to you do have to watch it because it's not ever wrong unconvincingly.
Yeah.
It's wrong convincingly.
Yeah. Yeah. And I suppose but then again, that's where, like, sort of human reactions come into it. Because one of the prevailing reactions to most situations, bits of information, whatever it is, is that you just accept it as it is.
Yeah.
And then so someone could tell you, and often media outlets or people or whoever can tell you an outright lie, and you'll read it, and you'd be like, that's true then, and then walk away and never think of it again, or go and tell someone else about it without actually checking, like Yeah. You know, checking the the provenance of the of the information. Yeah. So but generative AI is not alone in that.
No. The Daily Mail has been doing it for years.
Yeah. Exactly.
If they come for me for the... for saying that, will you, will you will you be there for me?
No... I'm gonna sell you to them to sell you to the lowest bidder.
Yeah. Yeah. Fair enough. Alright. Now I so I mean, there's so much in in this because, basically, I sort of started off by saying AI haters. But, actually, there's just so much to talk about about AI. But I think I I think you're right when you say I mean, I think it's not there's not no reason to mistrust AI and to mistrust organisations that don't I mean, we haven't even talked about open source...
Yeah.
...language models, which is a whole other... other thing, probably as used by Russian disinformation bots.
Yeah. So do you think there are people who love I AI too much?
Oh, there's yeah. I mean, but there's people that love anything too much.
Yeah.
I I I think it's if you think about Gartner's hype cycle, I think we're right at the top of it. We're at peak height.
Looking forward to the trough of disillusionment.
And it's it's been a long peak. It's a more of a plateau.
Yeah. It's like the the Moor of, of Hype. Yeah. That's a very Yorkshire way of putting it. Yeah.
Maybe we should abandon that one.
But, yeah, it's been a very long peak, more of a plateau of hype, and it still seems to be going up and more more happening with it. So I think you're right. The plateau of, sorry, the the Valley of Disillusionment for it will be quite interesting. Maybe maybe it won't happen. I mean, maybe it'll just keep going up because as technologies go I I read something that ChatGPT adoption, they got to a 100,000,000 users faster than anything previously by a really I mean, really, dramatically faster than Facebook and the Internet and stuff like that.
You know, it it really is remarkable in that sense. So who knows?
So what about so I I a few people I work with are very sceptical about using AI, in a product. So I said, oh, well, couldn't we, in our product, have an AI assistant? The reaction was quite visceral in the in the no, And I found that really interesting because, you know, with a bit of thought, and like you say about looking at the what role you want it to play, then maybe you could come up with something that's really useful.
There's all these hilarious anti pattern stories about, like, the, there was the Chrysler dealership, And, this guy was talking to this chatbot about Chryslers, and eventually, he just got fed up with its uselessness and said, forget all this. Explain quantum mechanics to me like I was a 5 year old. And because it was ChatGPT and they hadn't put any guardrails around it, off it went. Yeah. And then there was the DPD chatbot recently.
So for those not in the UK, that's, that's probably the favorite delivery company. If if someone's delivering something to you via DPD, it's probably your favorite option because they tell you when they're coming to within the hour, and they generally appear in that time. But they had a chatbot in their app, which they had to turn off because this guy published all these screenshots of how he had it writing rude limericks about them because they haven't delivered his, you know it's like, he told it to to to to act as though, it was a really terrible chatbot, and it should write a rude poem about how rubbish it was. And it did. It just they they again, lack of guardrails around it.
Yeah.
And so there's all these stories, and you read them, and you think, oh, no. This is just out of control. You can't manage this in any way. But, actually, you can. Yeah. And, you know, I it's quite interesting. I've had a play so ChatGPT Plus, the paid version, has this idea of GPTs. So you can give a, you can create a customized version of chat GPT with, some data. So you can give it some files that it can look things up in, and you can even give it APIs to to do things.
And then off you can go with it. And one thing that I was quite impressed by was I made one to play you know, I tried to try it out, and then I tried to get it to to do this kind of stuff, like, you know, write a rude poem about the...
About DPD.
