AI and its uses

00:21S1

Hello everyone. Welcome to Talking Tech. This edition available from April the 30th, 2024. I'm Stephen Jolly. Great to have you with us. Listening maybe through Vision Australia radio associated stations of RPA Australia or maybe the Community Radio Network. There is also the podcast. To catch that, all you need to do is search for the two words Talking Tech and [?Daniken], or come usually on a Tuesday afternoon just after it's been produced. Another option is to ask your Siri device or smart speaker to play Vision Australia Radio talking tech podcast, Vision Australia Radio Talking tech podcast.

Normally with me is Vision Australia's national advisor on access technology, David Woodbridge. Unfortunately, David can't be with us this week, so instead we'll bring you a conversation that David and I had about four months ago. Towards the end of 2023. Reflecting on the year's developments in technology, particularly from the perspective of people who are blind or have low vision. So here we go. This conversation, taken from the edition of December the 26th, 2023. 

So let's reflect today on what has probably been the major talking point in the tech world over the last year. And that's A.I., Artificial Intelligence. And I must say that I first really started to think about it sort of in current times, around November 2022, when a report on ABC was talking about this new thing, uh, this website open AI, and you could do really interesting things like get it to write letters for you or write essays. And it's all taken off. I think of A.I. really as taking advantage of the really high computing power that we have these days to be able to do things that we just couldn't imagine doing before, but really useful things. How would you describe it?

02:22S2

I tend to describe it as a computer system that responds to speech or typing input. So those two modes are primarily. The second thing I always think about is the fact that it can actually make decisions based on data. And the third thing that really comes off that data is that the fact that it can recognize patterns, because it has this whole database of information, that it can look up extremely fast and recognize patterns of your own personal use or in general use around the world. So that's the main three things that always sticks out for me with computer, artificial intelligence or AI.

03:03S1

Now if you ask some people what I use, they'd say, Oh, something tech that's got the capacity to destroy civilisation as we know it, but it can do a lot of very useful things, and let's not worry about the threats of it at the moment, because that's a long conversation some other time. What sorts of things is it doing?

03:21S2

It's actually quite amazing because when you think about AI, we're already using it. So people get sort of blown away or get nervous about chat, GPT and other systems, but we're already using it in everyday lives. For example, I've got a bit of a list here. So advanced web search that's actually AI, because it also brings up suggestions for you when you do a search, when you're watching Netflix, ABC, iview, I'm assuming, because I really haven't delved in too much. When you watch a show, it then comes up with other suggestions that other things that you may want to watch in the future or now. So again, that's pattern recognition and making a bit of a decisions for you.

Let's think about virtual assistants. So they're the classic A lady I won't say the name Google and Siri, which we always have in our smartphones. Again, to me that's a basic version of AI, because it's really command line driven. It doesn't have these large language models that it can look up speech recognition. So not just talking to your smart speakers or these virtual assistants, uh, but also controlling your computer and of course, dictating to your computer, which is actually pretty amazing. Automated translation. So for those that remember Hitchhiker's Guide to the Galaxy in the famous Babelfish, which is stuck in your ear for automatic language translation systems do that automatically for you now. 

And that's again, I, nowadays, Stephen, we can get an audio file of somebody talking and get a transcription of it, in text. So again, I've already mentioned smart home, but that's really in relation to virtual. But when I had this as a separate topic, your systems learn when you come home, when you want the air conditioner on, when you want the lights turned off. That's all about pattern recognition. And for industry it's predictive maintenance. So the AI systems will decide when you know the trucks need servicing or the machinery that's used for packing stuff in the warehouse needs maintenance. Natural language processing. 

And this is really where things like, uh, ChatGPT and Copilot and Bing stand out, because you can just have a bit of a general chat to these things now. And based on the keywords that you use, it decides what things you want to look up and give the information back to you. Content creation, which is one you already mentioned there. That's where, for example, if I wanted to write an email to a company about web accessibility or application accessibility, I can say to the AI, look, I'd like you to write a nice, gentle, but assertive email about the importance of web or application accessibility. And it's just the amazing stuff that it writes. So it says, you know, dear Sir or madam, we'd like to bring to your attention, etc., etc., etc. and it really does have all the main points, which is really amazing. 

This is the one that I particularly like when I'm using and doing research for the show is automatic text summarisation. So I can actually ask a AI to summarize an article for me, or some text that I've given it, and it will give me the main points. When we access our smartphones, we're using facial recognition. Again, I spell checking you might not think this is AI, but I is indeed one version of spell checking word suggestions. So this is not so much about word prediction, which is another form of AI. This is where you partially type in a word, and then the computer comes back and gives you a list of other words that are similar to the word you just started typing in, which is actually really amazing. 

Cybersecurity is another major one, weather prediction, because when you ask your weather system or your weather app, what's the weather going to be? That's based on computer generated weather tables around the world and looking at trends and that sort of stuff. Health and well-being, automated customer systems. You know, when you ring up those phone system and it says, how would you like to chat to? And you might say, well, for a mobile, you might say mobile or mobile internet. And then it'll keep asking you these questions, and I'll just say these ones quickly, because they're all sort of similar aviation, shipping trains and other forms of public transport, all again driven by AI automatic self-driving cars. Who doesn't want one of those automated drones? 

