Skip to Razor content

Episode 1 Transcript



Callum  0:18  

Today we're talking about data. And we're joined by the illustrious Patrick Murray.

Patrick  0:22  


Callum  0:23  

And the inimitable James Croft. 

James  0:25  


Callum  0:26  

Do you want to start by introducing yourself? Tell us a little bit about what you do here at Razor.

Patrick  0:29  

Yeah, ok cool, I'm Patrick. I'm Head of Strategy at Razor. My job is to make sure all the cool tech that we're making for our clients not just makes sense now, but it also makes sense in three years time when when it's all part of their digital transformation projects and making sure all those bits fit but excited to be in our pretty, pretty cool looking studio in our in our cave, and talking data.

James  0:52  

And I'm James Croft, Head of Innovation of Razor. And I do, I think what sort of Patrick's looking into and sort of their industry problems, and then turning those into tech solutions to solve problems that we have today. And then problems in the future like this little guy up here...

Callum  1:10  

I wasn't gonna mention it, but now you've now brought it to our attention. What is the hat about?

James  1:14  

So this is a knock sensing hardhat. So if we consider if you like, think about a lone worker, you know, they're on site, hit their head, knocked out. All the little gadgets on here, the little raspberry pi on the back, there's a little accelerometer, which is detecting that knock. So when, you know, someone bashes their head, you know, we've got this little red light up here, that's, you know, visual indicator, but send an alert up to the cloud, it's sending that piece of data that at this point in time, this person who's wearing this helmet has knocked themselves unconscious, basically, how

Patrick  1:50  

How many raspberry pies do we have around the office?

James  1:53  

Too many, too many, although we are running short, we are running short. The chip shortage in the world is causing us a bit of a headache, but we do have plenty around. So...

Callum  2:02  

How many is too many raspberry pies?

James  2:04  

There isn't! I, at home, I've at least got three or four just of my own. And that's not including work ones, though.

Callum  2:12  

And so where does that data go? And what do people, what can people get from it in use?

James  2:16  

So I mean, it's sending off to the cloud so that that data, or it's just a small data point, this knock happened at this point in time, this specific helmet. So you know, say, if you had like 20 or 30 helmets, all sending that data up, you can get a bit, you know, get some insights out of that. How many times is Joe Bloggs knocked, knocked himself unconscious?

Patrick  2:36  

I think that's a good example, that I mean, I think it illustrates that you don't need lots and lots of data points to get quite a useful bit of insight from it. Yeah, that's one person knocked on the head. And that can be all, that's all that needs to be. And that could save somebody's life quite easily. So I think that's a good, it's a good illustration of how that's the case.

Callum  2:36  

Absolutely. And is it only in the construction industry to think that this is applicable?

James  3:01  

I mean, the technology it's in that you can apply to literally anything. You know you could have that in like a bicycle helmet, you know, if you're, you know, off in the mountains somewhere, you know, unconscious, send an alert out. Now, there's so many applications for that piece of technology. 

Callum  3:16  

Good. I'd like to think that if you're knocked off your bike, you'd hopefully know about it before?

James  3:21  

I don't know, it depends on how unconscious you are.

Callum  3:25  

This is very true. It's very true. Now, big data. What does it mean? What is it? Why is it better than small data?

Patrick  3:31  

So big data, well is it better than small data, that's probably my question. I think something that data has never been short of is buzzwords. And I don't know if you find the same thing. Big, big data isn't a new thing. It's been around for a long time. I think Big Data is interesting. It's important. It's, in a nutshell, getting value from lots and lots of data. But we're talking terabytes. And it's something that the Netflix's and the Amazons are getting tons of value from. Whenever you go on Netflix and you see a recommendation, here's what to watch next. That is Big Data in action. That's it. That's big data, looking at your activity, comparing it to everybody else who uses Netflix and saying this is where this person fits, this is the kind of thing they like. I think a lot, particularly clients that we work in. It's important not to get hung up on the word big. You asked me where the small data is better or worse. I think it's near, I think it's as important. And, you know, if anyone is listening to this, an important takeaway that I would say is that even a small amount of data as we kind of illustrated with the, with the helmet there, can still be valuable. It doesn't need to be big data. It can be just putting the right amount of insight of information in the right person's hands. And that and that can be valuable and you can start there. Big Data can come later on, but I'd start there.