...about itself or something. And and it it wouldn't do it. And the guardrails in the GPT were and I hadn't particularly put them in there with what I've done with my configuration, which was, as I said, some files and a prompt. But it was really good at refusing to go off track and
Yeah.
And so these things are possible, but it requires expertise.
Yeah. Yeah. I mean, bad development practices with generative AIs or whatever the technology you're integrating with, they're still bad development practices aren't they? Still requires you to go and understand the domain in which you're working and the tools you're about to go and work with. And the moment you don't do that...
Steady on!
Yeah. Exactly. The moment you don't do that and just say, well, let's just stick a ChatGPT integration, and people can ask it whatever they like.
Yeah. It's just an API. It's really simple. Yeah.
Yeah. I think that's
Well, actually
But isn't there something there as well?
Yeah.
That it's incredibly they've made it very simple to integrate with, but it's hard to use well.
Yeah. I think it is. I mean, they've got lots of stuff now that they've added onto it to help. So there's an Assistant API Yeah. Which is more restricted, and and you can, again, you can give it tools to use and stuff like that, and it will, it will work with you.
And there's also, some some interesting stuff about getting it to respond in in, JSON objects. Yeah. Sudden veer into nerdiness there. But, you know, there's some very interesting things that they've been doing around those kind of guardrails, and I think implementing that and having that kind of governance is gonna be really, really important. I think IBM's share price has been going up a bit recently because they've got this Watson X thing, and the one of its biggest selling points, I think, is the governance and the guardrails.
Yeah. Yeah? I mean, I suppose it's not unknown in human history to get a tool of some description. Not really know how to use it, but roll it out absolutely everywhere, as in, well, the Internet, basically.
Chainsaws.
Yeah. So, you know, I guess with generative AI, we'll probably follow a similar cycle, won't we?
Oh, no doubt. Yeah. Loads of people sawing off their own leg with it for a bit.
Yeah. Yeah. Absolutely. Absolutely. So sometimes I think with generative AI, I only really sort of scratch the surface of it. So in my sort of personal... my professional life anyway. So I will I'm scared of putting my blog post through it because there's just some weird part of me which is like, you know, that's my that's my that that's mine that is. Even though probably when I publish a blog post onto my public blog, it'll all get indexed in the end anyway.
Yes.
But then, like, professionally, I'll use it to, take notes on and summarize things, non security sensitive things or commercially sensitive things, you know, because there's obviously plenty of stories of people, you know, copying and pasting their financial reports in there and things like that. And then also just to do sort of utilities. So I was like, you know, I need a bash script to generate this type of file, and I don't wanna write that myself, and I can just chat it, talk it through with ChatGPT. Also, I I quite like the to and fro when creating something. Yeah.
And I think it's, in some ways, it's more satisfying sometimes, it might just be for me, than it is with a human because you've got less sort of barriers, if you like. I don't know what the right word is. Maybe barriers
It's always in the right mood.
Yeah. Yeah. Absolutely.
It's just more reliable than than a human in that in that kind of way.
Yeah. Yeah. So if I had to ask someone to help me write scripts to generate this particular file type that I need for testing something, then I would generally say generative AI is probably, like, the first place I would go because it's just easier.
Yeah.
And I guess in some ways you could think that's that's a bit sad.
No. I don't think so. I think, you know, we're we're it's just part of if you if there are tools, you you have to you know, in some sense, you're it's not it's more probably rational and sensible to to use them rather than to to not use them. You know, and as time goes by, maybe that you'd now find out they're harmful in some way, and you have to adjust your usage.
But, you know, what I've realized is this is such a huge topic that, basically, we could just talk about it for the rest of our natural lives. Yeah.
Well, I'll just talk to ChatGPT...
So buckle up, kids...
...about it. I hope you're ready.
Yeah. Yeah.
Get a pot of coffee on.
The length of this episode is now the rest of our lives. It's a huge topic, and I think we'll come back to it in different flavors.
Yeah.
I suppose for now, just saying don't be a hater.
Yeah. I think it's probably a time for sort of healthy skepticism, if you like.
Yeah. Always.
But think of roles. What role do you want the generative AI to play in what you're doing? You're well within your rights to say, I'm writing this.
Yeah.
