Hello everyone. Welcome to Talking Tech. This edition available from July the 18th, 2023. I'm Stephen Jolly. Great to have you with us wherever you're listening, perhaps through Vision Australia Radio, associated Stations of Australia or maybe the Community Radio Network, there is also the podcast. If you haven't caught up with that yet, all you need to do is search for the two words. Talking tech and DNA can all 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. With me someone who can explain all this tech stuff really well. Vision Australia's National Advisor on Access Technology, David Woodbridge. David, as you suggested, it's probably about time we had a good yarn about
AI or artificial intelligence. So my first question to you is standing at a barbecue and AI comes up in conversation and someone says to you, What's AI all about?
David So the simple answer about AI is it's to teach or instruct or code. Computers in general use it in quotation marks, computers in general, to think and learn. So they access data, they interpret the data, and then they output some result out of that. So they look at lots and lots and lots of data bases and then whatever you ask for, they're sort of like try and link that statistically together and then give you an appropriate answer. But its basic premise is think and learn.
And the learn is what's a bit of a worry for people? And just briefly, because they could take us into territory people fear that we haven't anticipated.
That's right. It also depends on the data that the thing's accessing. So if the data is incorrect, then the output is going to be incorrect or slightly off the planet a bit. So that's the other thing, because, you know the old expression, Steve, from back in the old days, garbage in, garbage out. In some ways you can get the same thing with AI.
So what's the potential for people who are blind or have low vision with AI?
This has got tremendous benefit because if you think about the thing that you can actually ask this thing. So it's beyond what Siri and AI lady and those sort of things do, which is fairly simple tasks. But for example, I can say to I can you give me the main monuments that I'll be able to find in, I don't know, Paris or Rome or anything else, then I can say, well, look, you know, can you tell me more about X, Y and Z monument? And then we can keep going down the rabbit hole. You can get
as broad as you like. You can get specific like. And what I like about it is that you'll see more systems now that actually give you the links of where it got the information from and pulled it all together for you. It's also really good in being able to, if you like, properly construct things like emails or letters and that sort of stuff. And I've been using that for a while now when I've been sending app developers and web developers some feedback upon perhaps how non accessible
their websites are. I don't copy and paste what the AI has come back with, but what I have done is, you know, the ten points that it's come up with and I might have only thought about four of them is just absolutely amazing. So it's also reminding me that, you know, as a human, I don't know everything and the AI wireless know well, it knows things, doesn't know everything that may or may not be relevant to this particular email I'm sending. And it's almost between AI and
humans and for us blind people. So if you can get lost in websites because there are so many things that are quite inaccessible that this thing really defines and narrows it down for you a lot.
You've been doing a fair bit of reading on this lately. A number of well known people have had a bit to say about it. Bill Gates writes that there's no need to be afraid of AI.
Basically, what he said really in a nutshell was that, look, when calculators came in, everybody thought it was going to be the end of maths, that we wouldn't have had a, you know, add, subtract, divide and times anymore. Then when computers came into the classroom, it's like, well, that's going to be the end. People won't remember how to handwrite or do other things because the computer does it everything
for you. And now if artificial intelligence people are saying, oh, no, that's that's going to be the end of people who are able to to be able to research. But one thing that II doesn't do is that it's not intuitive, it's not empathic, and it's not particularly understanding what it really is good on in quotes, is finding facts that it has access to. And then you can do your own thing with it. But at the moment, there are
places around the world that are looking at guidelines. Lines to say to AI will Perhaps AI systems should only go a certain way and if they start going down the wrong little rabbit hole, then there should be guidelines to say, well, stop, you're not allowed to do that, or you're not allowed to process that data or whatever else it might be. So I think people in Australia we are particularly wary of AI, apparently according to some statistics, and I think it's just wait and see what happens
because humans are not stupid. We're going to be able to work out how to control things and I just think we just need to be positive about being able to develop something that was like a control over.
It's interesting that Bill Gates is having a view on this because, of course, his experience goes back a long way, back to the early days of Microsoft.
Yep. And keep in mind that the original term artificial Intelligence was actually coined in 1956. So it's not a new term that suddenly sprouted up. It's actually been around for you know, over 50 years now. So it's not a brand new term. It's just that it's getting more and more in our face about how important it is and how important it is to make sure we don't lose control over the thing.
What's the head of Amazon got to say about this?
By the article I was looking at, they didn't actually mention a lady at all. You know, they said, this is not your personal assistant. This is all the background data stats crunching in the background that they're going to make code available for third party developers to pull in relevant information to then help their corporate base systems be more AI driven in the future. So he was saying, you know, the bars of the world and the bings
of the world and the japes of the world. Yes, you know, they're in your face and everything else, but like, you know, meta, so i.e. Facebook and then like Amazon, they're also working on artificial intelligence as well. It's not so much in your face, but they're certainly working on it. So I think his point was to say, Please don't forget us because we're also working on an extremely hard.
You mentioned ChatGPT, the OpenAI people responsible for ChatGPT. They're very mindful of the possibility of AI getting out of control and doing something about it.
