1 00:00:00,080 --> 00:00:02,639 Speaker 1: There is no longer any escaping it. It will find 2 00:00:02,680 --> 00:00:06,280 Speaker 1: you AI. Today, we've got Janey Plant joining us as 3 00:00:06,280 --> 00:00:08,720 Speaker 1: we delve into AI all this week. Janey, you are 4 00:00:08,760 --> 00:00:11,559 Speaker 1: the chair of Women in Technology in WA and that 5 00:00:11,600 --> 00:00:14,240 Speaker 1: means you are pretty familiar with AI. 6 00:00:14,600 --> 00:00:15,720 Speaker 2: Yes, I very much him. 7 00:00:15,840 --> 00:00:21,560 Speaker 1: I'm terrified of it. There are benefits, there are risks, 8 00:00:21,560 --> 00:00:23,440 Speaker 1: and there are limitations. And this is what I'm hoping 9 00:00:23,440 --> 00:00:25,040 Speaker 1: to get from this week, to be a little. 10 00:00:24,880 --> 00:00:28,320 Speaker 3: Less terrified to understand it. The more you understand it, 11 00:00:28,360 --> 00:00:29,240 Speaker 3: the less terrified you. 12 00:00:29,280 --> 00:00:31,120 Speaker 1: I don't think I can afford to be terrified of it. 13 00:00:31,560 --> 00:00:32,440 Speaker 2: No, you probably can't. 14 00:00:32,440 --> 00:00:36,400 Speaker 1: There probably are bits to be terrified of. What What 15 00:00:36,440 --> 00:00:40,280 Speaker 1: are some of the downsides of using AI in society? 16 00:00:41,159 --> 00:00:44,480 Speaker 2: So I suppose some of the downsides is that it's 17 00:00:44,520 --> 00:00:48,320 Speaker 2: not foolproof, so it does make mistakes, and if people 18 00:00:48,400 --> 00:00:51,839 Speaker 2: are over relying on it all the time, then that's 19 00:00:51,880 --> 00:00:55,840 Speaker 2: going to cause problems. If you are using it to 20 00:00:58,160 --> 00:01:02,680 Speaker 2: make decisions that are critical, then that can also cause issues, 21 00:01:02,720 --> 00:01:05,080 Speaker 2: so that you have to kind of make sure that 22 00:01:05,120 --> 00:01:06,480 Speaker 2: you're using it for the right things. 23 00:01:06,720 --> 00:01:09,920 Speaker 3: Yeah, because I saw a report would have been about 24 00:01:09,920 --> 00:01:12,240 Speaker 3: two or three days ago where in China, they had 25 00:01:12,520 --> 00:01:18,360 Speaker 3: a panel of surgeons and doctors and the AI made 26 00:01:18,400 --> 00:01:21,560 Speaker 3: the diagnosis in two seconds. Yeah. 27 00:01:21,600 --> 00:01:26,319 Speaker 2: So even here at Uwa, I have seen research where 28 00:01:26,360 --> 00:01:30,840 Speaker 2: they are like feeding in like scans people that have 29 00:01:31,000 --> 00:01:33,479 Speaker 2: got cancers and things like that, and then they're using 30 00:01:33,520 --> 00:01:37,759 Speaker 2: AI to actually depict through to find like changes in 31 00:01:37,880 --> 00:01:42,080 Speaker 2: their cancers so that they can customize and tailor the treatment. 32 00:01:42,360 --> 00:01:44,520 Speaker 2: So instead of kind of going, you know, we normally 33 00:01:44,520 --> 00:01:45,880 Speaker 2: do A and then we do B and then we 34 00:01:45,920 --> 00:01:48,760 Speaker 2: do C, they might be able to see a particular 35 00:01:48,800 --> 00:01:51,680 Speaker 2: patient who they've done A, but in actual fact they 36 00:01:51,720 --> 00:01:54,960 Speaker 2: now need to jump to D because B and C irrelevant. 37 00:01:54,400 --> 00:01:58,760 Speaker 1: Now, right, So that comes under the heading of benefits AI. 38 00:01:59,160 --> 00:02:02,440 Speaker 1: But what about a lot of people, especially younger people, 39 00:02:02,880 --> 00:02:05,840 Speaker 1: are using AI to replace human relationships? 