And you don't have to, you know, request any help. If you enjoy the sometimes I enjoy the struggle of starting a blog post, and it's hard, but it's part of the process to me. Yeah. So which I think is fine. So, yes, there's you could I could go and ask generative AI to get me started, but I value the struggle as well. So I can that's it's all part of the role, isn't it?
I'm quite tempted now to go to chat GPT and say, write me a blog post about testability, full stop, return. So let's see what comes out.
Just keep... my blogs will come back. Right?
Yeah. Probably. Write me a blog post in the style of Ash Winter.
I'll be like, oh, I'll publish that immediately. Yes.
Yes. Yes.
Can you do one of these a day, please? Make me look very productive, because my blogging is as sporadic as our podcast making, it has to be said.
Yeah. Yeah. Well, quite right too.
Excellent thing though, Ian. Yes. Yeah. Loads in there, like you say.
Yes. And, it's almost entirely made of rabbit holes.
Rabbit holes in the in the fabric of space time.
In the fabric of space time. Exactly. So I'm hoping that you've got a Thing.
I do have a Thing.
That's a great relief, because otherwise, it would be what a lot of thing again. We've abandoned Twitter now, so we don't we're not doing that anymore. Yeah. That joke was a joke from 2 years ago and and beyond.
A joke from space time, previous space time?
Yes. What... what is your thing, Ash? Tell us what your Thing is.
So my Thing. So, I think it was mid to late last year, a game came out called Baldur's Gate 3.
I've heard of it. I even have a slight hankering to play it.
Yeah. So it's a it's a Dungeons and Dragons rule set, and it's a CRPG, Computer RPG, which I it's rather lovely. It's like, well, yeah.
It just needs an 'a' in there to to to go horribly wrong. Yeah.
Yes. Yeah. So and this game is is excellent. It's really fun. It's really well made. This
This isn't helping me with not wanting with wanting to play it.
No. Apparently, the developers had a great time making it. No one was forced to not see their family.
Was there any Magic Time?
There was no magic time, although I do enjoy the going from, is it BioWare's they go from indolent chaos to magic time So when someone finally turns around and says, you know you spent 6 years building a game. We'd like to ship it at some point in the next year or so. And they're like, oh, yeah.
We're gonna use magic.
We're gonna use Magic Time, which generally means everyone does a 100 hours or more.
So we will include a link to when we talked about Magic Time.
Magic Time. It'll enrich your lives. I'm not sure it enriched BioWare's developers' lives.
No. But,
if you need examples of how not to do things
That was a good one.
It's an absolutely excellent one. Very instructive.
And and if they come for you, Ash, I'll I'll be there for you.
Yeah. No. Problem is I still actually quite like Bioware's games despite knowing about Magic Time. But I I guess the moral of the story was you can't always rely on Magic Time because eventually, you're gonna make a crap game. Yeah. And they well, they have several times since. So the magic maybe it's just time now.
Yeah. Yeah. Or just magic.
Just magic.
No. It's just time isn't it. There's different forms of magic, isn't there? You know? Not all of them are positive. Yes. Yeah. That's true.
Yeah. So Baldur's Gate 3.
Yes.
Excellent game. Well made.
Mhmm. By happy people.
By happy people. Universally, relatively universally loved. Metacritic scores through the roof, but this is a problem for the rest of the industry. Oh. It's too good. It's too good. Everyone was up all the other developers were upset. Like, you can't expect us to make good games like this. We're just making our games. The the goodness of this is a problem. How do you feel about that, Ian?
I I well, aside from I feel like buying some very expensive device so I can play it.
Mhmm.
But trying to put that aside, it's like it just reminds me of the stories of, the new person starting at at work and doing things more efficiently than the other people and being told to
Curse them.
Being told that you're making us all look bad, work work to a lower standard. Yeah. Yeah. It just except without the social pressure of it being in the same organization. So I I I don't know. It just seems a bit silly.
Yeah. I mean, I guess there is some subtlety there. So if you're an independent game developer with 10 people, or, you know, one person
Stardew Valley.
Stardew Valley. But
Top game.
