BFM 8.9v Business Station. My name is Rich Bredry. Of course, this is Matt Splein. So we are back with our very own chatty man, Matt Armitage to explore the world of bots and blood tests today.
No, I have no idea what that means either. Hey Rich, well partly this is a follow up from the late September story that OpenAI would be introducing full voice capabilities with ChatGPT. That's the bot part, we'll get to the blood part later, you know, always say the blood to later. But anyway, the bot part, so this introduction of Chat is similar to the way that we interact with, you know, assistants like Serial X or Google Assistant, you know, all of those bots that we're already using.
And supposedly these updates were part of a suite of new features that would see ChatGPT becoming a multi-modal model. I hate times like that, but anyway, what it basically means is it can interact with text, voice and images, which again, we've both been playing with this week. And as is usual with these things, you know, I read the story in September, I think maybe we mentioned it on the show. And then of course, I'd promptly forgot about it until my own not so digital assistant reminded me to check in again.
And if you're wondering, the machine with the unenviable task of looking after my brain is actually of course Richard. So it was Richard who reminded me last week that the voice features were now live and available in the app. Yes, they were. But although they were not showing up for some people, and it turns out that for them, for it to show up in your app, you had to delete the app that was installed on your device and then reinstall it.
And then it would show you the voice chat thing. Yeah, exactly. And that's what I did. And it showed up.
And I'm kind of having the same issue with Metas new chat assistants as well, because that's not showing up in my app. And I'm wondering whether I have to delete and I really can't be bothered with Facebook apps because there's so much data or anything. I'm not going to do that anyway. So yeah, if the features aren't showing on your version of chat GPT, the app rather than the desktop version, as Richard said, delete it, download it again and log back in and it should appear and ready for you to chat.
So the new voice feature allows you to choose from one of five personas. Again, you know, like you can choose different voices for Siri and then you can have a pretty good natural conversation with it. And in terms of the the photo options, well, that's a bit more straightforward. You upload a photo and the bot then gives you feedback on what you've uploaded.
Yeah. Now, before we go into those functionality aspects, let's look at the wider landscape. And this thing I've been hearing a little bit about is interesting AI tools like chat GPT declining. Yeah, I mean, I read somewhere, I think it was on the Economist that said that consumer searches for AI have declined recently.
So I think we are seeing some kind of change. And we have to remember though that there are different sections to this. There is the business use case for AI. And of course, there's the consumer use case.
So these declines in searches that could indicate a number of things, that could indicate that people have played with these things and they're just kind of over it. Or that, you know, they've looked at the tools and realized that in most cases, you know, these are paywall products and any free versions only give access to, you know, very basic versions of the technology, which, you know, generally aren't very good. Or, you know, they've actually subscribed and they're using the tools so they don't need to search for them anymore. Or perhaps, you know, they've just gone the easy route and they've downloaded the Bing app.
And that means they've got access to chat GPT and the value image generation for free through the app that Microsoft is providing. So I don't necessarily think it's an indication that people are less interested. I think, yes, the novelty is wearing off, but in much the same way as it did with services like Siri and Alexa. You know, we're now getting to that ubiquity phase where we kind of take them for granted and use them for all the kind of mundane things that they were originally designed for.
Like timers and meetings? Well, to an extent, I mean, I think these more advanced tools, not necessarily such a mundane task, but to be honest, you know, the number of times I set timers with Siri and Alexa is quite so much. I mean, it might be mundane, but it's still really useful. There's a whole kind of world of AI assisted tools that people are integrating into their daily lives.
I mean, there's a whole bunch of them that I've integrated. I mean, I'm using chat GPT, reword, Grammarly, mid-journey, Feedly. I'm using the AI features and Adobe podcast. You know, there will probably be a bunch more that I'm not even thinking about.
And I'm sure you've got a whole kind of, you know, bunch of them as well. For sure. I'm using a whole bunch of the ones that you're using as well. I'm even considering subscribing to the whole Adobe Creative Suite because there's stuff on there that I find interesting that may in fact help me with my workflow moving forward.