That's right, because the term that now people are talking about is this thing called super AI. And I remember reading a science fiction book couple of years ago that mentioned the word AI in it, and it was basically getting to the stage where the AI basically decided that the human race was too dangerous. So I decided to start culling it. So it's not that bad at the moment, but I think what open AI people are saying is we're going to make sure that we have guidelines that
will restrict ideas in certain things. And if you wanted a doomsday scenario, Steve, I guess the one of them is that there is no way artificial intelligence is going to have access to to weapons of mass destruction. That would just be stupid if it suddenly decided that, you know, that the world or particular country needed got rid of, then it could take control of the weapons. And that's a really extreme example. But that's the sort of stuff that you would never give an AI system access to.
AI and the potential disruption to the education system as we know it. Ways of learning.
Again, I agree with this article because basically at the moment our education system is very focused on exams and memorizing particular facts or just, you know, repeating what you've been told and writing down an exam. So there's no
intuitive or creative thinking going on. And one of the things that have been finding what AI does and things like ChatGPT is it makes the person more critical in their thinking so they can do reasonable amounts or a lot amount of research based on GPT, also with other fact checking as well and come out with some amazing results. So it's like AI is going to make research a lot easier. It does not mean people are going to end up being stupid and say, Well, this thing just
wrote my exam for you. It's the opposite way around because you've got access to research and pulling in all types of different data. You then as an intelligent human being, can then make up your own mind.
An MIT Massachusetts Institute of Technology study talks about ChatGPT increasing productivity.
Yeah. Now look, I know they're talking about ChatGPT, but let's just, you know, pull it back and make it a bit more general. Let's just say AI in general. And what they found was that and they did, you know, your test, your test group. So one test group had the AI, the other test group didn't, and they had to write a range of different articles like memos and
advertisements and sensitive emails and that sort of stuff. And what they found was the people that use the AI system spent less time actually compiling the email or the
advertisement than people that. Weren't using AI. So again, it's one of those things where, yes, you know, I could sit down here and I could, you know, ponder for 30 minutes on how to write a delicate email to, you know, for me, for a developer to basically say, you need to pull your socks up or I can back into AI and say, Can you write me a nice email? Which points out how incredibly important is accessibility
and your app is not meeting those requirements. And when it spits back the result, I can then massage that text. I can change the wording around, but at least I've got the body. Because you know, sometimes when you're actually writing, you just want that beginning sentence or that beginning idea that then sparks the rest of your writing for you.
That's true. Let's turn to news reading now. And certainly it's started in Asia where AI is being used to generate news bulletins.
This is a weird one. So. So the way this works is and the guy, for example, of this lady in India who, you know, she may not be able to be in the newsroom. So what she does is that she reads the news articles, you know, into a recorder that's sent back to the studio and then that studio, then using her voice puts that voice with this AI, artificial intelligence generated avatar on the screen that looks like her. And then what they also do is they use her
voice to then read in different languages. So in India, it was English and the other major languages spoken in India. And then that person that's reading that in her, you know, the different language using her voice is also a different avatar. So you've got her avatar and another avatar reading the news in a different language. And it's all done by artificial intelligence. And just one more point on this. What
somebody was saying, I think it was at Queensland University. Well, it's all well and good to have these translation things in a different languages, but there are nuances in different languages and different words are important than others and you would still need some sort of linguistic on board to make sure that the translation to a different language wasn't lost in going to from English to another language.
Is the Robodebt debacle in Australia an example of where I could get out of control?
It effectively is. And the simple reason was there was no human being overseeing it properly. By the way, this thing had a 27% failure rate, so that's how much it got wrong and it was literally matching data sets and what it thought where people like debt was actually incorrect. And this comes back to the thing I said earlier on. These things have no empathy, no compassion, no intuitive, no insight, no nothing at the moment. So who knows? Maybe I
will have that eventually. Every time we do things with these, there needs to be guidelines, which we've already talked about, and there always needs to be humans involved, particularly at such intimate things as you owe the government money.
Thinking of our situation as people who are blind or have low vision, you're optimistic about the future role of AI.
I am. If we can use AI to say what's around me, what's the most important thing near me? Can you describe my surroundings? Or, you know, you might pick up a photo and say, Can you describe my child in this photo that my wife took when when they were two years old? Being able to interpret different types of data to make it easier to access information. We're on the cusp of a great movement forward. So I'm really looking forward to the next ten years or so.
And now coming back to the present, just before we go, a reminder of where people can find details of what you've been talking about in this and previous editions of the program.
As always, you can check out our website, which is David Woodmere dot Podbean pad com David.
Would be dot podbean pod b e a n to write to the program.
You can write to me at Vision Australia where I work, which is David Dot Woodbridge. How it sounds at Vision Australia. Org.
David Dot Woodbridge at Vision australia.org. This has been talking tech with me has been Vision Australia's National Advisor on Access Technology, David Woodbridge. I'm Stephen Jolley. Take care. We'll talk more tech next week. See you.