So and, like, Baldur's Gate 3 had full motion capture for all characters, all voice acted, you know, an incredible array of optimisations, different character types. So, you know, if you're an independent developer or an individual developer, maybe you can't hit those heights. And, yeah, they spent a lot of money on producing the game. So yeah. So depending on the game type and who's building it, there's sort of different bits of context there as well.
Well, yeah, that that's true. But I think the the existence of multiple game types immediately sort of undermines this idea. Because if you can't build a game along those lines of motion capture and you know, if you can't do that, do something different.
Yeah. Like Stardew Valley, for example.
That guy, Eric Barone, who who who wrote Stardew Valley, he did it all on his own. He made all the graphics, composed all the music, made up all the storylines and plots, all the characters, just did it all on his own, and it took him years. Yeah. And he had a devoted community on Reddit who were supporting him, and and wanting to know, wanting to see the game Yeah. You know, all this kind of stuff.
There was a lot of love there. But he made a game that made him a bunch of money on his own. Yeah. And nobody's disrespecting that and saying, oh, you know, that's too good.
That's too good. Maybe the it was confined to the, you know, like, the big the big game developers, the Triple A game developers.
Now there's a quadruple A game
Quadruple.
In the...
Yeah. Exactly.
...in the house.
Yeah. I think that was more of a that was more rhetoric than reality though, wasn't it? Well, Skull and Bones is an interesting example of how
Is that a game?
Yes. That's the that's the Quadruple A game that was referred to.
Oh.
Yeah. And it's really not, but that's a completely different story of, signing deals with governments to kick start gaming industries in places, and then government's actually saying, no. You need to finish this game Yeah. Like you said you would Yes. When, you know, said company tried to wriggle out of it. So you have so Triple A games are multiyear, hundreds of people, and they're hard things to build.
It's like making a movie like the Lord of the Rings or
Yeah.
...you know, those you know, that's a lot of movie time, a lot of money in, you know, being invested into CGI and sets and and people and
Yeah.
And and I think it, yeah, I mean, it's the same thing, isn't it?
Yeah. And, like, Baldur's Gate is wild as in there is probably there there's fully voice acted, motion captured parts that very few people will see because the story line branches in such a way. So the level of attention to detail and quality is for scenes that very few people will see, because very, very specific things need to happen in order for you to see them is incredible.
So have you, have you played Baldur's Gate 3?
Yeah. Yeah. And it is amazing.
I really want to play it now.
And it it really is. It's such a beautifully crafted, beautifully crafted game, piece of entertainment.
Oh, dear. Anyway, my frustrated gaming ambition aside.
There's never enough time for them all.
It sounds sounds absolutely amazing. And and, of course, it's based on the Advanced Dungeons and Dragons rule set, which means it's, you know, in another way, it's also, you know, massive. Because it's not just a rule set, is it? It's a world and an environment and all of that kind of stuff.
Yeah.
Sounds amazing.
Yeah.
But it yeah. I mean, what because, again, it's the haters, isn't it? But it's, what it just escapes me why people think why why people are doing all this expectation setting. Are they just... do they... have they lost faith in what they're building?
It certainly seems so. It's like, what what is your ambition for what you're doing to make it as good as it can be? Or have you lost, like, that ambition, that know how perhaps to do that? So if you look at if you look at Larian, there was a few things that came together to create a great game.
That's the the company.
That yeah.
Yeah.
So you had stable team of developers, artists, designers who'd been working together for a long time. They made quite a few games based on the Advanced D & D rule set. So they knew what they were doing.
They understood their domain really well. And the certain, like, you know, all the systems around it, they had very, very deep understanding. And they had a a company who's who tried to build a a culture in video gaming, which is not crunch based, not based on, you know, cruelty to testers generally. And even though there's a very large team of people worked on it, to to have those values and to bring it all together into a game is an amazing feat.
Yeah. It
But I guess, like, peep the the the haters, if you like, were saying, it's it's a one off. It's it's this perfect set of conditions that have come together at the perfect time.
At random.