You know, and the thing is though, it's every time I look at for this kind of news, there's always something that's peaking my interest. Yeah, not just peaking your interest, but also pulling things out of your wallet, which is probably the most important thing. But that is, you know, that's kind of the point. They've become very routine and very, you know, easily usable for quite a large section of people.
I think there is that other group of people who are maybe a bit disappointed with the tools, because they've listened to all the media hype. They've tried out some of the new toys. They've found them entertaining for a couple of hours, but they don't really see how they can integrate them or use them in their wider lives. Or, you know, they simply don't understand enough about what's going on under the hood to apply them to their lives.
So there's still a lot of misunderstanding around artificial intelligence and machine learning. There's still this trend to anthropomorphize the tools. I hate having to say that word. I nearly got it through it.
I got it through it. Okay, today. But I think these new kind of adding adding that voice element, adding that ability to speak to it naturally, this could be the functionality that kind of tips the, you know, the naysayers over the edge. But I'll go back to that photo upload thing first, partly because I didn't really think it would be particularly useful.
Yeah. And so this is when you upload an image to the bot and it comments on it, right? Yeah. So I started with the image that I created for last week's show, which actually used Dali to create in the first case.
It's a pastiche of a Victorian shopping arcade. So it analyzed the image. It broke down the elements. It indicated what era the image was from, the Victorian era based on the shop brands and the clothes that people were wearing.
It commented, for example, on the color treatment of the image, saying it was, you know, it's a sepia image. So it could be from that past time. Now, could it tell that it had been created by its own companion AI? Well, that was an interesting thing.
No. I mean, it gave me a great checklist of things to look for in images that suggest whether they're real or whether they've been, you know, confected by AI. But it couldn't tell with any certainty that the image was no more vintage than its own data pool, which kind of goes back to the point that we've made a few times, you can use AI to trick other AI in this case, essentially using the same AI to trick itself. The tools that help to check for machine generated text, for example, can be fooled by simply upscaling the text in another AI and pasting it back into the one that said it was fake to begin with.
So I tried it with another few images, real images this time, although I guess the idea of real photographs is a bit strange as well, because they're just recordings. But anyway, I put in a garden scene, and it was actually pretty good at identifying the plants. It correctly guessed that it was a tropical garden. And I was impressed, because when I've tried reverse engineering photos in search engines, it's always been, you know, really hit and miss.
I mean, one of the famous examples, early examples of that was me uploading my own face to a very early version of Google's image search, and it came back with a basket of kittens. So that was one that was always made up, we were very happy. But you know, and it even noted that there was a cat in the picture that I uploaded, and it made a point of saying that it's a cat and not a plant. Well, I mean, that's a good start, I guess.
But looking forward, I mean, we've seen examples, but how do you see this type of tool being used? Well, in terms of the pictures, I mean, obviously, we know Chachie PT, it's not live linked yet. Yes, when you use it through the Bing app, it can access news stories, but it's still a trained closed model. I think its data goes up to mid 2022 or something like that.
So its knowledge is limited in that sense of, you know, what's going on in the real world. For example, it easily identified the Eiffel Tower, but it couldn't place the Palace of Justice here in Putra Jaya. Although it, you know, it correctly flagged that it's an example of Islamic architecture. It was happy to tell me about Islamic architecture in general, and various motifs that were found in the building.
So, you know, there are definitely educational purposes for this. You can use it, you know, for arts and culture science. Apparently, it can give you fashion tips. I don't really know how that works.
Although it won't comment on people. So it won't tell you that, you know, somebody's ugly or badly dressed or anything like that. You can use it for visual storytelling. So you can upload the picture and it will tell you a story about or come up with a story about that building.
So I asked it to create a story about the, again, the Palace of Justice image. And it wrote me a short story about the Palace of Whispers and a centuries old Sultan called Rafik. I mean, it was a bit generic, but you know, it was quite charming. So actually, the image search function is a lot more useful than I thought it was going to be.