James  4:56  

And what I'd add to that as well. I think people misinterpret what big data means. I think they think it means a lot of the same data, as you know, say, with that helmet, you know, having millions of that specific kind of data point, but big data is about everything about every piece of data. It's not just one thing. And it's bringing those sources of data together to be able to do something with it.

Callum  5:20  

Interesting, and how can data big or small influence the decision making process in industries?

Patrick  5:26  

Well, I think it's that, I think it's putting the right, it is putting the right data in the right person's hands. And, you know, we talk a lot about if you've seen any of the numerous YouTube videos, starring myself, 

Callum  5:42  

I've seen them all. 

Patrick  5:43  

Yeah. So you'll be as familiar as anybody with the the, what we call the DIKW pyramid, data, information, knowledge, wisdom. Talk about a lot. Without the benefit of having a nice slideshow to take you through. If you want a triangle, can you picture a triangle?

Callum  6:02  

Yeah, I can just about picture a triangle, one of them. 

Patrick  6:04  

Yeah, three points. What it illustrates is that data is always at the bottom, it has a foundation. And it's important, it holds the whole thing up. But what you're interested in is the wisdom that you can get from it. That's the actionable insight. And the way I always like to illustrate it, if you take an example of data, it could be something like red light, if I said the word red light, that could mean anything. It could be that neon sign, it could be it could be anything. If I say if I turn into information by saying something like the red light on Sidney Street in Sheffield, the traffic lights on Sidney Street have turned red. Now, we were all on Sydney Street in Sheffield. But if you've never been there, you could still probably picture the same thing. That becomes knowledge by being in the hands of someone who can take action on it. So a piece of knowledge would be I'm driving on Sidney Street in Sheffield, and the traffic lights turn red. That is now a piece of actionable insight that you can apply the wisdom to of I should put the brakes on. Now that is a pretty brief, silly example. It might not be relevant to industry, but that the concept still holds up. It's about understanding what are those bits of wisdom? What's the bit of actionable insight and understanding the data you need to enable it. And that could be big or small. But it's important to start with that. And then you can, you can go from there. 

James  7:24  

I mean, I think, I think a lot of people try to gather a lot of information, like a lot, well not information, a lot of data, but don't know what to do with it. And even that data that they have might not be relevant to the problems they're trying to solve. And that's one of the problems with big data, or is that, what can we get? Let's get some of that. Let's get some of that. Let's pull it all together and do something with it. Rather than thinking about what they, what they can do with it.

Patrick  7:49  

I've had that conversation. I've met people who've said, Oh, we've done, we've done data, we collect loads of data, but it's not very valuable. So what are you doing with it? Oh, well nothing we just put it in a database. And then you've got to pay someone to look after it, you got to, you know, and then when you finally do come to it, you've neglected it for a year, and it's all in a state and you have to clean it and you've spent a lot of money. And it's before you embark on whatever it is you're doing is thinking, what is the action? What is that one piece of, what, what's that question we want to answer, right? 

Callum  8:16  

Is that the first step then for businesses to understand what they want to do with it, because I know data gets thrown around as a buzzword. We want to do something with data, we're not sure what we want to do. Is it having that actual insight and identifying that first? Starting from top down?

Patrick  8:29  

It's as simple as that. I think, for all the buzzwords, it's a pretty simple concept. It's, get that, record that thing, put it in the right person's hands, make a decision. Why are we saying I'm having these conversations, really simple exercise, you're sat in a meeting. Leave an empty chair in a room. What, if that, if that, if sat in that chair was your data, what questions would you be asking to it? Because that's, that's your clue. That's a clue to the thing that you're trying to understand to solve the problems that you're actually facing. It's not data for data's sake. It's we're sat in here making these decisions every day. And this is having a real impact on what we do. If we, if something or someone can answer this question, we'll make that decision. And that tells you what you should be working on. It's as simple as that.

