1 00:00:00,000 --> 00:00:02,250 Speaker 1: This is a C N. A podcast 2 00:00:04,510 --> 00:00:07,770 Speaker 1: launched in november last year, chat GPT has taken the 3 00:00:07,770 --> 00:00:10,760 Speaker 1: world by storm. It's been called a souped up chat 4 00:00:10,760 --> 00:00:13,530 Speaker 1: bot that can do whatever you wanted to need an 5 00:00:13,540 --> 00:00:17,759 Speaker 1: essay on 20th century japanese politics. Check what about a 6 00:00:17,760 --> 00:00:20,850 Speaker 1: resume for your new job done or you need help 7 00:00:20,850 --> 00:00:25,170 Speaker 1: with a complicated math problem. You're stuck at easy peasy 8 00:00:25,579 --> 00:00:28,160 Speaker 1: millions have tried their hands at it and have come 9 00:00:28,160 --> 00:00:32,050 Speaker 1: away suitably impressed. We did it to asking it to 10 00:00:32,050 --> 00:00:36,980 Speaker 1: write an introduction about how chat GPT has changed education 11 00:00:37,220 --> 00:00:38,599 Speaker 1: and here's what it said. 12 00:00:38,990 --> 00:00:42,070 Speaker 2: I am an Ai language model and do not have 13 00:00:42,070 --> 00:00:46,120 Speaker 2: the ability to change education on my own. However, my 14 00:00:46,120 --> 00:00:49,210 Speaker 2: capabilities can be used to assist in the development of 15 00:00:49,220 --> 00:00:55,350 Speaker 2: educational technology and resources such as personalized learning programs, virtual 16 00:00:55,350 --> 00:00:57,780 Speaker 2: tutoring and automated essay grading. 17 00:00:58,020 --> 00:01:02,380 Speaker 2: Additionally, my ability to understand and generate human language can 18 00:01:02,380 --> 00:01:06,630 Speaker 2: be used to improve communication and collaboration in online learning 19 00:01:06,630 --> 00:01:11,080 Speaker 2: environments overall, I can be a powerful tool for enhancing 20 00:01:11,080 --> 00:01:13,850 Speaker 2: and supplementing traditional education methods 21 00:01:14,959 --> 00:01:17,750 Speaker 1: and just for the record, we did not change a 22 00:01:17,750 --> 00:01:20,450 Speaker 1: word of that immediately. I'm thinking how this would have 23 00:01:20,450 --> 00:01:22,610 Speaker 1: been awesome back in the day when I was in 24 00:01:22,610 --> 00:01:27,090 Speaker 1: university helping me churn out my essays. But here's the dilemma. 25 00:01:27,100 --> 00:01:29,480 Speaker 1: How will we ever be able to tell if something 26 00:01:29,480 --> 00:01:32,819 Speaker 1: written was by a human or by computer and if 27 00:01:32,819 --> 00:01:35,540 Speaker 1: we can't then how can we still grade students on 28 00:01:35,540 --> 00:01:38,800 Speaker 1: their essays? If we're not even sure they wrote it? 29 00:01:38,920 --> 00:01:41,840 Speaker 1: Will everyone sunday sounds smart 30 00:01:42,000 --> 00:01:44,530 Speaker 1: and there in lies the bigger question, how will this 31 00:01:44,530 --> 00:01:47,310 Speaker 1: change the way we learn, teach and tests 32 00:01:48,150 --> 00:01:51,520 Speaker 1: with me today to non bots or humans as we 33 00:01:51,520 --> 00:01:54,840 Speaker 1: like to call ourselves. I have Toby Walsh, she is 34 00:01:54,840 --> 00:01:58,070 Speaker 1: chief scientist at U N S W dot Ai. That's 35 00:01:58,070 --> 00:02:02,090 Speaker 1: the University of New South Wales New Ai Institute. It's 36 00:02:02,090 --> 00:02:04,780 Speaker 1: a pleasure to join you and your listeners today, Jonathan 37 00:02:04,780 --> 00:02:07,930 Speaker 1: Sim is an instructor with the department of Philosophy at 38 00:02:07,930 --> 00:02:11,340 Speaker 1: the National University of Singapore. Hi everyone happy