Jesse Mulligan:
RNZ National - it's time for Tech Tuesday and Daniel Watson from Vertech IT Services joins me. Hey Dan.
Daniel Watson:
Good day. Hey, how are you doing?
Jesse Mulligan:
Good. You're talking about Microsoft today, which is a bit of a sore spot for me. I feel like I spent half of my life waiting for Microsoft products to load. I got used to steering at a blank screen, but it may be by hardware more than the software. Well, if the value of their company and the wealth of their founder or anything to go by and they seem to be embracing the future pretty quickly.
Daniel Watson:
Yeah, there's the big two businesses out there in it, Microsoft and Apple, and people have their fanboy and their loves to hates. The thing is that they do actually produce quite a good kit. That's why people use them. And there's other reasons going back to historical ways that how they managed to get themselves into that dominant position in the marketplace. But things have been clicking along with not much too different in things because there was the drive to virtualise things and then there was the drive to the cloud and we've been steady state for half a decade or so. And now the big thing is artificial intelligence or machine learning and large language models. Those are the things which are starting to bubble up and we're now starting to see it actually being, what's the word…
Jesse Mulligan:
I guess, incorporated into some of their other standard products right
Daniel Watson:
Into the hardware. Right? It's like, oh, okay. So what's quite interesting now it's not exactly new. We're talking about neural processing units (NPU’S). You might be familiar with graphics cards and computers, which have been around for quite some time. And that offloads from the main central processing unit, which is the general number cruncher taking all the 2D and 3D calculations off to the side so it can ramp that up. And these days you've got a gaming computer. The grunt that's in the graphics unit is approaching that of the actual central processing unit because for gamers it's all about how quickly you can get the frame rate up.
Now Apple has put new processing units in its chips for a year, so it's not brand new there. But now Microsoft is looking to put these into its surface pro devices. And these NPUs, they're optimised for handling the really complex mathematical that are an integral part of artificial neural networks. So they're just crunching vast amounts of data in parallel, which is useful for things like image recognition and natural language processing because the computer is having to pattern recognise and try and match it all in. And humans we're quite messy. None of this..we've got different accents, we've got different ways that people pronounce words, and it's trying to take all of this really analog stuff and then recognise it as like, oh, it is that word or that is this object in what would be the great blurry analog world of vision and voice.
Jesse Mulligan:
And so that is a form of machine intelligence that works better if you have a unit that is specialised to that task. So the CPU, the main kind of brain just gets to focus on general computing tasks.
Daniel Watson:
Yeah, so in order for software engineers to be say, look, I want to add in this kind of particular feature into a computer, it's going to have to take some processing time from the main central processing unit, or if it's got this NPU, it can go, oh cool. Well actually since that can now do that, I'm going to add in a whole bunch of other stuff that I would like to do as well. So it's taking a bit the breaks off. So if you open your phone by looking at it, that kind of thing, that Apple's been doing it for a while. Microsoft's been doing it for a while, the hello. But it's like it takes a few seconds to go, oh yeah, that's you and lets you in. Part of that is that it's also draining power from the battery life. So in order to do those kind of things, which are seamless from you as a user perspective, the computer's actually having to suck juice out of the battery to run it through the CPU to recognise what's going on through the camera and go, that looks like Jesse Mulligan. Okay, I'll open up by putting this off to the side. Then that can sip battery life only when it's needed. And it's optimised for doing that. And whilst your computer is starting up, that's also doing the recognition in parallel so it's all happening at the same time. So it’s much more seamless.
Jesse Mulligan:
Got it. And that's mostly happening under the hood, and it means that when future applications that really draw heavily on AI start becoming more common, then they just run seamlessly on our computers. So all going, well, we won't notice any difference, but people might've noticed something called copilot popping up in their Microsoft software that have I got the news has that just started getting shipped with the new windows?
Daniel Watson:
Yeah, they're calling it a preview and it's the open version as in it's only going to act upon what you put into it. There's also a paid Microsoft copilot, which is probably more aimed at businesses to be fair, where you have knowledge workers who are looking to really greatly reduce the time they spend on certain tasks and to improve on what they're doing with those tasks already. So the new co pilot preview I've played around for, it's done some nifty things that has saved me a bunch of time already straight there. But where I think the really cool stuff that's going to happen with that is say, I never really learned Excel, right? Somebody showed me one showed you how to do pivot tables, and I was blown away. I was like, wow, that's awesome!
But where something like co-pilot can help you out is if you're using the paint version, it's sitting within Excel and you could provide the prompts to it to tell it what you want to achieve in natural language. What I'd like to do is create a graph of all sales to this section of customers over the last 12 years looking for patterns with regards to seasonality. Then select those out, and then I would like to produce that in a bar. Graph managers explaining it like that. You don't need to know any formulas. No, it's going to dive into your folder where you might have all these previous years Excel spreadsheets or reports. It'll suck the stuff out, bring it in and put it out for you to go, does this look about right? And then you can adjust it from there, which I just hate. Things like Word formatting and you get stuck in a document and you think, I'm going to dolly this up a bit, and then three hours later you're trying to get the paragraphs to align just right.
Jesse Mulligan:
Those applications have always had help services, but you could be in Excel and what would you like help with a table. Sorry, we don't have any articles on tables. And you're like, come on, Excel is all tables. So having AI that can take a dummy, the words of a dummy like me, translate it into computer talk and then put it back into dummy words. If you get into the habit of using it, you'd have to think that would be an advantage.
Daniel Watson:
So seeing these kind of chips going into new ranges of computers, it shows the way that they're leading. They're going, okay, this is clearly, we are going to start developing a lot more in applications that are going to need that grunt to be able to, even in teams and meetings like this, which we're doing over Zoom the program, even if you're not using a flash headset that's got lots of trickery already built into it, having those kind of intelligent voice services at the NPU turbocharges, it can be in real time removing background noise and adjusting the gain levels and clarifying the speech without you just already inbuilt into the computer. Or if you're a creator and you're starting to put together a project where you're wanting to pull together different bits of content that you've created before the AI could actually be suggesting the stuff to you, oh, I see that you're building out this kind of proposal. Here are the three other proposals that you've done recently. Would you like to copy some of the slides across from it? Those are the kind of things that will make a real big difference to people who working in that kind of field
Jesse Mulligan:
And the best thing is just to dip your toe and have a go.
Daniel Watson:
Yeah, have a crack.
Jesse Mulligan:
Yeah. Thanks Dan. Great to chat to you. Thanks for the updates. Dan Watson from Vertech IT Services on Tech Tuesday. Catch you in a couple of weeks.
Have you had a security incident in the past and you have a nagging feeling that not much has changed to prevent it from happening again? Contact us today.
See our cybersecurity services
Explore our co-managed IT services
Need help with process optimization?
Join our systems coaching