All this is gold.
I think we might have missed it.
Yeah. So well, you could just say it again. And then when you inevitably say OpenAPI by mistake, we can we can reenact it.
What I'm trying to do is integrate OpenAI with the OpenAPI AI.
No. It's gone wrong again.
It's gone wrong again. There's been a pigeon following me all day.
But you really like pigeons?
I do like pigeons, but when I was at home, there was a pigeon sat on the on the chimney cooing. When I get here, there's also a pigeon sat just outside.
Do you think they're watching us?
Yes.
I have to say though that this is a popular haunt of pigeons, and the sound of pigeons in this room is not uncommon.
Right.
Hopefully, they're quiet enough that we won't have to be editing them out. We won't have to identify the frequency of pigeon calls and put in a notch filter.
Maybe that's what we should put in AI generated voices. A pigeon coo.
Yes. That can be like the watermark.
Yeah.
And then Donald Trump will always carry a a pet pigeon everywhere with him so he can deny that he said whatever it is.
Maybe that's... if... or maybe we could flip it around so you know it's real if there's a pigeon because everyone has to carry a pigeon.
Yeah. Because you could arrange for a pigeon to be there when it's real, but you can't arrange for one not to be there.
That would be an amazing world. Everyone had to have a have a pigeon with them on their shoulder.
Just to validate all of the...
Just to make sure.
...statements.
It's not an AI generated voice. I see no problems with that approach.
No. I think, that's the most scalable startup idea I've ever heard.
To be fair, I've worked for worse startups.
Worse than the the carrying a pigeon everywhere...
Yeah.
...start up. And what do you do with all the pigeon droppings?
You just keep them. You... You just go to the dry cleaner's a lot.
Yeah. I guess you would.
Yeah. Yeah.
Mind you, pirates did alright with parrots, didn't they?
See? It's got a historical precedent, this idea.
Yes. Although I have discovered that, the only reason we think pirates have a Dorset accent is because of the actor that played Long John Silver in the 1953 adaptation of Treasure Island.
Yeah. That makes sense, doesn't it?
So all of this "arrr me hearties" comes from, this this one chap, infusing his background into his part.
Oh, yeah.
So, you know, every day is a school day around here.
Every day is a school day. So maybe we could go back to the golden age of piracy, just with pigeons with Dorset accents. Arr.
I'm not so sure about Yorkshire...
So
...pirate.
Where... a place that I used to work, the the one of the the task managers used to build, like, these, spreadsheets, like, tools to, like, estimate and and, like, generate test cases.
This is gonna be a journey back to the point.
Yeah. And, and automate as well, because it was in a very strange technology stack. But he always used to put like a big so if the tool was called test automator, he would put a big r r on the end. And if it was and the the tool to generate test guesses was called TestMate with a big e on the end. So it was test autofar and test matee.
Oh my goodness.
You know, I I was happy to work there just for that just just just for the, the level of joy that that brought me when I first saw it. I was like, this is the best job I've ever had. It wasn't. It wasn't. But but that
But it that blinded you to it. From you stayed there much longer than you should have.
Absolutely. Absolutely. I would think it's time to leave, and then someone at the other end of the office would shout test b t.
That that
And that I'd be like, right. I'm here for another 6 months. Yeah. Yes. Can't let that go.
Were they manipulating you a bit? Probably. Probably. Yeah.
Although I continue to manipulate them by getting to invent more things using the same pattern.
Oh, I see. So you were responsible for perpetuating Yeah. Absolutely. The pattern.
Absolutely.
And now I'm saying perpetuating the pattern. I'm hoping that the The popping. The pop filter provision is potentially perfect.
So I have 2 announcements.
Announcements?
Announcements.
So in the, chapter markers for the last episode, there were 2 relating to announcements. One is forgotten announcements and the other one is remembered announcements. And that was your previous announcement. So announce away.
These are remembered announcements.
Yeah. I
I don't know the ones I've forgotten. So
I've forgotten about those.
Yeah. So I finished the 30 days of AI in testing. I did all 30 days of exercises.
And how has that affected you?
So I think my general sort of understanding of generative AI, prompt engineering, the different tools that are out there, things like hallucination and various other concepts, have come along quite a long way. Before, I was kinda talking about them in the very general sense, but now I think got a few more specifics.
Yeah. Yeah.
In total, I wrote 15,000 words on the subject in the 30 days.
Oh, wow. So you could have published them? So they're published on my blog, which I'm sure we can pop a link into the show notes. What? A link to your blog?
My blog.
Your personal blog of personalness.
Well, it's my company blog.
Oh, fair enough.
So Of company blog. Which is mainly me. But so, you know, the circle is complete. Yeah. So it was when I looked at it in the in the round afterwards, it was like, I've learned a pretty amazing amount.
That I mean, that's a lot of words.
It reminded me of the, was it the NaNoWriMo? That where you write a
Oh, yes. Yes.
You try and write a
a 50,000 word book in a month.
Yeah. So I didn't get close to 50, but I was still pretty pleased with 15.
Yeah. I mean, that's that's a hefty output. What will people gain by by going and reading your blog? What will they come away with?
Well, hopefully, I won't my my overriding feeling afterwards was rather than using AI to try and squeeze more efficiency out of what we currently do in the testing in the testing world, we could use it to do something slightly different and maybe show what really great testing looks like. A bit and using AI to do the things that it's good at, the more structured work, like, helping us to automate as we explore. Mhmm. And sort of tying those together rather than saying, well, we're just gonna use AI to generate all of our test ideas.
Yeah. And Replace all of our testers.