to be here. 39 00:02:11,350 --> 00:02:14,180 Speaker 1: All right guys, let's get a crack at this. First 40 00:02:14,180 --> 00:02:17,240 Speaker 1: up your reactions when you first heard about chat. GPT. 41 00:02:18,186 --> 00:02:21,145 Speaker 1: I must admit I was also quite impressed. I've worked 42 00:02:21,145 --> 00:02:23,686 Speaker 1: in the field of Arctic intelligence now for more than 43 00:02:23,686 --> 00:02:26,186 Speaker 1: 40 years and it does seem to be a bit 44 00:02:26,186 --> 00:02:29,216 Speaker 1: of a step change making this technology accessible to the 45 00:02:29,216 --> 00:02:32,016 Speaker 1: people I reminded of when I saw the very first 46 00:02:32,026 --> 00:02:35,316 Speaker 1: demo of the iPhone Steve jobs up on stage and 47 00:02:35,316 --> 00:02:36,716 Speaker 1: that sort of feeling this is going to be a 48 00:02:36,716 --> 00:02:40,046 Speaker 1: technology that's going to enter all our lives and change 49 00:02:40,056 --> 00:02:41,376 Speaker 1: a lot of things that we do 50 00:02:41,392 --> 00:02:44,062 Speaker 1: and it has indeed changed many of our lives, Jonathan 51 00:02:44,072 --> 00:02:46,422 Speaker 1: the first time I played with chad G P. T. 52 00:02:46,422 --> 00:02:48,572 Speaker 1: I got it to write a story about a pig 53 00:02:48,572 --> 00:02:51,742 Speaker 1: and a robot making friends with each other and I 54 00:02:51,742 --> 00:02:54,862 Speaker 1: was very surprised at how compelling the story was. But 55 00:02:54,872 --> 00:02:57,502 Speaker 1: what it did tell me is it is a game 56 00:02:57,502 --> 00:03:00,052 Speaker 1: changer And I believe we're actually at the start of 57 00:03:00,052 --> 00:03:04,401 Speaker 1: a big revolution in education. Well let's help our listeners 58 00:03:04,401 --> 00:03:05,512 Speaker 1: understand a little bit about 59 00:03:05,528 --> 00:03:07,548 Speaker 1: Ai maybe Toby are you able to give us a 60 00:03:07,558 --> 00:03:10,188 Speaker 1: quick and broad understanding of just what A. I. Or 61 00:03:10,198 --> 00:03:13,778 Speaker 1: artificial intelligence is? One way to make an Ai researchers 62 00:03:13,778 --> 00:03:17,078 Speaker 1: squirm is to ask him or her to defy what 63 00:03:17,088 --> 00:03:21,368 Speaker 1: Ai is. But I will seize the challenge. Nevertheless, broadly 64 00:03:21,368 --> 00:03:23,818 Speaker 1: speaking I think most of my colleagues would agree it's 65 00:03:23,818 --> 00:03:26,998 Speaker 1: trying to get computers to do things tasks that when 66 00:03:26,998 --> 00:03:29,648 Speaker 1: humans do them we say they require intelligence. So 67 00:03:29,663 --> 00:03:32,804 Speaker 1: you can break that down into perceiving the world, understanding 68 00:03:32,804 --> 00:03:37,564 Speaker 1: speech recognizing objects with your eyes, reasoning about that world. 69 00:03:37,574 --> 00:03:41,534 Speaker 1: Coming up with plans and making decisions and then acting 70 00:03:41,534 --> 00:03:44,354 Speaker 1: in that world. So that's why robots end up in A. I. 71 00:03:44,354 --> 00:03:46,484 Speaker 1: Because you know for them to move around the world, 72 00:03:46,494 --> 00:03:49,414 Speaker 1: they need to have some intelligence. So chat GPT is 73 00:03:49,414 --> 00:03:51,664 Speaker 1: the latest program to come out of open AI. And 74 00:03:51,664 --> 00:03:53,754 Speaker 1: that's a research lab out of California 75 00:03:53,950 --> 00:03:56,270 Speaker 1: Ai is not really that new. It's been around for 76 00:03:56,270 --> 00:03:58,190 Speaker 1: a while and it's at the level where you can 77 00:03:58,190 --> 00:04:01,670 Speaker 1: play chess it has beaten chess masters. So what makes 78 00:04:01,670 --> 00:04:04,560 Speaker 1: chat GPT so different to things. I think it's the 79 00:04:04,560 --> 00:04:07,060 Speaker 1: scale and the accessibility. 