40 00:02:06,560 --> 00:02:08,359 Speaker 3: This cannot be good. 41 00:02:08,639 --> 00:02:12,320 Speaker 2: Yes, no, it absolutely can't, and I suppose it can't. 42 00:02:12,360 --> 00:02:15,519 Speaker 2: So it doesn't have feelings obviously, So now I think 43 00:02:15,800 --> 00:02:18,520 Speaker 2: I think Sarah was saying on Monday that that's a 44 00:02:18,600 --> 00:02:21,000 Speaker 2: kind of a watch this space. But at the moment 45 00:02:21,040 --> 00:02:24,160 Speaker 2: it certainly doesn't have feelings. It doesn't have values or 46 00:02:24,240 --> 00:02:29,040 Speaker 2: real world experience, so it can't truly understand you're like 47 00:02:29,040 --> 00:02:32,040 Speaker 2: a personal context or love, like I can't love your kids, 48 00:02:32,160 --> 00:02:35,280 Speaker 2: can't take responsibility in the same way that a human can. 49 00:02:35,440 --> 00:02:38,560 Speaker 2: So if you like things like trust, which is obviously 50 00:02:38,600 --> 00:02:42,600 Speaker 2: critical for relationships, empathy, ethical judgments, that kind of stuff, 51 00:02:42,639 --> 00:02:45,640 Speaker 2: that all sits still very squarely with humanity. 52 00:02:45,720 --> 00:02:47,760 Speaker 1: When your AI boyfriend says how are you and you 53 00:02:47,800 --> 00:02:53,239 Speaker 1: say fine, he is not going to understand the subtle 54 00:02:54,120 --> 00:02:56,040 Speaker 1: are what fine means? 55 00:02:56,160 --> 00:02:59,079 Speaker 2: So that's actually that's actually really interesting. Just when Russell 56 00:02:59,120 --> 00:03:01,720 Speaker 2: said cannot read the time, I mean you can ask 57 00:03:01,760 --> 00:03:03,640 Speaker 2: it to read the tone so you can say to it, 58 00:03:03,680 --> 00:03:06,320 Speaker 2: you know, I want you to you know, like like 59 00:03:06,440 --> 00:03:09,680 Speaker 2: my tone is annoyed or something. You can kind of 60 00:03:09,680 --> 00:03:13,440 Speaker 2: give it like paramative one says that no, so you 61 00:03:13,520 --> 00:03:16,040 Speaker 2: have to kind of yeah, so you are literally kind 62 00:03:16,040 --> 00:03:17,799 Speaker 2: of giving it context. 63 00:03:17,919 --> 00:03:21,040 Speaker 3: It's really back, yes, back to the risks to get 64 00:03:21,240 --> 00:03:24,000 Speaker 3: you know what, what about in a critical situation where 65 00:03:24,000 --> 00:03:26,560 Speaker 3: AI makes a mistake, and so we were talking about 66 00:03:26,560 --> 00:03:29,720 Speaker 3: the medical side of things before, who's responsible? 67 00:03:29,960 --> 00:03:33,040 Speaker 2: So that's an excellent question. So at the moment, we 68 00:03:33,080 --> 00:03:36,920 Speaker 2: don't have any specific regulations or laws in Australia. We 69 00:03:36,960 --> 00:03:39,240 Speaker 2: do have some guardrails and guidelines, so we've got the 70 00:03:40,080 --> 00:03:43,480 Speaker 2: government's eight Ethical Principles, and then there's a National AI 71 00:03:43,600 --> 00:03:45,600 Speaker 2: Framework and so on and so forth, but none of 72 00:03:45,640 --> 00:03:52,400 Speaker 2: those are like compulsory. But in any case, even without 73 00:03:52,400 --> 00:03:56,040 Speaker 2: those specific laws, what we do have are laws that 74 00:03:56,160 --> 00:03:59,840 Speaker 2: generally apply across the board. So we have copyright laws, 75 00:04:00,080 --> 00:04:03,560 Speaker 2: have privacy laws, we have anti discrimination laws. You know, 76 00:04:03,640 --> 00:04:08,480 Speaker 2: all those sorts of laws. They all apply to everything anyway. 