Yeah. Yeah. So, hopefully, it left me feeling quite hopeful that there might be other ways to do this rather than just saying, well, just generate me a set of tests that I don't understand, and and then say that it's tested when we know when we all know it's not. And then following on from, speaking about TikTok
Oh, yeah.
Last time and that LinkedIn is my only, social media remaining out of all the social media giants that are out there.
In fact, I did did write something in the show notes last time about we would include our social media links or in Ash's case, link. Are you going to tell me that I'm already wrong?
Well, so LinkedIn has announced that it's going to start doing shorts, YouTube Shorts. Oh, like Reels.
Like YouTube...
Vines.
Sorry. I thought you meant, like, American underpants.
American
Complete pivot from
So, I'm I'm feel I feel slightly torn now because, you know, the thing that I'm trying to escape from is chasing me onto all platforms. The short form video, which I fear so much, And even on LinkedIn. But I feel like LinkedIn will have the worst of them all.
Yeah. People's motivations for posting things on LinkedIn sometimes don't lead to what you really want to spend your time consuming.
Yeah. So I was reading something the other day about LinkedIn. Apparently, there's like a pattern emerging that people will post very controversial things about, say, software development, But then and then if you reply, refuting that, they then give you a very nuanced answer and say, well, it depends, doesn't it, on the content? And to make you look unreasonable, it's a very bizarre I'm not quite sure what you get out of it.
Well, I was gonna say, I've made 14 people look unreasonable this week. Yeah. It's a triumph.
Unless that's what you are. You know? You've got some kind of bonus tied to making people look unreasonable. But I was just like, what a very strange pattern to emerge.
It does sound quite odd. Although, maybe it might, it might we we we'll all be answering posts on Linked in an attempt to appear reasonable. So reasonable that whatever the reply is, it can't possibly seem
to be unreasonable. Just the replies just become meaningless because they're also reasonable.
Yes. Yeah. Well, I'm just gonna what's the blandest possible replies I can make to this post?
Well, you know, nothing means anything, and it all depends on context. And there's so many contexts out there that, you know, you can't pick one thing for anything to mean, can you?
I think you've just proposed a post that could be made in reply to anything on LinkedIn, on the whole of LinkedIn.
Maybe I'll just reply to everything on LinkedIn with that.
Yeah. I wonder if LinkedIn has APIs and you won't even have to do it yourself.
Perhaps I can use an AI to call the API.
Go on. It's a trap. It's a trap.
It's always a trap. So I'm not gonna say it.
No. No.
So they were my 2, announcements.
Well, so your second announcement to summarize is you're considering getting rid of your last remaining server.
Yeah. Very much.
Very much. And after I found you a Mastodon server that would be perfect. And it even has famous testing people on it. Famous testing people. It's it's very nerdy. You should you should you should give it another look.
There's a very gentle peer pressure Yeah. About signing up for a master's on.
Yes. Yes. Very gentle.
I'm just not ready yet. Let's think about some things.
Steady on. You're gonna rush into things after only 10 minutes of wittering. Honestly, Ash. Feel sufficiently warmed up. Yes. I I might have to do a special edit to make that true. There's only 10 minutes of wittering. So, yes, let's let's talk about things given that that is our ostensible reason for being here. So I have a thing, Ash. Excellent.
Let's hear it. So my thing is entitled, I I'm not very good at coming up with narrowly scoped things. So this is again quite a broad area. But, managing IT service providers. Mhmm. And this kind of comes from having been on both sides of the kind of fence. I worked for a very large IT service provider organisation for a very long time. And then more recently, in my role that I play with some companies as what the cool kids call a fractional CTO. But what that really means is Oh, yeah. Part time.
But fractional apparently sounds more cool. So that's what we need to say. I've seen
that rather a lot as well. Yeah.
The fractional
thing. I'm gonna be fractional, I think.
Yeah. I mean, the other one is it's very close to being fictional.
Fictional.
I quite like the idea of a fictional CTO.
That sounds good. I've met Like David Huffs. Sorry? I've met a few of those.
Yeah. Yeah.
Yes. Science science fictional CTOs.
But in that role, often the things I'm doing is trying to help companies without sort of much technical capability manage agencies or other people who they're giving quite a lot of money to Yeah. To do do development. And I've seen that work really well. I've also seen it go quite wrong. So I just thought it might be interesting for us to talk about.
Yeah. Yeah. I guess I've probably seen it slightly differently, as in being on within, like, the development team, if you like. Yeah. And seen lots of different scenarios with different providers where you've got either a single consultancy providing a team to build a thing or multiple consultancies on the same team to try and build a thing.
Oh, that's
not And
that's Yeah. That's that's that's not pretty.
And then, like, mixes of, permanent people. Well, we're all permanent people.
Wish we were.
Yeah.
We're all temporary people. Temporary people. Ultimately, actually, essentially. So
a a mix of people who work for the host organization, and then people from consultancies and contractors and things like that. So I've worked for both consultancies and been an independent contractor coming in off my own volition and through other consultancies as well. So I've seen a few patterns, I think.
Yeah. I think, so we should be able to find something to say about this. I'm I'm guessing from the 12 paragraphs you immediately put in the show notes as soon as the topic was put there. I mean, my experience from the other side has generally been as part of large teams fielded by a big company. Yep.
And generally, if there are other people involved, it's quite often you might have the old independent contractor filling a a gap that the big company can't source, or you have big companies that own different things. So I remember in one scenario, my company was delivering software and then but they were delivering it to another company's infrastructure Oh. Which was quite interesting. And it ended up that the, and and of course, infrastructure delivery is, is always interesting. It's been a bit, reformed now by the cloud.
Yeah. But it used to be, you used to see people printing money doing doing infrastructure because they would sign a 10 year deal. And in the last 3 years of the deal, it'd be just printing money because they were charging 10 years ago prices per gigabyte for data storage and things like Yeah. Things like that. Yeah.