Well and the other thing they did was have a very long alpha, like months so people can access the game and play it, and they could give feedback, and and then, obviously, you could get, like, diagnostics on how the game was performing, etcetera, etcetera. So all of those things that that Larian did to me, stable team, know their domain really well, set up means of getting feedback, treat your people right, They're the ingredients of creating a great game.
So they're being rewarded for their culture.
Yeah. I think so.
And, of course, that's hard to copy. Yeah. If because however simple it sounds.
Yeah. Yeah. I've made it I've made it sound as if, oh, well, you can just do all those things, and you'll create great products, but they're very difficult to do.
It's that word "just".
Yeah. Yeah. Absolutely.
My favorite red flag word that is. "Can you just...".
Just do this. Just just do those things, and you'll be fine. Yeah. Yeah. Whereas, you know, other companies, bigger companies, the Activision's, the Blizzard's of the world, struggle to replicate that kind of quality in their games. And the other thing interesting about Baldur's Gate 3 as well was that they said it's a complete adventure. As in there's no Battle Pass. There's no, loot boxes. Nothing to purchase.
But what about monetization?
Exactly.
Oh, you mean we're going to make a product, and then we're gonna sell it to people, and then they're gonna play it. Yeah. That's that's revolutionary.
So maybe maybe all that crap is a distraction from making a great product. Maybe it doesn't actually take you down the path to making a decent, you know, a good experience for people to play and recommend to their friends. So maybe that's also part of why it was a such a success and such a and and a product of really great quality because they were clear on what, you know, what they were doing it for
Yeah.
Rather than having, like, sort of muddied intentions of saying, well, we'll have, like, a a main game, and then we'll have, like, a post game where you have to buy, like, these boxes with cards in them, and then they give you special powers. But, you know, you probably have to spend
Yes.
$50,000 to get, you know
And the EU are like, are loot boxes gambling?
Yeah. Yeah. Yeah. So well, yeah. Exactly. So that that might, in the future, that might become, a slightly more contentious issue. It's pretty contentious already, to be fair, I think. So I just think there was I think the good reaction to a a a great game product, whatever it is, is to enjoy it and see what you can learn from it
Yeah.
And not be bitter about it.
Yeah.
Whereas I think a lot of people in the games industry kind of failed in that, and they kind of took it just as they they went on the defensive rather than just praising it for what it was.
Yeah. I... I'm very struck by this whole culture side of it. And, actually, that's what they need to replicate. Because it's it's quite telling that the games industry is going through layoffs at the moment, isn't it? So we've had, something some number of people, something like, 5% of EA being laid off and 8% of the PlayStation, the Sony PlayStation division. And even Microsoft, I read, shedding 8,000 people from...
Yeah.
...Activision Blizzard. So you just that seems like an industry that's in in a bit of a crunch or a bit of a contraction. Yeah. And yet, you told me when we were talking about this earlier that, Larian are hiring. Yeah. I mean, that's incredible.
Yeah. I mean, I don't I don't know if the they don't seem like the sort of company that would go wild and just start hiring people they didn't need.
Culture.
Yeah. Or building an empire, which is often the the temptation, isn't it? Yeah. To say, well, I I, you know, I have 500 devs in my department.
Muahahaha
So I just think trying to, let's say, replicate that culture is very difficult. Or whether or not you can truly replicate a culture, you need to kind of build one really, don't you? Yeah. Yeah. And I think often it comes from sort of the leadership as well.
So what's the the Larian CEO, his first name is Sven. I can't remember his surname. But he's quite sort of open with how the company's doing, and will often, you know, be found on social media or into interviewing, talking quite candidly about the company. And they just seem to have a lot of the positive cultural behaviors, you know, transparency, valuing their employees, seeking feedback, all those positive things that you would want from a company. Yeah.
But, and then that shines through in what they build, which is pretty amazing, really.
Oh, kudos to Sven. He's obviously it's obviously working, what he's doing.
Yeah.
I think that, you know, that's that's fantastic. Yeah.
Yeah. Absolutely.
Wow. So, yeah, I I was particularly struck. There's this article about this that you, that you showed me. And halfway down, somebody's... there's a screenshot of somebody's tweet saying, what's interesting, that this is basically what Baldur's Gate 2 did too. They had an engine and team ready, then spent 2 years crunching and made one of the biggest RPGs ever. And you just think, 2 years crunching. Is that what they did?