And finally, robotics, which is the the one that scares people about, you know, you think of, Isaac Asimov and other famous science fiction writers. That's the one that always gets people worried because they think, well, robots are going to take off the world. That'll be the end of the human race. So there are all the main general ones.

08:10S1

Let's talk now then, about specifically where it's making a difference for people who are blind or have low vision, where AI is making a difference.

08:19S2

So the ones that are making a huge difference, and the ones that have already said sort of cover these, but these are more specific ones that I guess to remind people. So object recognition and of course these are mainly all based around your camera and your smartphone. So that's tell you what sort of objects may be around you. And then we've also got distance from objects. So how close is the object away. So that's almost like you know a good old orientation mobility to you know beep beep or now someone's away indoor navigation as opposed to GPS. This is indoor navigation where systems learn a route inside a building and then can take identification markers of items around you to position you in a, in a building sort of very similar to object recognition is scene detection. 

So scene detection is a way where you can take a picture of somebody's backyard or a park, and then it will tell you the position of things in the park, for example. So where the swings are, they'll identify the trees, the seats and so on.

09:20S1

Now this next one really appeals to me on your list. The barcode or QR code identification.

09:26S2

Again, being able to point your camera. And particularly there's an app that makes it very easy to access barcodes. I don't believe it does QR codes at the moment, QR codes, but again, that's actually very effective. The one that I don't like because it's it's a factor of light, is color recognition because the based on what the light's bouncing back from the object, you can get different versions of color. So that's a thing that I think needs to be worked on. Speaking about light, light detection. So you can tell whether a light's on or off or a device is on and off by the little LED lights on the device. Pebble detection. And I use this quite a lot handwriting recognition as opposed to the other one, which is also optical character recognition OCR for reading text. 

So handwriting is very cool. Voice control. So again, this is probably more advanced than just general speech recognition. This is also being able to control your assistive technology, whether it's a screen reader or screen magnifier, switch control, etc. and interact with the computer. So you're using the computer, the program and your assistive technology. It's really, really amazing.

10:33S1

And tell us about some of the apps that really excel with AI.

10:37S2

These apps are actually going to be quite well known to people. And the first one we have to mention for both iOS and I'm pleased to say Android now is, of course, the Seeing Eye app from Microsoft. And that's a suite of applications inside it called channels. So you've got things like short text, OCR, document text for OCR product recognition, which is barcode. You've got people detection, you've got scene detection, light detection, currency detection, color detection and so on.

11:08S1

I just want to put in a quick word here for the document reader in seeing AI, because seeing AI is really improved. I put my rates notice in front of it the other day and. Instead of it reading me the whole document, I said, can you tell me when the due date is? Can you tell me the amount, total amount owing, etc.? It's really good.

11:32S2

Yeah, it does work really well, particularly when you can actually, I don't like the word interrogate, but you can ask questions about the document that it's just read, and you can really narrow down on the information that it's got. So, so to me, that's the main one that stands out, you know, really for the last 12 months, of course, the one, the main one. Besides seeing I now on Android, you've got the lookout app, which does stuff primarily very similar to seeing AI. And then there's other, you know, assistive tech ones and of course, the, the couple of the main ones that I use for, you know, I chat type stuff is of course the chat GPT because it's got that voice mode now where you can speak to it and it speaks to you, being the new version that's now called copilot, that's a live search, plus a large language model, and the one that I use on the Mac all the time now is actually Copilot itself via the web page on the Mac, which of course is built in to Windows 10 and 11 now. So that's actually pretty cool.

12:37S1

Perplexity is another very popular one that certainly works in the iOS environment. And we mustn't forget to acknowledge the Be My Eyes virtual assistant. Be My Eye.

12:48S2

Yeah, that's actually really set the bar extremely high. And of course, that's where you can take a photo of something and it gives you back all the information. And you can also ask it questions as well. So I think, you know, given that this stuff's jumped up in the last year or so, I think it's going to be really amazing where this stuff goes in the next, you know, next year or so even.

13:07S1

All right. So that's, I and I wonder what we'll be saying about it in 12 months time.

13:14S2

I think we'll have our socks blown off in another 12 months. Besides the web and mobile phones, this, for me, is the next third thing in 30 years of doing my job. Then I'll be really, really, really excited about.

13:27S1

So that's I our take on it at the moment. Before we go, a reminder of where people can find details of what we've been talking about in this and previous editions of the program.

13:38S2

As always, you can check out my blog site, which is David Woodbridge Podbean pod bean.com.

13:46S1

David Woodbridge Podbean pod b e a n com to write to the program.

13:52S2

You can write to me at David Dot Woodbridge - how it sounds - at Vision Australia - dot - org.

13:58S1

david.woodbridge@visionaustralia.org .... or Jolley Stephen at gmail.com ... Jolley Stephen,  j o l l e y s t e p h e n at gmail.com. This has been Talking Tech. We've been hearing from Vision Australia's National Advisor on Access Technology, David Woodbridge, reflecting on developments in technology throughout 2023. I'm Stephen Jolley. Take care. We'll talk more tech next week. See you.