Yeah. Which they were entitled to do because of the contracts that, you know, were front loaded with with work, so they would lose money in the beginning. But, yeah, it it could get that kind of thing can get messy, but I'm quite interested in what you're saying about these kind of mixed teams because any consultancy is kind of doing the work but also trying to sell more work at the same time.
Yeah. And
if there's someone else doing the same thing at at cross purposes in the same team, that must be interesting.
Yeah. The whole selling of further work, it never really sat all that well with me when I was in a a consultancy. I'd rather just do the work well, and then that would generate further opportunities. I was never really, you know, in that sense of a no. I never really saw it as sales.
It's just excel at what you do and deliver well, and then hopefully more work will come. But other consultancies don't subscribe to that philosophy, do they? They say, well, I need to there was one particular place where I was at where, another consultancy's, sort of salesperson stuck very close to the head of development for the for the host organization. Yes. They were in every meeting that they were in, they were always sat next to them.
Yeah. And that's another
10 people for this. They need another tester. You need this. You need that. So that's, like, the the flip side to just getting on with the work and delivering it.
Yeah. It's like nuclear deterrent, isn't it? One side has one, the other side needs to have one as well.
Yeah. Yeah. So I always found it, say, I always found it really uncomfortable. I went down to a, yeah, the the main trade was in book publishing, which they were great at, but they were absolutely terrible at software. And they had lots of consultancies in there, and then those consultancies were not doing a very good job of testing their own work.
So they they brought the testing consultancy that I was working on working for at the time in to help. And so I kinda sat there and thought, well, I know what the testing consultancy I was working for wanted me to do. They wanted me to say, you need lots of testers from our consultancy, and they need to go on to all these teams, or better still, a separate team somewhere else with lots of people on it, writing lots of test automation and doing performance testing, loads of things you don't need in order to, you know, to to get the work. But to my mind, that was already what was happening with the other consultancies. They were already behaving like that, just, you know, focused on the particular products that they were developing.
So I was like, I don't understand how having another consultancy will solve your too many consultancies problem.
And they
were like, I didn't expect you to say that. I was like, no. You probably didn't, but I'm just not sure.
I worked for some people who would expect me to say Yeah. Pretty much.
But these other consultancies had already agreed to do all this work. Yeah. So I was like, well, can't you just, like, ask them to do what they said they were gonna
do? I did.
So so I I just I wasn't the person for that job, I don't think.
It's the whole thing though of competition between sort of and and marketplaces and all this kind of stuff where competition is not inherently a bad thing. But it it's you you have to it needs a lot of regulation to avoid being people can't see a system without gaming it, can they?
Yeah. But it wraps around to what, even if it's in a big organization, it's like what you were saying about being as a freelancer going into smaller places to help them deal with agencies. It's kind of like the same problem. It's because the host organization doesn't know what to do, so they bring someone in to help them find out what to do. But especially if you come from a bigger consultancy, you don't have the same aims as it's not necessarily only about helping the the organization who's brought you in to deliver what they need to deliver because you need to make some money as well.
So I don't know. Maybe there's freedom from that if you're an individual freelancer like yourself because you can literally just come in and do and recommend something that's a bit more tailored to what they need. And if you get, like, a few months worth of work out of it, you'd be delighted.
Well, yeah. One one interesting thing is this idea of the two extremes of this art, the throwing it over the wall and waiting and then something comes back approach. And then there's the I'm not
familiar with that approach as a test
No. Of course not. You would never have seen that. I'm sure other parts of the life cycle are much more much more likely to have seen that. And then the other end of the scale is the team is completely integrated.
And, you know, you you you are on Slack together or or you're, you know, you you can see that the the the the team's processes and and participate in them. I I think that and actually, I personally much prefer that second way of doing it. Yeah. Because it's reducing the length of feedback loops. It's making everything, but the improving the clarity between the shared clarity between everyone. But you still see the first one sometimes.
Yeah. Yeah. So one of the the contracts I was on, we went to this this organization and they made us wear, like, different colored lanyards. And then another place wouldn't give us passes, so you had to ask to go to the toilet. Oh. Yeah. And it's like, what this this, like,
this This is how much we trust you.
Yeah. This sort of strange hostility towards any sort of separate entity that that they had invited in, but I guess not everybody had invited us in. So that kind of thing doesn't make you feel like it doesn't, like, leave you with warm feelings towards those organizations. Whereas other places, one certain particular place that I worked in Leeds, you were like you're part of a team. You know, no particular restrictions on They didn't have, like, sort of divisions by, you know, some of the testers would have would be consultants and some would be permanent peep permanent, members of staff.
And it was just like it was nice. And everyone you felt like a big team and stayed there for quite a long time and always remember it fondly. So having that sort of integrated nature really, really helps.
And would you say I think I already know the answer to this, but would you say it helps the results or the outcomes be better?
Yes. I think so. I think so. I think if it's hard to ask consultancy contractors to buy in to what an organization might be doing because of that slight conflict of interest, that we talked about in terms of getting more sales. Yeah.
Getting a contract extension, whatever it is that you're after. So but to get them to truly buy into, like, what you're trying to deliver Yeah. Is is quite difficult. So bringing them in a bit more, I think that really, really helps. Because when you ask them to make the leap, you know, to to to to deliver the thing that you really, really want or hit the goal that the organization's after, they already feel more affinity with the organization. Yeah. So it's easier to do.
No. I think that no. I mean, I I I think, actually, people work better when they're happy. Yeah. And so not having to deal with the stress.