77 00:04:08,640 --> 00:04:11,720 Speaker 2: So you know, it's no AI is a tool. It's 78 00:04:11,760 --> 00:04:14,400 Speaker 2: no different in that way to any other tool that 79 00:04:14,440 --> 00:04:16,600 Speaker 2: you're using. So you need to make sure that it's secure, 80 00:04:16,720 --> 00:04:20,279 Speaker 2: that you're not you know, like leaking people's private data 81 00:04:20,440 --> 00:04:21,360 Speaker 2: all that kind of stuff. 82 00:04:21,520 --> 00:04:23,479 Speaker 1: I've always been curious in as sort of a chicken 83 00:04:23,520 --> 00:04:28,600 Speaker 1: in the egg situation. AI. They say it's artificial intelligence, 84 00:04:29,040 --> 00:04:33,320 Speaker 1: but doesn't it always sort of have to be programmed 85 00:04:33,520 --> 00:04:37,839 Speaker 1: very very originally somehow by a person at a human 86 00:04:38,440 --> 00:04:41,200 Speaker 1: So when does it sort of you know, people say 87 00:04:41,240 --> 00:04:43,440 Speaker 1: it's going to take over and stuff, but don't we 88 00:04:43,760 --> 00:04:46,080 Speaker 1: have to make it do what it does? 89 00:04:46,440 --> 00:04:52,080 Speaker 2: So initially it's basically based on data sets. So some 90 00:04:52,200 --> 00:04:55,120 Speaker 2: of the AI tools that you would have heard of 91 00:04:56,040 --> 00:04:58,479 Speaker 2: based on a data set. And we'll just call it 92 00:04:58,480 --> 00:05:01,919 Speaker 2: the internet because we're not certain exactly what it's based on, 93 00:05:02,520 --> 00:05:09,000 Speaker 2: but obviously, because the Internet is based on humanity, then 94 00:05:09,120 --> 00:05:13,479 Speaker 2: obviously there's going to be problems with what you kind 95 00:05:13,520 --> 00:05:15,760 Speaker 2: of get out of it, and so things like bias 96 00:05:16,000 --> 00:05:18,760 Speaker 2: and you know that kind of stuff is going to 97 00:05:18,760 --> 00:05:22,040 Speaker 2: be sort of an issue. But what it does after 98 00:05:22,120 --> 00:05:24,200 Speaker 2: that is you can sort of use subsets. So if 99 00:05:24,240 --> 00:05:26,400 Speaker 2: you're an organization and you want to be able to 100 00:05:26,480 --> 00:05:29,320 Speaker 2: use only your own data, you can definitely do that right, 101 00:05:29,839 --> 00:05:31,680 Speaker 2: and then it's really just about what you've got to 102 00:05:31,680 --> 00:05:34,520 Speaker 2: make sure your data is, you know, cleansed, like it's 103 00:05:34,520 --> 00:05:36,320 Speaker 2: not just a bunch of rubbish, you know, that kind 104 00:05:36,320 --> 00:05:38,600 Speaker 2: of stuff, because obviously it's that garbage and garbage out 105 00:05:38,720 --> 00:05:42,320 Speaker 2: kind of thing. So you can kind of isolate the 106 00:05:42,400 --> 00:05:44,200 Speaker 2: data that you're using, which is what a lot of 107 00:05:44,640 --> 00:05:49,280 Speaker 2: like specific organizations would do. But if you're just kind 108 00:05:49,320 --> 00:05:51,440 Speaker 2: of looking at the Internet, then yes, you need to be. 109 00:05:51,440 --> 00:05:53,120 Speaker 2: If you're looking at sort of tools that are just 110 00:05:53,200 --> 00:05:55,880 Speaker 2: kind of