James  9:13  

And once you know what it is that you're trying, you know those questions that you're trying to answer. That's when you go and get that data, you might already have some of it, but maybe bits of missing data that you need to answer that question. That's when you start making the change, you know, adding sensors to machines or you know, whatever it might be to get, to get that data out.

Callum  9:32  

Things are moving pretty fast with technology at the minute, is it too late now for businesses to get started on their journey?

Patrick  9:39  

God, innovation. What do you think?

James  9:41  

No, no, I wouldn't say so. It's one of those things, it, where I'm gonna go with this is probably a bit like left field but um, adding another buzzword in, AI is like is booming again now we've gotten you know, your emergence of things like ChatGPT. All of those require data. So whether, you know, I'm not saying that, you know, people should go down the route of exploring AI right now. But if that's if that helps you solve the problem you're trying, you're trying to get out the questions from your data, then you still need the data. You can't start from nothing with AI.

Callum  10:19  

You mentioned probably one of the biggest buzzwords, hot topics in data and technology at the minute ChatGPT, Open AI. Should we be worried? 

James  10:29  


Callum  10:29  

Should we be excited? 

James  10:31  

You should be excited. 

Callum  10:34  

Is it going to take your jobs? 

James  10:35  

You definitely should be excited. No, it's meant to be an assistive tool. When I speak to clients, and we're talking about AI I've tried to push that it's an assistive tool. So that the people who are not doing those manual jobs now of you know, trying to work out, you know, predicting when a machine is going to break, you know, it solves that problem so that you can focus on something else, something more important, more valuable. It, I'm not gonna say it can't replace jobs. It can. But the idea is that you use it as an assistive tool, rather than it actually fully replace them.

Patrick  11:09  

Yeah, it's, I think it's an interesting one. I think in, if you look at previous revolutions, I suppose, industrial revolutions, where we've had some kind of automation that has maybe taken some kind of manual job and replaced it, you know, you might have some kind of farming machinery does the work of, you know, 100 people. I think AI is an interesting one. But I don't know what your thoughts on this, James, but it affects a lot of different kinds of jobs. And I agree, I think it is absolutely a it's an assistive tool that should make people better. I think it will be interesting in 20 years time, I don't think it's going to be, okay we don't need this developer now, because we've got this AI tool. But in a company that has 100 developers today, I'd be interested, what you think is, is that 100 developers in 20 years time?

James  12:04  

I think it'd be 100 developers focused on something else. You know, part of the challenge that we have today is that the complexity in building any application is getting harder and harder, which obviously add more complexity. But I think if we can focus that down and use AI to solve those problems, the things that shouldn't be that hard building a website, building an API, so that then these engineers become data scientists and...

Patrick  12:29  

...building websites.

James  12:31  

It shouldn't be hard. And that's the thing, it shouldn't be hard, but it can be.

Patrick  12:35  

We should have a go.

Callum  12:36  

James Croft says your job is safe.

Patrick  12:39  

But I think with developing, you know, building websites probably is quite hard for myself, and I don't know about you Callum. 

Callum  12:47  

I'd give it a go. 

Patrick  12:48  

But that point, I was standing that the hardest part of what, what I, what I feel that we do is understanding a problem. Yeah. It's, it's having the vision to how you get from where you're starting from to that problem being solved. That that kind of problem, I don't think is as well suited to AI. You know, AI can help write some code, because you know, you're working in pretty well defined boundaries with that, you know, an AI can be well trained and looking at a big, a big data set of how that code might be written. But when you're starting to solve problems that require what humans are good at: creativity, more strategic thinking, dealing with uncertainty. That's when you're always gonna need humans. I think, ChatGPT did a really interesting one. And we're getting tonnes and tonnes of people from all kinds of backgrounds, industries, levels of tech maturity,

James  13:42  

We had a client find us via ChatGPT.

Patrick  13:45  

Yeah that was one. They put into ChatGPT was at Sheffield companies who can work with AI?

Callum  13:51  

The best Sheffield company.