I mean, imagine I mean, that story working somewhere where they wouldn't give you a pass, and you just think you're paying me however many, £100 a day or whatever Yeah. To be here, and yet you you think it's okay for me to waste time Yeah. In trying to that that lack of that level of trust, I mean, you could almost argue if you can't trust your supplier that much, then you should get a supplier that you can trust.
Yeah. I was very much, why am I here? In that in that particular example, I didn't really do anything either Yeah. Because I couldn't access anything. And people were quite relatively hostile.
So I was just like I remember, a few a few days into it, I spoke to my then line manager. I was like, I don't wanna do this. I'm not really doing anything. I'm just turning up in an office and not being welcome and just sitting there all day and then being sort of fobbed off, you know, when I go to speak to someone. And it's just like, what a waste of time.
Yeah. Yeah. It's that's exactly what it is.
Yeah.
So one final thing that I was quite interested in with this is kind of this idea of functional consultancies. In other words, they only do one thing. So we've talked a bit about test testing consultancies. Yeah. Is it does that still make sense?
I think they're going out with fashion a bit more, I think. More like consultancies are offering development teams with with the whole sort of suite of skills that you need in order to deliver. I think it's quite rare that you get actual specialist consultancies. Maybe the specialisms have drilled down a little bit, so you might get security pen testing specialists. Yeah.
I don't think that sort of consultancy sort of sits in its own niche, if you like. But I think now I think I think some some sort of clients have been more interested in hiring a team to build something from my time at, Equal Experts, and they do the they do that that is their model. So they you get a cross functional team in which has a team lead, some testers, some developers Yeah. Data people, you user research people, whatever it is that, you know, they've decided that that they need, and then you build the thing. And it's a it's a nice way to do it so you kind of miss out all that strange multi consultancy games that get played on development teams sometimes.
And it's a nice you've got like a remit and a team that are well understood, and you can just sort of get on and deliver the thing. And that's that always seems quite nice. So to have a separate sort of functional you know, the testers are from here and the developers are from here and the analysts are from somewhere else, it makes less sense, I think.
Do you think it would still make sense in sort of giant programs? So, I mean, I'm working in very small companies Yeah. Now. So I'm not really seeing this kind of project. But, I mean, I know, you know, people who are using SAFE, for example, to run very big Yeah. Kind of delivery programs. Is that where that fits in now?
Big delivery programs. I mean, there's a problem with those anyway, isn't it? In their bigness.
Well, they I think they fail a disproportionate amount of the time. Yeah. By by some measure. I mean, fail is a is a fairly harsh word, but, you know, failing might mean overrunning or Yeah. Or going over budget
or or the first anticipated goals or whatever it is. Yeah. I don't know. I try and avoid those scenarios, to be honest, in terms of being sort of caught up in a very large program of work. I don't know.
It just doesn't sound like a very healthy place, and I think it probably pushes, especially in terms of service provider, some really bad behaviors because all the, you know, the more people you add to a project just makes it later and later. All those things start to happen, and it all starts to balloon out of control. And before you know it, you've got, you know, 3 or 400 people on a program of work. Yeah. Yeah.
Reminds me of video games. Video games public video video games development houses, they actively celebrate pulling everyone onto a game and then literally like a few months later, they're like, the game is gonna be delayed. And it's like, oh, really?
Oh, yeah. Oh, we're back to magic time again now.
Yeah. Pretty much. It's like, what an amazing revelation you've just come out with. So I I I don't know. I think bigness comes with its own problems, doesn't it? And then if you then engage service providers to deal with the bigness of the thing that you want Yeah. Then it just gets bigger and bigger and just gets completely out of control. Bloated monster of a a program. Yeah. I think so.
I think so. Because it's like the concept of flexing for demand, isn't it? It's like, well, you know, let's have such a consultancy on the side. And then if we decide we wanna do, you know, twice as much things this year, we'll just flex with that. And I'm always a bit like, well, similar in a similar vein to the the bigness thing.
It's like, well, is the problem that you you the pipe is only so so wide and you can't prioritize correctly, so you just try and, like, make the pipe bigger but in a false way?
Well, just by keeping it the same size but increasing the pressure. Yeah. Pretty much. Well, the analogy just keeps giving and giving.
Yeah. So I'm a bit like, whenever I hear that, it's like, well, we've got this consultancy and they, you know, we flex based on need. Because in my experience, most of the time, it's just like you just end up with them there all the time. So I I'm just like, how do you obviously the pipe is is not what you need at the moment, but you don't know how to make it better, and you don't know how to prioritize what you're doing, which let's face it, that's what it always comes back to, isn't
it? Yeah. It does. Yeah.
So I think it's it's it's that is also like a bit of a a red flag for me when it comes to service providers, which is used for flexibility. So it's like you're using it to cover a failing in your the way that you build your product or the way that you, you know, you decide what to work on, And it's just costing you loads of money. So I'm a bit like, it's not the greatest.
I haven't got much to add to that, really. I completely agree with it. Okay. I just just one more thing that's on my mind about this, and that is when do you decide that enough is enough and fire your service provider? Yeah.
Because sometimes you see it just kinda gets worse and worse until the relationship breaks down irretrievably and everyone's reading contracts instead of talking to each other. Yeah. But it'd be nice to sort of be able to realise, oh, okay. Well, let's move on before you kind of got to the the start sort of stage. But it's a at some point, you gotta figure out if it's not working.
Yeah. Yeah. So while I was at the test people, a somewhat rogue element of the test people, let's say. A rogue element. This this would be
a thing. Have to edit the words the test people out of every sentence that you
Well, the test people don't exist anymore. Okay. So we don't we don't need to edit. Yeah.
They can't sue us.