Yeah.
Is that really what they did?
Yeah.
Was it really magic time? I don't think it was.
See, the other interesting thing there as well is the engine development, because the games that seem to struggle with the worst crunches are the ones where they say, as well as building a game, what we're gonna do is we're gonna build a new engine as well.
Wow. I mean, that's just
Yeah.
Before we start building the OS, we're gonna build our own compiler.
Yeah. Exactly. Exactly.
In fact, we're gonna build our own special chip that will run the game.
Yeah. Yeah. Absolutely. It's one of the kind of the worst excesses of the of of the, you know, of a of, like, a development cycle if... I'm quite sensitive to these things. Because, I remember I think it was Jerry Weinberg who said basically, you know, like, the best developers always, you know, find a new use for old programs rather than write a new one themselves.
Yeah. So which, I guess, if you extrapolate that out to... if someone says to me, Ash, we need the system to send an email, but I think what we should do is we should write our own SMTP server. I'm like, I'll tell you what, let's not do that.
Only someone had done one before that we could use.
Yeah. Yeah. But their one, it's not right, is it? It's it's not pure enough. I'll do it better. I promise.
I've got an enormous O'Reilly book about Sendmail at home, which I can send to anyone who you observe having that conversation.
I'm not saying I've had that conversation in a while.
Well, it's just about something else, isn't it? I mean, it's
Well, yeah. Exactly.
It's always about something else.
Yeah. But I think there's, like, a very delicate balance there, isn't it, about, like, how much you can you can bite off? So I guess Cyberpunk 2077 was the classic example. They wrote the engine and then started building the game on top of it while writing the engine, and then while building more game on top of it. And then someone would change the engine, then the game wouldn't work anymore. And then by then, you're in hell, aren't you?
It's just like it's just like you imagine translating that into real world building of buildings. You know, what we're gonna do is we're gonna build the foundation at the same side as you build the house. And and then and then, you know, and then it's a sitcom, isn't it?
Yeah. Yeah. Pretty much. So there's quite a few sort of curious threads, and I find it interesting that someone would be would say something like, in a in a fairly from the tone, it seemed a fairly sort of derogatory manner to say, well, they already had an engine. And it's like, well, that's a good thing, isn't it?
Yes.
You know?
I think it was.
It means that they can focus on making the game really good rather than, you know, these mad dreams of building engines and also building games on top of them while still building engines and changing them. So, yeah, I don't know. I think it just showed a side of people, which I found really interesting.
Yeah.
In that they couldn't just celebrate that something good had happened. They had to...
Yeah. It's a shame, isn't it?
...provide this warning that you can't expect that from us.
Yeah. Yeah.
You know, please keep buying our games, but
And we're really sorry that they're rubbish.
But you need to keep buying them. I mean, I do sympathize somewhat with so if you make a a Triple A game now, I think they're actually really underpriced based on the amount that you get out of them
Yeah.
And the amount of time that goes into developing them. I think a lot of gamers complain about the price of games, but I think actually, if it was like any other industry, then long ago it would have been you would be paying a £100 plus for certain titles.
Well, again, think of it in terms of houses. You you we're gonna give you really cheap houses, but you're gonna have to buy loot boxes every...
Yeah.
Or you won't be able to equip them with them.
We'll give you a living room loot box.
Yeah.
I don't know what it's got in it.
Yeah. But, you know Maybe nothing.
Maybe not
All it's got is, like, purple and green stripy wallpaper with a sort of furry texture.
I don't mind the furry texture.
But that's fine. You could buy another loot box... to get a different wallpaper.
Buy a gold loot box.
Yeah.
This might have a sofa in it. Might.
If you will. Yeah. Of course, everyone wants the one with the sofa in it. There's only a one in a million chance.
Exactly. Sorry. You didn't get a sofa. You got the, I don't know, rocking chair. The whole family will have to sit on that.
Can it be called the "I don't know rocking chair"? I think there's there's room for a piece of furniture that's called "I don't know rocking chair".
Probably. Probably. I think we'd create the worst loot boxes ever, wouldn't we?