Patrick  13:52  

The best Sheffield company. And it put Razor. And then he talks about some interesting challenges. And he's probably the kind of person who, you know, we talk, we talk to all kinds of companies, but he wasn't, you know, talking to him, he wasn't the typical person, you might expect to be asking us about something that is so cutting edge. And it's sort of shows the mainstream appeal that it's had. But ultimately, it's a big statistical language model. I think a lot of people, because you interact with it, and you can, you can have quite, you know when you're conversing with it, you would think it is a person. 

James  14:30  


Patrick  14:30  

Like it's very, it's amazing. I find it amazing. But ultimately, it's a statistical model that is deciding which word comes next in the sentence. Yeah, based on information that already exists.

James  14:43  

It's pre-trained as well. That's one thing around GPT which it, as, as a model, like an AI model you couldn't use for specific scenarios in your own industry. It's good for like human readable content knowledge, but it's not it's not gonna be able to infer that your machine is going to break. It doesn't know that because it doesn't have that information. It's pre trained on the internet of knowledge.

Callum  15:08  

Who trained the model who trained the ChatGPT model?

James  15:11  

So there's a company called Open AI. And they, they basically research all of this, this space, they are engineers, and data scientists who are creating these models. So ChatGPT which I think everybody is more familiar with. There's just a surface layer for a, an AI called GPT. And GPT is the language model. And they're continuously iterating on it, I think just a couple of weeks ago, maybe last month, they introduced GPT four, which is a newer, more advanced model than the previous one. And there's some interesting stuff that they do. I don't want to go too much into it. But some interesting research that they do around the AI, and then, you know, moulding it before they actually put it out there for people to use.

Patrick  15:57  

Yeah, it was interesting. I was reading an article this morning. And it was a newspaper in Ireland which was apologising, because it turned out a piece that had been submitted to them had been generated through ChatGPT. I think it's an interesting question around does that, does that matter? Would you think less of an article you were reading if you knew it was written by a journalist or if it had been written by ChatGPT? Does it matter?

James  16:24  

I think this is where this is an assistive tool and the two come together? You know, because if if you're using the AI tool to just generate content, you're throwing it out there, you're not fact checking it, you don't know if it's correct, you know, everything that everything that it's responding with is just an assumption based on the prompt you give it. And unless you're, you know, you're fully into your data sciences, and you're doing prompt engineering, which is basically, you know, saying the right phrases to get the right answer. What it can generate could be absolute garbage.

Patrick  16:54  

Well, yeah, I mean, don't know if you found that? I find, because I sometimes use it as basically a Google, you know. If I've got a question I want, I just asked you to ChatGPT. And you're not digging through.

Callum  17:07  

I also know that you've used it to create your own sitcoms.

Patrick  17:11  

Own sitcoms, it's not as funny as we are, obviously. You get it... but I find sometimes when it's something when I ask it you about a subject that I do know about, especially when I asked it to give me kind of a long form detailed thing. There are elements where Oh, actually, I don't know if that's questionable. But it says in a very, very convincing way. I'm not actually tried GPT 4. But...

James  17:42  

Not yet. It's on our tech radar.

Callum  17:46  

Think of the sitcoms you could write with GPT,

Patrick  17:48  

Well if this doesn't work out

Callum  17:54  

You mentioned in the chat, data engineers and data scientists can just break down those rules and what they mean and what their responsibilities are.

Patrick  18:00  

Yeah. So I think that often confused, often overlapped roles. I think people hire data professionals in some ways, and I think some ways, hopefully can fulfil both roles. I'll tell you in the most simplest way to be a data engineer is to be a data plumber. It's the person who puts all the pipe work in and does all the work to get data from its raw, unrefined stapes at wherever it was made, into a place where it can be analysed. And we can talk more about those little bits look like but that's what the engineer does. A data scientist is as probably as the name suggests, a scientist, it's someone who asks questions about data and goes in and finds the answer, who forms hypotheses, he does statistical analysis. And it's very much more on that, you know, machine learning, statistical modelling side. We have a question about this data. Let's go find out about it. But I think, there are three roles that are massively in demand at the minute we have our own. But, and ones are often confused, but very, very distinct skill sets.