Yeah. So a certain element tried to engage with the customer in slightly different ways. So they said, we'll have a 6 month contract, but it will have a rolling 2 week get out for both sides. Oh. So what will happen is we'll demonstrate the work that's been done every 2 weeks.
And then the customer can either say, that's enough. We're happy now. Or carry on. Or the consultancy could say, we want out because we'd had a few jobs in the past where it was like, you just need to write some tests for this and that turns into
Just.
Yeah. Exactly. And then that turns into days days weeks weeks of virtually everybody in the company trying to test this thing that was just a couple of days work. Oh. See but the the the host the the consultancy didn't want that cause you can't budget with that, can you?
No. If someone says after 2 weeks, it's like from a technologist point of view, the client could have said after 2 weeks, that's perfect. That's all we need. And with that would have been a great 2 weeks worth of consultancy. But it's you can't put that in your forward plan for the next year, can you? There's a definite thing that you're gonna get.
So this this rogue element,
It's quite intriguing, isn't it?
Yes. Without, inquiring too deeply about any particular people that may have may have surprised the rogue element. How long did it go for?
It carried on for quite a long time. Yeah. Thing is it was a valuable element as well.
Yeah. So
that helps. Which allowed it some
roguish tendencies. Some roguish tendencies. Yes. I I when you say, testing people with roguish tendencies, like, I I just feel that that's leading me somewhere. I'm not quite sure where where it can be.
But, yeah. I mean, that seems like, actually, a really a really good way of doing things. And, of course, it sort of avoids the question of when is the testing finished, which I know, you enjoy having a view about because we we we had quite a a reasonable dedicated a reasonable amount of the last episode to to to to that question. But yeah. So that that's a really good, that seems like a really quite nice way of doing it because it just it means that you're basically normalizing asking the question of, is this still valuable?
Yeah. Every you know, in a relatively rapid time frame.
Yeah. So Yeah. Absolutely. I mean, sometimes there's interesting examples where the the host organization will will take, like, the cheapest option as well. So because, you know, you were talking about helping to do diligence for on a for smaller clients on their digital agencies or what whoever they're engaged with.
Mhmm. And maybe they've got a few bits. So I was in one particular place, and they had a a few bids. It was a third party streaming video provider. We were asked to assess them technically.
So I I watched them, I asked for access to their source control and they didn't use it. And then I said, we want you to deploy your services to one of our test environments. So I got them access and said, can I sort of hang out with you while you're doing it? They were just, like, changing things on the fly on in config on this server. It's like, you know, have you not got, like,
a a pipeline for this or anything like that?
And they were like, no, I just need to change this config item. Oh, it still doesn't work. Oh, maybe this one. Maybe this one. I was like, how can you remember what you've changed? How do you know what worked?
And they said, if only there was some technology that could help us do this in a structured and, and managed way, but but it just doesn't exist.
Yeah. Exactly. They were like, yeah. I don't know. We we need a miracle. But they were cheap, so they got the so they got the gig.
And how well did it go?
Terribly. It never worked
No.
Until until and this is also the common thing as well. We had to go and do it. So the so the people who worked at the organization were just like, basically, you can't do it, but you've you've provided the product. So, you know, we put all the
config into source control and built a pipeline for it,
which I'm sure that they went and they would then went and went
and bought a wonderful idea. A jolly good idea.
So there's some terrible examples like that. I don't know. I think perhaps I'm slightly damaged goods when it comes to service providers because I've just seen too much. The horror is Horror is real. Is too deeply embedded.
The carnage.
Yeah. Yeah. But I think that in there, in all the experience, there's there's like a pattern. So I think I wrote in the notes. So I've said, if we could, set a minimum technical competency level which is comparable to the host organization Mhmm. Before heading through the door. Interview any new starters that come along from the consultancy. Yep. Treat them like humans and part of the team, but probably don't put them in long term leadership and strategy positions.
Seems fair.
And then have a plan to be without them soon, whatever soon looks like for you. 6 months, a year, but not 20 years because that's not soon, is it?
It's hard to imagine a perspective from which that is soon.
Yeah. Yeah. So my final, in my in my comment on the show notes, it just says, I just haven't seen too many thoughtful approaches, to be honest. Just endless horror.
This is like Jerry Springer, isn't it? Yeah. Here's my final thought.
Yeah. But if you got lots of different technology people to come and talk about this subject, I think you'd get something quite similar.
Yeah. I think you're right. I mean, I think we seem quite aligned about a lot of this.
Yeah. Yeah. And, also, I'm mindful that you probably got if you got someone who works for a big consultancy with lots of tales of horror on the other side.
Oh, yeah. So Clients are worse.
Yeah. Exactly.
Who needs it, mate? Yeah. Yeah. Hey ho. Anyway, that that was my, as usual, very large and wooly thing. But, I feel like there's some sense been made of some of it. So that's good. Thank you.
Great thing.
Thank you for making sense of of my thing. So, in the last episode, we had an agile interlude between the things.
Oh, yes. Of course.
So I'm just wondering, you know, should I try and to get out of my chair or something just just for laughs?
Maybe you could try and get your pigeon. Put it on your shoulder so I know that you're not an AI.
Well, you could do that, but, I feel that you're biased in this instance because no one except for you has ever given me a sticker of a pigeon.
I mean With my head on it.
With your head on it. I mean, what is the story? Why is there a pigeon with your head on it?
So Gwen diagram likes pigeons. I have a head for dogs.
She likes your head as well. Yeah? Yeah. So These are 2 of my favorite things.
And a chap called Chris Champ said, I've got you some stickers done. And it was a pigeon with my head saying testability. Yeah?
I've got one. I still look bemusedly at it from time to time Yeah. Just to say that this has a story, but I've never before known what the story was.