But as I said before, you know, that that killer application is the voice interaction. And again, you know, you would want to turn me on to it. So how have you found experimenting with voice conversations with it, especially in terms of, you know, the humanness inverted comments of the responses? It actually shocked me as to how realistic it sounded with, because there's human-like pauses, there are arms, there are arms.
And I think it genuinely shocked me when I used it for the first time that I ended up having, and I say this in air quotes, a conversation with the bot to the point where I think I was conversing for about 45 minutes about, you know, my hometown and the surrounding areas and where I should visit and whatnot, stuff I already knew. But it was interesting to hear it from a bot telling me where I should go on. And it's probably sourcing all of this data from Wikipedia, stuff that I already know. But it was as if it was coming from a real person genuinely.
I know, and it's quite freaky. I mean, I haven't managed a 45-minute conversation, but then I don't manage 45-minute conversations with people. This is probably the longest I talked to anyone in any given week. But I mean, just sort of backtracking a bit.
So I mentioned there are five personas. So far, I've only checked out Sky. I don't know if you've checked out any of the others yet. I have.
Yep. I landed on Sky though. She reminds me of the voice from her, which is potentially why I stuck with it. Ah, is that why it's chimed a bell in my head?
Yeah. So that's probably why I haven't moved on. And I've found it very impressive as well. When we talk to assistants like Alexa, those are a very limited step towards this spring-free world, this world where we speak to all our devices rather than interact with them through a control device.
And you can't compare that kind of interaction talking to Siri or Alexa with that experience with chat, GPT. One of the recurring things on the show has been voice and the advance of large language models using natural language processing. And it's, you know, as my conversations were short of yours, but it's amazing how your interactions with the machine change once you're not typing anymore. I found that asking questions was much easier and more natural than when I try and type them.
And I found or I am finding that it's making the process of researching and searching for answers to things a lot simpler. Because, you know, you're not thinking about how to phrase a typed sentence for a machine. You know, already that act, that translation of getting it out of your brain and typed onto a screen, it slows you down. Your thoughts happen so much faster.
And you don't have to go through this process of translating your thoughts, typing them and adding this layer of trying to phrase them in a way that you think that the machine is going to understand. Isn't there still an issue when you're speaking? Yeah, of course. I mean, the machine still misunderstands things, but it's so much easier and less frustrating for you as a user to just change the way that you phrase something, to try another verbal command, than it is to look at your sentence and think, oh, how am I going to change that to get it to do what I want it to do?
And going back to that part I mentioned about the people who try out these tools and don't see them as anything but a novelty. I think this is where it will grab those people, because it is much more like having a normal conversation. And I'm finding that it's changing the way I think about asking questions, because I can go off on tangents and come back. I'm not, you know, working through a checklist of questions.
As you said, your conversation is just taking you down strange parts, it's giving you information, you're able to riff on the information and you go off in all these different directions. And, you know, it really, I think, matches my kind of tangential learning, which I find, you know, a little bit frustrating with the kind of linear approach of search engines. So during your trials with the voice, what do you think were some of the limitations that were most apparent to you? Well, I mean, just to go back to you mentioning her, you know, overall, it is very kind of sci-fi.
It's like speaking to the ship's computer in the, in the start, right? The conversations themselves are very easy. And as you said, there's those natural elements of the speech, the pauses, the arms and arms. So they don't have that robotic element that you get when you speak with a lot of digital assistants.
You can still tell, you know, it's not, it's not human. I don't think it's trying to be human, but it is normal. So in terms of limitations, the verbal misunderstandings, which of course is to be expected, you know, these machines will get better at context over time. But they're never going to be perfect because people aren't perfect at context either.
We'll get things wrong when we're talking to each other and misunderstanding each other. I guess the thing I notice most is the speed of the responses, not in terms of the gaps between you asking the question and it formulating the replies, but in actually getting through the responses, because, you know, we used to just scanning through things and speech is a little bit slower than reading. So it was surprising when I've listened to an answer that would feel like it took, you know, a minute or two minutes or whatever. And then you look back at the text and you see how short the text is.