James  18:00  

Your data scientist is the one that's probably creating your AI models and machine learning models. You know, that's how I identify

Callum  19:22  

And your engineer as your plumber.

Patrick  19:23  

Is your plumber. Doing your data pipes and do this together. 

James  19:29  

I mean, you need that though. You do need, you do need the plumbing, the two are like the complementary roles. You need both, you can't have one without the other.

Patrick  19:35  

Don't get me wrong and again, it kind of goes back to what I said before about how we, you know, the challenges we face when we're creating a piece of software you know, the tech is the tech you can learn about the tech. The hard thing with data engineering is getting the data in a state to answer the right questions and also understand what those questions are. Which leads back to what I said before when you whenever you doing a data transformation or trying to do things with data, there the things you need to be thinking of, it's not just getting you from point A to point B, because that's what you do end up with, well, I loaded data in this idea, and I've got it here. You know what, so what? It's getting it in a form that right now we can create the dashboards that, you know, interrogate that data when we want to, we can ask the questions a bit. So, both are important.

Callum  19:36  

Talking about the applicability for data projects in certain industries, something that comes up a lot, particularly in manufacturing is sustainability. How can data projects be used to support sustainability?

Patrick  20:35  

Well, with with sustainability, you can't improve something, you improve the things that you can measure, and... you've got got to be able to measure what you're improving, to know that the things that you're doing are having an impact, you know, data on its own, won't make you more sustainable. But it can tell you the places that you need to start looking at, and then go and look at those. You know, something, like you sort of alluded to there, particularly the manufacturing industry, particularly with what's happening with energy bills, how expensive it is, you know, we're talking to companies who have had really, it's had a really profound impact on their bottom line, you know, leaving a machine on when it doesn't need to be, leaving a light or air conditioning, when it doesn't need to be. Over the course of a year actually has a pretty profound impact on the bottom line. And by using the right data, it tells you where to start looking, where to take action.

James  21:28  

And you need, I mean, there are two sides to this, I guess on the consuming energy, you need to be able to monitor it, how do you monitor it, you know, there, you could get, you know, the core power that's come in from everything, but that doesn't give you enough information to target specific areas, and then you need to start sensoring up, you know, specific machines and then working out if, if it's that one that's drawing too much in too much power, then there is a flip side of all this that I think about when we want to think about sustainability, and it's actually the software that we build as well, you know, if you flip it on its head, from an AI perspective, if you build an AI model, that is, you know, going to predict how much energy usage and you need a lot of data to do that. So there's a lot of computational power. So then you've got a different machine, but you've got computer, a server, whether it's in the cloud or not, you know, running for X amount of time.

Patrick  22:22  

It's easy to forget, isn't it I think when you work when you work in the cloud, yeah, you're not it's not the days where you've got to, you got to serve in the corner wearing away getting hot, and you go  that's a lot of energy. It does go like out of sight, out of mind that it is your sole responsibility, isn't it to make sure...

James  22:35  

Yeah, I mean the companies like Microsoft are doing things with Azure cloud, you know, that they've got their system sustainability efforts that they're, you know, sort of working towards. But I think it's still on the people who are building the software, to make sure that it's you know, performing. And it has all sorts of things in the background, not just at the at the core in the actual manufacturing,

Callum  22:57  

And are there other different ways that you can visualise that data. And is that important?

Patrick  23:02  

Ah visualise it, it's arguably the most important bit. I think we, at Razor we have that internal capability to do the UX and UI design. And we don't just apply that when we're building a piece of software, or  building a website, if we're creating a Power BI dashboard, it has to work well for the user. Everything eventually has to interact with the human or else you know, there's no point. And so we'll get back to...

James  23:31  

Yeah, I was gonna say.

Patrick  23:32  

Not taking our chances, and until it does. People need to be able to interact with it, understand  it, that it needs to be intuitive, otherwise all, all the stuff all the plumbing, that's come before, that could be perfect. But if it ends up on the dashboard, in an interface that isn't intuitive to use, that people just give up, give up on, they're not gonna take action, and you may as well have not done all that. And I think it's something that is often an afterthought, I think, especially here with, with tools like Power BI or whatever you end up using it in, people can just throw something together because you think Okay, is there, there's your graph, but yeah, it has it has a big impact on the actual impact that thing makes so yeah, we always think about that first thing.