Yeah. Although someone saw it through, an event a little while ago and then said, I like pigeons as well. So he had to he was an actual, like, pigeon fancier. So he had like racing pigeons and things like that. And he showed me pictures.
So would you say Gwen was not an an actual pigeon fancier?
Well, she doesn't keep pigeons. She just
Observes them.
Observes them. Enjoys them.
Oh, well. She should come and visit this office sometimes because there's no shortage of of pigeons No. To be had here. No. Oh, well, that was better than an agile diversion. Oh, pigeon interlude. So I'm hoping that, after all that, you will have a thing as well. Because we've only been going for 3 quarters an hour and we need more content.
That's true. That's true. I do have a thing.
What is your thing?
I want to talk about AI voice cloning.
Oh, AI again. Fantastic.
Yeah. Well, I think as well, as it matures, or as or it doesn't mature, or society doesn't mature, or everyone's immature, I don't know. I
I yeah.
Or we just build things without much thought about what they actually mean, and then we just unleash them.
So This is interesting. If you bang these atoms together, it makes
a really big a really big bang. Bang the rocks together, people. That's the secret. Yes. So I think it has the potential to be one of those where we have a very, very advanced voice cloning functionality offered, which, has many, many uses. Some of them good, some of them bad. The the technology itself obviously is, is is different to the intent of using it for whatever purposes that you want to use it for. So but the the thing that interested me was that, OpenAI
Very well said.
Very well said. So they've created this new capability, and it's extremely powerful. So you can you can take 15 seconds of someone's voice, and it will create a clone, with the tone inflection, apparently. I'm a tester, so when people say, oh, this thing will do this, I'm always like, will it?
Will it now?
Yeah. Will it now? Will it really? I'll tell you what. I'll tell you if it does that.
Well, we could feed it a good few hours of What A Lot of Things, and then we could just publish or it could just send it away to make episodes on its own. They probably come out more regularly than ours.
Well, yeah. That's true. That's true. We'll be stuck for hours editing.
Yes. Oh, no. Don't don't even joke. Anyway, sorry. Yes. Serious thing.
So 15 seconds of audio from a person and then you can create a clone of their voice, which is incredibly accurate, or so the claim goes. So there's a few examples of this. So the my favorite one was James Earl Jones allowed Disney to clone his voice to use as Darth Vader.
That's saving him a load of inconvenient acting.
Yeah. Exactly.
But presumably, he did that with some sort of contract that every time he did it, he would get money.
Yeah. So he's not paid.
Unlike where, some other, media big media conglomerates are going.
Yeah. Yeah. So I guess that's the, that's that's the worry, isn't it? So you got, like, the positive uses of the technology. Although, you know, masquerade an AI pretending to be Darth Vader is well, I guess it is a positive use if you like that kind of thing.
What kind of thing is that? Like entertainment.
No. If you like Star Wars, it's a positive thing. Although there was another example where, for example, someone had a vascular brain tumor, which made them lose their voice.
Mhmm.
But they still had recordings of them. So they put it through, a voice cloning AI programme. And they got to speak again.
Did you know that your iPhone has that capability? You can record your voice into your iPhone, and it is then able to, not all that well it must be said, say things in your voice and you can then use it for that eventuality.
Yeah. So I didn't know that. I mean, maybe this is like the next evolution of that
Well of
your capability.
I think there's a lot of room for that capability to improve, so I think you yeah. You're probably right.
Yeah. Yeah. And then but then you had on the flip side of that example of something really positive. So, you know, someone can they can they can speak to their their family members again using their own voice, which, you know, I think we can probably agree is a very positive thing. Yeah. But then, it was used to create a a scam phone call using Joe Biden's voice to people.
Yes. I I think we should be specific that this wasn't OpenAI's technology that was used. It was there were quite a few companies that have variations of this kind of thing and some, like, open source stuff as well, I think. But, yeah, I mean, that that usage was clearly I mean, it was I think they were telling Americans not to bother voting.
Yeah. Yeah. So appealing to our natural sense of apathy.
Yes. Yes. Yeah.
So but then so open API with their products. Oh, did it? So then open AI
API with their their AI.
So OpenAI's approach is to release this to a limited amount of people because they're mindful of what the what the impact might be, especially in an election year, which is interesting, isn't it? Because I don't think chat GPT, as in the the generative AI, was it was pretty much just released widely without too much, like, sort of limitations put on it.
I can't I can't remember, back in the beginning of it whether there was a beta. I I certainly wasn't involved in it. Yeah. But with their more recent products, I think mostly for capacity management purposes Yeah. Yeah.
A lot of the newer you know, the, when the GPT 4 turbo API came out Yep. That that was you know, you had to ask to be allowed onto that and then they expanded it out. But I think I'm not sure that they were doing any of that for the purposes of stopping people from from misusing it. I think they were doing that to manage their capacity of their ability to deliver it. So I'm not sure, but maybe this is a a sort of new new one for them.
Yeah. So it feels like a a a a nice change, a sort of welcome change in approach. Because they're they basically argued that you need to give society time to adapt. I mean, I don't know how long society takes to adapt. I'm British, so it takes about a 100 years for us to adapt to anything, doesn't it? Pretty much. 20 mile an hour zone.
Oh.
100 years.
Don't mention the board.
Yeah. Exactly. So I'm I'm not sure that giving society time to adapt well, I suppose it depends on the society, but because there are certain implications. So one of the examples given is that, say, if you if your your bank still offers phone banking Yeah. Then that could be very vulnerable to, voice cloning, whoever offers it. Not necessarily OpenAI, but whoever offers it.
I mean, potentially, although you need to know other things than just sound sounding like you.