You're quite surprised because it kind of feels like an eternity when it's being read loud. Yeah. Yeah. Now, how are businesses likely to incorporate these advances do you think?
Don't ask me. I don't do business stuff. No, I'm just messing. I know that.
I mean, there are lots of applications, things like internal chat bots, you know, for company use. So imagine a machine that's loaded with with schematics that's actually conversing with an engineer who's out in the field in real time, you know, the engineers saying, I can see the blue wire, but not the red wire. You know, when you've got your head inside a huge piece of factory part machinery, you can't always go and look at a screen, but you can talk and you can listen. So I think there's a, you know, a huge number of applications for companies to use it for their internal processes in terms of, you know, business to business, business to consumer.
Look at me with all the cool words. Yeah. You know, there's there's massive potential as well. You know, going to a brand site, clicking the microphone button to talk to the bot about the goods and services, rather than going through all the process again of typing.
That thing of easily being able to find out what isn't isn't in stock, you know, saying, Oh, is that blue t-shirt in stock and the machine just saying, yeah, you can buy it, just click much easier for you to raise service or other issues that you have with with companies. I mean, I'm not going to go into that side of things today. So depending on how things work out, I want to have a look at the business case for implementing AI and machine learning technology and also because in between recording this and it being broadcast, I'm actually doing a panel on this with the business applications for for BFM's breakaway this week. So hopefully that will feature a bit in next week's episode.
And of course, you know, there are some sectors that are racing ahead with this technology. Others seem to be hedging their bets, but I have to say that I'm really impressed with these new integrations. They do change the game. And once we see them being extended beyond, you know, the premium tier, once we see them coming more into everyday tools, like search engines like being, that's where we're going to see the worlds of search and internet browsing change rapidly.
There, I got through it in just one half of the show. And you know what, it's been fairly sensible so far, which worries me. So let's see how weird things will go after the break. Yeah, folks are tuning to Matt's point here on BFM 89.9, the business station.
BFM 89.9, the business station. My name is Rich Breadry. Of course, this is Matt's plane. So we are back with our very own chatty man, Matt Armitage to explore the world of bots and blood tests today.
No, I have no idea what that means either. Hey, Rich. Well, partly this is a follow up from the late September story that open AI would be introducing full voice capabilities with chat GPT. That's the bot part.
We'll get to the blood part later. You know, we save the blood to later. But anyway, the bot part, so this introduction of chat is similar to the way that we interact with, you know, assistants like serial X or Google assistant, you know, all of those bots that we're already using. And supposedly, these updates were part of a suite of new features that would see chat GPT becoming a multi-modal model.
I hate times like that. But anyway, what it basically means is it can interact with text, voice and images, which again, we've both been playing with this week. And as is usual with these things, you know, I read the story in September, I think maybe we mentioned it on the show. And then of course, I promptly forgot about it until, yeah, until my own not so digital assistant reminded me to check in again.
And if you're wondering, the machine with the unenviable task of looking off my brain is actually, of course, Richard. So it was Richard who reminded me last week that the voice speeches were now live and available in the app. Yes, they were. But although they were not showing up for some people, and it turns out that for them, for it to show up in your app, you had to delete the app that was installed on your device and then reinstall it.
And then it would show you the voice chat thing. Yeah, exactly. And that's what I did. And I'm kind of having the same issue with Metas new chat assistants as well, because that's not showing up in my app.
And I'm wondering whether I have to delete. And I really can't be bothered with Facebook apps because of so much data or any phone. I'm not going to do that anyway. So yeah, if the features aren't showing on your version of chat GPT, the app rather than the desktop version, as Richard said, delete it, download it again and log back in.
And it should appear and ready for you to chat. So the new voice feature allows you to choose from one of five personas. Again, you know, like you can choose different voices for Siri. And then you can have a pretty good natural conversation with it.