James  24:19  

Imagine going down though, from an innovator you know, little things were running through my mind going down the AI route. I'm not saying GPT could do this, but if you can ask GPT to create your dashboard, you tell it what you want. I want, I want a pie chart that says this and first from your data source you know, imagine if you could do that though, if there's if there was a tool out there.

Patrick  24:41  

Not taking jobs though.

Callum  24:42  

I just love to see the cogs working.

Patrick  24:48  

The heat he's giving off

Callum  24:53  

Terms like data warehouse and data lake, what's that all about?

Patrick  24:58  

Yeah, so both, they are both different things for different needs. An important thing to point out is the difference between structured and unstructured data. Structured is something that I think we're all very used to working with in business. Yes, that's your spreadsheet. That's stuff that can sit in rows and columns, your numbers and whatever. Unstructured is your images. It's your PDF document or basically everything else. And the data lake can do, can do the lot. It has everything in its raw format. And then the data warehouse is much more of that, how can we put structured data in a way that we can query it and set it up for what we are asking. So just to just blow your mind, there is a third option, which is a combination of the two. Can you guess what it's called? Data Data, a data warehouse. A data, a Lake House, the data lake house, is this kind of the new system where we're in between that tries to both leverage that ability to have, you know, all kinds of data in a row format. But then, sitting on top of that is sort of a layer that you can interact with as if it was a data warehouse, if it was structured, so you can kind of you know, change the focus from having that pipeline architecture to kind of having it all in one place. And that's something that we're seeing more and more, particularly when there is a lot of data to deal with. So there's, yeah it depends on the application, but often solutions have a bit of a bit of everything. Going back to that point about data scientists, that lake is mainly where they're working. They want to work with things that are unrefined, because as soon as it's been refined, you don't know what you've lost. You might be asking the question that the data that was gonna answer it got cleaned out, somewhere back over here, you're missing a trend. So all part of data solutions.

Callum  26:48  

If data were a superhero, what would its superpower be?

Patrick  26:52  

If data was a superhero, what superpower would it be? I think probably the, I think the superpower that people often think it has, is that it can predict the future. I think they think if we get a you know, if we throw some AI in there and set up our data warehouse, where I build this thing that we can ask a question, it will just we can start predicting everything's gonna happen three years out and make all the decisions then and which is not what it can do. Nobody can do that. It's about informing your decision making as much as possible. So you can make better decisions, you know, no one's ever always going to make the right decision really, it's can we make better decisions. So if it was a superhero, it would be able to predict the future slightly more often than a regular person, but still not to an extent where you could, you could kind of put your money on it.

Callum  27:49  

It's almost like you've had a chance to prepare for that.

Patrick  27:52  

Yeah, it's like it was a really insightful way of turning that question on its head a little bit.

Callum  27:57  

So should we believe the hype then? Data?

Patrick  28:01  

Should we believe the hype of data? And should we believe the hype data? We should, I think, going back to my big pyramid, if I could, if I could just for a second, go back to my maps of the triangle, triangle, three shapes. We should, the hype of data, no. Data on its own is not useful. It's an overhead and something that you've got to pay to store and look after. And honestly, it doesn't do anything. The hype of turning that into real insight you can take action on, I believe in hyper that. That hype of trend to elevate people, making people better. I believe the hype of that as well. So I think data no, but the effects when you when used properly. I think there's some there's some hype there to really

Callum  28:54  

It's almost like unrefined oil. 

Patrick  28:56  

Yeah. Data... Yeah, you know the phrase data is the new oil. The guy came up with that is it Sheffield University 

Callum  29:03  

Is that right? 

Patrick  29:04  

Just off the road. So little fact for you, little soundbite.

Callum  29:09  

Sheffield home of data. Well, Patrick James, thank you very much for your time.

Patrick  29:16  

Thank you for having me.

Callum  29:17  

It's been the most insightful and I hope you enjoyed listening and watching. Until next time, goodbye.