Yeah. Yeah. But it opens up because
Well, it would be evidence, wouldn't it, This is your voice saying that you want to do this thing.
Yeah. Yeah. So I think that's a it's an interesting thought to allow society to adapt. I don't know whether or not it'll work in practice. I think adaptation is it will be necessarily non uniform across different bits of society.
But there is that tension, isn't there, between the freedom side of it? So it's like open source software. I mean, we it's impossible to imagine what the world would be like now if we had never had open source software. And for whatever his failings might be, you know, Richard Stallman coming up with the, the GPL license, That's really moved the world forward and, you know, the variations of that. And then you can see that there are plenty of open source large language models like, LAMA 2.
So Meta have been very consistent about open sourcing their stuff, and there are other open source models. But you kind of think how do you manage the tension between allowing that freedom that really helps innovation to happen Yeah. And putting but also giving people tools with which they can do mischief. And I suppose it's the age old question of technology, isn't it? I remember, a friend of mine, I may have mentioned this before, a chap called Alec Muffet, who's been many things, but he's a security researcher and other things like that.
And he wrote in the nineties, he he released an open source programme called Crack, which would take Unix password files and then try it's dam disc to guess all the passwords in. And you could view that as he did as a tool for system administrators to improve the quality of people's passwords. Or
On the other side?
You were arming the hackers and he was I think he he was amusingly roundly condemned in some of the more pompous newspapers in this in this country anyway for it. But there genuinely is great value in in that in that tool at that time. There was great value in it for people who, you know, were running systems, and the people who were trying to break into them already had Yeah. Password guessers. Yeah.
So I wrote my first password guesser when I was, I suppose, 19 on a sudden workstation. It wasn't very hard to do. Yeah. People who were doing trying to do bad things and I wasn't trying to do bad things. I was just interested in learning.
But people trying to do bad things at that time, they already had that. Obviously, if you look at the various countries around the world who want to undermine western democracy, they've already got Well, including western democracies. Yes. Including of ourselves, we seem to be. Yeah. Good point. Well made. But they've got language models. They can use the things that already exist to, to do voice synthesis. Or they could just get a really good impersonator Yeah.
For goodness sake. I mean, you know, that's not there have always been people who've been able to imitate other people's voices. Yeah. Yeah. So I don't know where I'm going with that, except that my instincts with these things is normally to on the side of letting people have it. Yeah. Because then you find out what what it can do, good and bad.
Yeah. Yeah. I think, like I said, having a an early access phase is to to sort of choice. I mean, I guess it depends who you choose to have it. Yes.
Is is, you know, in terms of software in general, is a is a pretty decent practice. But that only gets you certain types of information. You might get some operational information, or you might give it to people that you, you know, that you trust to use it well, and within acceptable sort of moral parameters, if you like. Yeah. But then once you start to open it out a bit more, the sample of people that you've given it to gets much much wider, and you don't know where it's gonna go.
An interesting thing there is is how fast it is. So there's a kind of dream, isn't there, that you can have a real time conversation using voice with an AI? And you can you can talk to chat gpt. You know, there's on the app, on the on the phone apps for it, you can have a voice conversation with it, but there's a lot of waiting while it makes pretty patterns on the screen while it's thinking what to say and then generating the Well,
you know.
The way to say it.
We all need to pause a bit more before we talk, don't we?
Yeah. Apparently, this is the key to stopping saying which I have to edit a reasonable amount of urms out of all of our all of our episodes.
Is it not what you mean?
I haven't quite learned, how to to not say, but, yeah. I I think you're right. We do I think pausing during speech is actually a good thing to be able to do. But there's also, there you go. That was a genuine one.
There's also the idea that if if it's able to generate speech in real time because the if it if it doesn't have to be in real time, then you send it some text and then, you know, 30 seconds later it comes back. Then in that time, a language model can can read it and say and you could say to GPT 4 or something, is this seem to be malicious? Yeah. And, you know, you you could say, well, I'm not doing that because I I think you're being malicious. But if it if it's supposed to work if it's able to work in real time, you couldn't really do that.
Yeah. It would it would be a different you'd have to come up with something different if you wanted to stop people misusing that.
Yeah. Because I think in for this sort of generation of of of voice cloning, the idea is to add, like, a sonic signature as a watermark.
Oh, very clever.
So you can then go and detect whether or not it's, an AI that's generated. But only if the generating program obviously inserts the signature.
Yeah. An open source one could be. Although in yeah.
Yeah. So from probably from the the the bigger players in the space, you'd probably find that this signature would be would be inserted. Whereas if you but I guess it's the same with the the the password cracker, isn't it? It's like, well, if you are going to be nefarious with it, then your tools are probably way ahead Yeah. Of all the other tools out there for doing that, for, you know, for doing whatever nefarious thing that you want to do.
So you've you those those people would have had a password cracker for years years years, which would
And, actually, I think what Alec wanted to do, and I it's a bit dangerous to speak for him obviously, but but was to give everybody the best one Yeah. So that the hackers didn't have a the the the miscreant side of the equation didn't have a better one. Yeah. Yeah. So it was the best one.
And and that's you sometimes hear the sort of Sam Altmans of the world saying stuff like, actually, you know, the way to to manage this is to make sure that the the goodies, not the comedy trio in seventies, but the the the the people on the side of, I'm struggling to to to define what I mean by this now because I think there aren't any bloody goodies. But there there are. There must be. But but the the goodies should always be able to have the best AI. Yeah.
And now I'm just thinking of Graham Garden and Bill Oddy for having the best AI. I should I shouldn't let my brain out, really. But, yeah, you want you want the side of the the the the people who aren't trying to destroy stuff to have the best resources or at least equal Yeah. Resources. And you sort of think, I I don't know, that seems like a principle that might might be useful Yeah. To take into account.