And in terms of the photo options, well, that's a bit more straightforward. You upload a photo and the bot then gives you feedback on what you've uploaded. Yeah. Now, before we go into those functionality aspects, let's look at the wider landscape.
And this, I think I've been hearing a little bit about is interesting AI tools like chat GPT declining. Yeah, I mean, I read somewhere, I think it was on the Economist that said that consumer searches for AI have declined recently. So I think we are seeing some kind of change. And we have to remember, though, that there are different sections to this.
There is the business use case for AI. And of course, there's consumer use case. So these declines in searches that could indicate a number of things. It could indicate that, you know, people are played with these things and they're just kind of over it.
Or that, you know, they've looked at the tools and realized that in most cases, you know, these are paywall products and any free versions only give access to, you know, very basic versions of the technology, which, you know, generally aren't very good. Or, you know, they've actually subscribed and they're using the tools so they don't need to search for them anymore. Or perhaps, you know, they've just gone the easy route and they downloaded the Bing app. And that means they've got access to chat GPT and the Valley image generation for free through the app that Microsoft is providing.
So I don't necessarily think it's an indication that people are less interested. I think, yes, the novelty is wearing off, but in much the same ways it did with services like Siri and Alexa, you know, we're now getting to that ubiquity phase where we kind of take them for granted and use them for all the kind of mundane things that they were originally designed for. Like timers and meetings? Well, to an extent, I mean, I think these more advanced tools, not necessarily such a mundane task.
But to be honest, you know, the number of times I set timers with Siri and Alexa is quite a bunch of, I mean, it might be mundane, but it's still really useful. Yeah. There's a whole kind of world of AI assisted tools that people are integrating into their daily lives. I mean, there's a whole bunch of them that I've integrated.
I mean, I'm using chat GPT, reword, Grammarly, mid-journey, Feedly. I'm using the AI features in Adobe podcast. You know, there are probably a bunch more that I'm not even thinking about. And I'm sure you've got a whole kind of, you know, bunch of them as well.
For sure. I'm using a whole bunch of the ones that you're using as well. I'm even considering subscribing to the whole Adobe Creative Suite, because there's stuff on there that I find interesting that may in fact help me with my workflow moving forward. You know, and the thing is, though, is every time I look at for this kind of news, there's always something that's peaking my interest.
Yeah, not just peaking your interest, but also pulling things out of your wallet, which is probably the most important thing. But that is, you know, that's kind of the point. They've become very routine and very easily usable for quite a large section of people. I think there is that other group of people who are maybe a bit disappointed with the tools, because I've listened to all the media hype.
They've tried out some of the new toys. They've found them entertaining for a couple of hours, but they don't really see how they can integrate them or use them in their wider lives. Or, you know, they simply don't understand enough about what's going on under the hood to apply them to their lives. So there's still a lot of misunderstanding around artificial intelligence and machine learning.
There's still this trend to anthropomorphize the tools. I hate having to say that word. I got it through it. Okay, today.
But I think these new kind of adding that voice element, adding that ability to speak to it naturally, this could be the functionality that kind of tips the, you know, the naysayers over the edge. But I'll go back to that photo upload thing first, partly because I didn't really think it would be particularly useful. Yeah. And so this is when you upload an image to the bot and it comments on it, right?
Yeah. So I started with the image that I created for last week's show, which actually used Dali to create in the first case. It's a pastiche of a Victorian shopping arcade. So it analyzed the image.
It broke down the elements. It indicated what era the image was from, the Victorian era, based on the shop fronts and the clothes that people were wearing. It commented, for example, on the color treatment of the image saying it was, you know, it's a sepia image. So it could be from that past time.
Now, could it tell that it had been created by its own companion AI? Well, that was the interesting thing. No. I mean, it gave me a great checklist of things to look for in images that suggest whether they're real or whether they've been, you know, confected by AI.