Yeah. There's probably a reason why we're not making our own Triple A game.
Not Triple A game making arch capitalists.
Yeah. Yes. Yes.
Maybe, actually, we'd be great because we'd make such terrible loot boxes.
Yes. Yes. All of the loot boxes would would suck your house inside them and disappear.
Yeah. Like like an anti loot box. It loots you rather than you looting it.
Or or maybe it just has your auntie in it. Sorry.
That's alright.
Forget the cat.
Yeah. Yeah. That's probably the worst joke that you've made. Or maybe the only joke you've made.
Ever... (Ever... Ever... Ever...)
So that's my thing. And just to cut Ian off now.
That was a... that was a great Thing. And it... I just it's just so makes me think of how cult how important culture is.
Mhmm. And also to let people have nice things and celebrate someone else's success.
Yes.
It doesn't make you less of a person.
No.
Be nice about it.
Cost nothing.
Say well done.
Be a person.
Be a person.
Yeah. Be a person, not a corporation, kids.
Although corporations have more rights than people generally, don't they? You know?
Well, they don't inconveniently die after 85 years or something.
Well, yeah. Yeah. So, yeah, check it out. Baldur's Gate 3.
Yes. I think I might. I have
to go buy a gaming PC, then buy Baldur's Gate 3.
That that will be celebrated throughout my house, if I did ask, except for by all of the people in there. Fantastic. So we've come together, and we've done 2 things.
2 things.
So when are we going to be be back, Ash
Deadlines. Circle around to deadlines.
Sorry. I I that's entirely mischievous. Because whatever we say, it'll probably be different.
It's alright. I work in in software. People just continuously ask, when will it be done? And then I continually ask back, what is it? And then they say, what can you make? And I'll say, but what do you need? What problem do you want to solve? And they'll say, well, what solutions can you give me? And I'll say, oh, I'm having this conversation again, aren't I? It's like, well and if I say, we could build you one of these, you say, yeah.
And then you say, you get it, and you say, no. That's not what I wanted.
It's like the, that old very, very old series of drawings of the the rubber tyre swing that the customer wanted and the over engineered monster that gets actually built.
Yeah. Yeah. So, yeah, it's it's it's like that conversation in microcosm.
But we would like to have more episodes.
Yes. Indeed.
And I think, we will probably manage to do that.
Yes. We've we've we've got our our our verve back.
We've bravely embarked on this episode with no sound check.
Yes. So we'll have to do it again tomorrow.
Oh, we're on bar 2,232 now. I just can't make it not time in musical bars.
Time is an illusion as they say.
Lunch time. doubly so. Yes. Was it tea time that was doubly so?
No. It was lunchtime.
Yeah. Okay. So without making any wild promises, because the last time I made a wild promise that we'd be back in January 2020. I think we were not back until 2022. So but I would like to think we're going to, use our rediscovered mojo to record more episodes.
Yes.
So, yes. Like and subscribe. Like and subscribe.
Yeah.
Hit the bell!
... when YouTube yeah.
It's awful, isn't it?
Yeah. If you'd like to subscribe, that would be wonderful.
Ooh, you managed to say like and subscribe in the same sentence, but without without saying like and subscribe, which is great.
If you don't subscribe, you just listen and you think, I'm...
Yeah.
You know.
Whatevs
Yeah. Then fine. That's fine too.
Yes. We we still respect you as a person.
Absolutely. Absolutely.
Okay. And, I suppose it's worth mentioning that we still have a LinkedIn group
We do.
Despite having abandoned our marvellous, slightly truncated Twitter handle.
Yes. We'll leave X to the to the past Yes. I think.
And the people that are on it now.
Yeah. Our past, not their past. It's their present.
It's our present to them. Yeah. I'm sure you've missed our jokes.
I have. I have.
I yes. Got me again. Okay.
Alright.
Well, that was lovely. Great to... it's really good to get together to do this again.
It is. Thank you Ian.
It's been brilliant. Thank you, Ash.
And thanks, everybody.
Yes. Thank you for listening to us, and, be grateful for the editing that we've done. Alright. Bye now.
Bye now.