Yeah. So, like, yeah, like you say, having, like, equal tooling, the equal ability to either to combat. So at least you're coming at it from a, from a a a where you can sort of stand on your own 2 feet if you like in terms of Yeah. Yeah. Rather than having like some kind of wild inequality or at least believing that you've got the best tools.
So I think that's one of the things as well with, especially with, like, with security problems in systems is that, you know, you believe because you buy, like, some kind of, you know, tooling from some vendor that they've got it covered, and they don't most of the time. So I think the belief that you have the best tools when the people who in who who you are in opposition with won't even they won't tell you about the tools that they have. They exist in the dark. So it's it's get it's it's relatively naive, so I see what what what mister Moffat was doing there and saying, well, at least Moffat. Yeah.
Moffat is at least the at least the the the playing field is level now. Yeah. Yeah. I don't want to be
An AI hater.
An AI hater. Yeah. I see the value in voice cloning as Darth Vader. And, also, you know, if you've lost your voice for whatever reason or you've not been able to speak for a long time, then you're gonna find the Darth Vader button. I find your lack of faith disturbing. That's uncanny.
Tried to just slide Darth Vader, didn't it?
So I'm trying not to be an AI hater.
I think that's good. I I I also don't want to to be an AI hater, as was probably apparent. Yeah. But it's a difficult it is a difficult and complicated question because it open source language models sort of defeats the idea of only billionaires can have this.
Yeah. Yeah.
And I kind of approve of that. It also puts it in the hands of bad people. Yeah. And I don't like that, but I don't. It seems like it's like free speech, isn't it? If you have free speech, then you have to accept people saying bad things.
Yeah. Yeah. But
that doesn't necessarily mean that free speech is a bad thing.
No. But I think with with with voice cloning, it's like, okay. There's some very positive sides to it. And, yeah, I can imagine a few applications which would be very bad. But OpenAI, with their tool at least, have said we're gonna limit the release of it in this election year, which to me is a fairly responsible thing to do. And so I think we can hopefully take that as a sign of the future for that particular organisation. That they're gonna try and be put responsibility first.
Which we
The amazing things that they do.
Which we can certainly applaud. So what do you think open API will be doing?
Nothing.
No. Well, not nothing.
Nothing at all with voice cloning.
With fair enough. Sorry. I'm going to mock you about that some more probably.
It's okay. I I I will recommend that we edit it out.
Yes. And, of course, I will just blindly take your recommendation.
Well, it's easier that way as well, isn't it?
It's easier for me certainly. Yeah.
Absolutely. So that was my thing.
I I love that thing. I'm very interested in the AI things at the moment Yep. As you can probably tell. And I think that kind of brings out all of those tensions between that freedom and, freedom of access and all that kind of thing with the sort of bad things that that people sometimes want to do. Yeah. So I think, yeah, fantastic. Thank you.
And like you said, some incredible technologies have resulted from open source and and the principles around it. Like, for example, you can get a library which will add any number of zeros to the left side of a number and you don't have to work out how to do it yourself. But I always liked that one because the the maintainer took it down because he got annoyed. And then everyone was like, oh my god. None of our systems work anymore.
Yes. Yes.
How do you add zeros to the left hand side of a number? I don't know.
Yes. Which makes me, now think about the or as I suppose lots of people would like me to call it. Think about the but maybe that can be a thing next time. Yeah. Maybe that can be a whole thing because there's an awful lot of in there.
Definitely.
In fact I might have to bagsy that one.
Not if I put it in first.
Dibs. Oh, how childish we are.
That's the problem, isn't it? It's great isn't it?
I always feel like, when we were on Twitter, we had a a good mainstay joke that we can make about our Twitter handle being what's a lot of thing.
You've got to change the joke.
We always... you don't have to change them. Sometimes it might be recommended. But, now that we don't have a Twitter account (or that joke) anymore, despite the fact I managed to drag it out for one last outing just that just there. But if you want to talk to us about any of the things we talk about, we would love that. Although, I'm gonna say the only way you can do it is by, using our LinkedIn group.
We must have an email as well.
Well, we have an email, but don't think either was retained.
No. I don't look at that. Well, maybe we can somehow have a better email solution.
Well, I'd like to. I'd like to, connect the domain properly.
So maybe for next episode, we can say, contact us at this email address.
We could just say it now because we can then set it up before we publish the activity.
Oh, no. No.
It's the temporal magic of time.
No. We don't do that because then that means that there'll be a rift in the fabric of space time if it doesn't happen.
It's not a deadline, Ash. It's, you know, it's only an email it's only an
email address. So I spend virtually my whole software development life because people always say, well, I'll tell you what, if we tie these two things together, then that'll be amazing, won't it? And then you're like, well, no. Because then we can't release one of them, can we? And people are always like, oh, well, I've made this dependent on this.
I've made this pull request, but it's branched off the other branch. I'm like, don't do that. Because then they have to go together. And then someone says, why can't we release? And it's because you've made things that have to go together. Yes. So don't do that.
Loosely coupled components, kids. Yeah. It's the magical the magical magic.
The magical magic time.
Is your system working? No. There's no magic time. Loosely coupled, highly cohesive components. That's what that's what you need to to build.
And that's what magic time is all about?
No. No. Magic time is about never sleeping until you die. It does get quite magic after the first few days. Yeah. I know. I know. I know. So much. Yes.
Let's finish that.
So That's what you just finished. The email address will be in the show notes. Please email us because we really, really want to hear from you.
Yes. We do. I don't know how we don't know how to end an episode, do we?
No. We just slowly talk and talk and talk until it peters out. Yeah.