But it couldn't tell with any certainty that the image was no more vintage than its own data, which kind of goes back to the point that we've made a few times, you can use AI to trick other AI in this case, essentially using the same AI to trick itself. The tools that help to check for machine generated text, for example, can be fooled by simply upscaling the text in another AI and pasting it back into the one that said it was fake to begin with. So I tried it with another few images, real images this time, although I guess the idea of real photographs is a bit strange as well, because they're just recordings. But anyway, I put in a garden scene and it was actually pretty good at identifying the plants.
It correctly guessed that it was a tropical garden. And I was impressed because when I've tried reverse engineering photos in search engines, it's always been, you know, really hit and miss. And one of the famous early examples of that was me uploading my own face to a very early version version of Google's image search, and it came back with a basket of kittens. So that was one that was always made up, very happy.
But you know, and it even noted that there was a cat in the picture that I uploaded, and it made a point of saying that it's a cat and not a plant. Well, I mean, that's a good start, I guess. But looking forward, I mean, we've seen examples, but how do you see this type of tool being used? Well, in terms of the pictures, I mean, obviously we know chat GPT, it's not live linked yet.
Yes, when you use it through Bing app, it can access news stories, but it's still a trained closed model. I think its data goes up to mid 2022 or something like that. So its knowledge is limited in that sense of what's going on in the real world. For example, it easily identified the Eiffel Tower, but it couldn't place the Palace of Justice here in Putra Jaya.
Although it correctly flagged that it's an example of Islamic architecture, it was happy to tell me about Islamic architecture in general and various motifs that were found in the building. So there are definitely educational purposes for this. You can use it for arts and culture science. Apparently, it can give you fashion tips, I don't really know how that works, although it won't comment on people.
So it won't tell you that somebody's ugly or badly dressed or anything like that. You can use it for visual storytelling, so you can upload the picture and it will tell you a story about or come up with a story about that building. So I asked it to create a story about the, again the Palace of Justice image, and it wrote me a short story about the Palace of Whispers and a centuries old Sultan called Rafiq. I mean, it was a bit generic, but it was quite charming.
So actually the image search function is a lot more useful than I thought it was going to be. But as I said before, that killer application is the voice interaction. And again, you were the one who turns me onto it. So how have you found experimenting with voice conversations with it, especially in terms of, you know, the humanness inverted comments of the responses?
It actually shocked me as to how realistic it sounded with, because there's human-like pauses, there are arms, there are arms. And I think it genuinely shocked me when I used it for the first time that I ended up having, and I say this in air quotes, a conversation with the bot to the point where I think I was conversing for about 45 minutes about, you know, my hometown and the surrounding areas and where I should visit and whatnot, stuff I already knew. But it was interesting to hear it from a bot telling me where I should go. And it's probably sourcing all of this data from Wikipedia, stuff that I already know.
But it was as if it was coming from a real person, genuinely. I know, it's quite freaky. I mean, I haven't managed a 45-minute conversation, but then I don't manage 45-minute conversations with people. This is probably the longest I talked to anyone in any given week.
But I mean, just sort of backtracking a bit. So I mentioned there are five personas. So far, I've only checked out Sky. I don't know if you've checked out any of the others yet.
I have. Yep. I landed on Sky. Oh, she reminds me of the voice from her, which is potentially why I stuck with it.
Ah, is that why it's chimed a bell in my head? Yeah, so that's probably why I haven't moved on. And I've found it very impressive as well. You know, when we talk to assistants like Alexa, you know, those are a very kind of limited step towards this screen-free world, you know, this world where we speak to all our devices rather than interact with them through a control device.
And you know, you can't compare that kind of interaction talking to Syria, Alexa, with that experience, with chat, GPT. You know, one of the recurring things on the show has been voice and the advance of large language models using natural language processing. And it's, you know, as my conversations were shorter than yours, but it's amazing how your interactions with the machine change once you're not typing anymore. I found that asking questions was much easier and more natural than when I try and type them.
And I found, or I am finding that it's making the process of researching and searching for answers to things a lot simpler, because, you know, you're not thinking about how to phrase a type sentence for a machine, you know, already that act, that translation of getting out of your brain and typed onto a screen, it slows you down. Your thoughts happen so much faster. And you don't have to go through this process of translating your thoughts, typing them, and adding this layer of trying to phrase them in a way that you think that the machine is going to understand. Isn't there still the same issue when you're speaking?