using that massive data set, then you do 111 00:05:56,000 --> 00:05:58,520 Speaker 2: need to be sort of cognizant of what some of 112 00:05:58,520 --> 00:06:01,880 Speaker 2: the issues can be around that. But it does learn 113 00:06:02,200 --> 00:06:04,640 Speaker 2: from things that you do, and it does learn from 114 00:06:04,760 --> 00:06:07,400 Speaker 2: the content that people are putting in. So if you're 115 00:06:07,520 --> 00:06:09,919 Speaker 2: if you're typing things into it, it is learning from that. 116 00:06:10,040 --> 00:06:12,120 Speaker 2: If you haven't told it that, you don't want it 117 00:06:12,160 --> 00:06:12,760 Speaker 2: to learn from that. 118 00:06:14,160 --> 00:06:16,880 Speaker 3: I guess given all that, and you look at all 119 00:06:16,920 --> 00:06:20,520 Speaker 3: the sides at the moment, would you say the risks 120 00:06:20,920 --> 00:06:23,920 Speaker 3: outweigh the benefits. What is it? Is it good? Is 121 00:06:23,960 --> 00:06:24,400 Speaker 3: it bad? 122 00:06:24,680 --> 00:06:27,960 Speaker 2: So I would say that I am optimistic. I heard 123 00:06:28,000 --> 00:06:30,160 Speaker 2: someone say the other day they are cautiously optimistic. I'm 124 00:06:30,160 --> 00:06:33,160 Speaker 2: not sure if I'm cautiously optimistic. I am optimistic. I 125 00:06:33,160 --> 00:06:35,160 Speaker 2: think that we do have a long way to I 126 00:06:35,200 --> 00:06:37,719 Speaker 2: think education is a really big part of it. We 127 00:06:37,960 --> 00:06:40,760 Speaker 2: need to make sure that people understand it. They understand 128 00:06:41,120 --> 00:06:44,279 Speaker 2: the risks, the nuances, how to use it properly, what 129 00:06:44,480 --> 00:06:46,840 Speaker 2: not to use it for, when to rely on, and 130 00:06:46,920 --> 00:06:50,240 Speaker 2: so on and so forth. So as yeah, absolutely so 131 00:06:50,279 --> 00:06:51,880 Speaker 2: as a lawyer, one of the things that we're seeing 132 00:06:52,000 --> 00:06:54,120 Speaker 2: is people turn I mean doctors would have had this 133 00:06:54,160 --> 00:06:56,800 Speaker 2: previously with doctor Google, right, So we have people turning 134 00:06:56,839 --> 00:06:59,320 Speaker 2: up and going, oh, here's my legal advice, and it 135 00:06:59,400 --> 00:07:02,480 Speaker 2: sounds authoritative, and so you've got to kind of say yes, 136 00:07:02,560 --> 00:07:05,560 Speaker 2: but in your situation, it doesn't. You know that doesn't 137 00:07:05,560 --> 00:07:08,960 Speaker 2: actually apply, and it can be quite tricky to sort 138 00:07:09,000 --> 00:07:12,560 Speaker 2: of get people to understand that, you know, the legal 139 00:07:12,560 --> 00:07:15,880 Speaker 2: advice that you're actually getting from a lawyer is totally 140 00:07:15,920 --> 00:07:18,520 Speaker 2: different and well worth the money that you're paying for 141 00:07:18,560 --> 00:07:21,360 Speaker 2: it to what you can get from you know, chat, 142 00:07:21,400 --> 00:07:22,560 Speaker 2: GPT or something like that. 143 00:07:23,320 --> 00:07:26,400 Speaker 1: We shall continue to watch this space. Then, thank you 144 00:07:26,520 --> 00:07:28,560 Speaker 1: Jamie today. 145 00:07:28,600 --> 00:07:32,080 Speaker 3: Helping us on our journey to understand the brave new 146 00:07:32,120 --> 00:07:35,160 Speaker 3: world of AI. It's coming for us. You can run, 147 00:07:35,440 --> 00:07:36,320 Speaker 3: but you can't hide.