Yeah, of course. I mean, the machine still misunderstands things, but it's so much easier and less frustrating for you as a user to just change the way that you phrase something, to try another verbal command, than it is to look at your sentence and think, Oh, how am I going to change that to get it to do what I want it to do? And going back to that part, I mentioned about the people who try out these tools and don't see them as anything, but a novelty. I think this is where it will grab those people, because it is much more like having a normal conversation.
And I'm finding that it's changing the way, I think, about asking questions, because I can go off on tangents and come back. I'm not, you know, working through a checklist of questions. As you said, your conversation is just taking you down strange paths, it's giving you information, you're able to riff on the information and you go off in all these different directions. And, you know, it really, I think, matches my kind of tangential learning, which I find, you know, a little bit frustrating with the kind of linear approach of search engines.
So during your trials with the voice, what do you think were some of the limitations that were most apparent to you? Well, I mean, just to go back to you mentioning her, you know, overall, it is very kind of sci-fi. It's like speaking to the ship's computer in Southra, the conversations themselves are very easy. And as you said, there's those natural elements of the speech, the pauses, the arms and arms.
So they don't have that robotic element that you get when you speak with a lot of digital assistants. You can still tell, you know, it's not, it's not human. I don't think it's trying to be human, but it is normal. So in terms of limitations, the verbal misunderstanding, which of course is to be expected, you know, these machines will get better at context over time.
But they're never going to be perfect because people aren't perfect at context either. We'll get things wrong when we're talking to each other and misunderstanding each other. I guess the thing I notice most is the speed of the responses, not in terms of the gaps between you asking the question and it formulating the replies, but in actually getting through the responses, because, you know, we used to just scanning through things and speech is a little bit slower than reading. So it was surprising when I've listened to an answer that would feel like it took, you know, a minute or two minutes or whatever.
And then you look back at the text and you see how short the text is, you're quite surprised because it kind of feels like an eternity when it's being read loud. Yeah. Yeah. Now, how are businesses likely to incorporate these advances, do you think?
Don't ask me, I don't do business stuff. No, I'm just messing. No, I mean, there are lots of applications, things like internal chatbots, you know, for company use. So imagine a machine that's loaded with schematics that's actually conversing with an engineer who's out in the field in real time.
You know, the engineer said, I can see the blue wire, but not the red wire. You know, when you've got your head inside a huge piece of factory plant machinery, you can't always go and look at a screen, but you can talk and you can listen. So I think there's a, you know, a huge number of applications for companies to use it for their internal processes. In terms of, you know, business to business, business to consumer, look at me with all the cool words.
You know, there's massive potential as well, you know, going to a brand site, clicking the microphone button to talk to the bot about the goods and services, rather than going through all the process again of typing. That thing of easily being able to find out what isn't isn't in stock, you know, saying, oh, is that blue t-shirt in stock and the machine just saying, yeah, you can buy it, just click. Much easier for you to raise service or other issues that you have with companies. I mean, I'm not going to go into that side of things today.
So depending on how things work out, I want to have a look at the business case for implementing AI and machine learning technology, and also because in between recording this and it being broadcast, I'm actually doing a panel on this with the business applications for BFM's breakaway this week. So hopefully that will feature a bit in next week's episode. And of course, you know, there are some sectors that are racing ahead with this technology. Others seem to be hedging their bets, but I have to say that I'm really impressed with these new integrations.
They do change the game. And once we see them being extended beyond, you know, the premium tier, once we see them coming more into everyday tools, like search engines like being, that's where we're going to see the worlds of search and internet browsing change rapidly. There, I got through it in just one half of the shirt. And you know what, it's been fairly sensible so far, which worries me.
So let's see how weird things will go after the break. Yeah, folks are tuning into Matt's explain to you on BFM 89.9, the business station.