80 00:04:07,220 --> 00:04:09,790 Speaker 1: So this is built on top of GPT. Three which 81 00:04:09,790 --> 00:04:13,560 Speaker 1: is the largest chat pot we've ever built. It's had 82 00:04:13,560 --> 00:04:16,880 Speaker 1: pretty much the whole of the contents the text you 83 00:04:16,880 --> 00:04:20,170 Speaker 1: can find on the internet, all of Wikipedia, all of reddit, 84 00:04:20,180 --> 00:04:23,250 Speaker 1: all of the U. S. Patent database and anything text 85 00:04:23,260 --> 00:04:25,909 Speaker 1: has been pretty much poured into it and trained on. 86 00:04:25,980 --> 00:04:27,930 Speaker 1: And second is they've worked really hard to make it 87 00:04:27,930 --> 00:04:30,640 Speaker 1: accessible and to deal with some of the failings and 88 00:04:30,640 --> 00:04:31,940 Speaker 1: we're probably gonna discuss those 89 00:04:32,290 --> 00:04:36,110 Speaker 1: more details shortly of existing chatbots the ability for them 90 00:04:36,110 --> 00:04:38,300 Speaker 1: to make up things or to go off on a tangent. 91 00:04:38,300 --> 00:04:41,020 Speaker 1: So they put quite a bit of effort with humans 92 00:04:41,029 --> 00:04:44,349 Speaker 1: training to improve the quality of the output. And it 93 00:04:44,350 --> 00:04:46,830 Speaker 1: does show it's much more focused and much better at 94 00:04:46,830 --> 00:04:49,710 Speaker 1: giving you better quality answers but not perfect. It still 95 00:04:49,710 --> 00:04:54,140 Speaker 1: makes some pretty large mistakes and real logical howlers and 96 00:04:54,150 --> 00:04:56,250 Speaker 1: makes just stuff up and set it in a very 97 00:04:56,250 --> 00:04:57,359 Speaker 1: confident way. 98 00:04:57,990 --> 00:05:01,349 Speaker 1: Sometimes that's half the battle won just saying it confidently. 99 00:05:01,360 --> 00:05:03,390 Speaker 1: But a I was meant to you know has been 100 00:05:03,390 --> 00:05:06,770 Speaker 1: around we've seen many manufacturing plants, it's you know, robots 101 00:05:06,770 --> 00:05:09,350 Speaker 1: run with it. It's taken a lot of the menial 102 00:05:09,350 --> 00:05:12,839 Speaker 1: day to day boring tasks out of our lives. But Jonathan, 103 00:05:12,850 --> 00:05:14,099 Speaker 1: are you surprised about 104 00:05:14,110 --> 00:05:17,760 Speaker 1: how sophisticated chat GPT is in particular and is it 105 00:05:17,770 --> 00:05:20,550 Speaker 1: evolving at a point where it is coming close to 106 00:05:20,550 --> 00:05:23,430 Speaker 1: what we think like as human beings. So I like 107 00:05:23,430 --> 00:05:26,700 Speaker 1: to actually think of chat GPT like a model student 108 00:05:26,700 --> 00:05:30,900 Speaker 1: that has memorized many many model answers right when it 109 00:05:30,900 --> 00:05:33,620 Speaker 1: comes to the kinds of work that we humans do, 110 00:05:33,620 --> 00:05:37,420 Speaker 1: if it requires low levels of low thinking skills, then 111 00:05:37,430 --> 00:05:39,650 Speaker 1: chat GPT does a fantastic job. You know, if it's 112 00:05:39,650 --> 00:05:43,460 Speaker 1: just recalling information or explaining a concept very, very good. 113 00:05:43,470 --> 00:05:43,919 Speaker 1: And in fact 114 00:05:44,060 --> 00:05:46,590 Speaker 1: right now there are many people, my friends and friends 115 00:05:46,589 --> 00:05:49,120 Speaker 1: of friends in the industry who are actually using chat 116 00:05:49,120 --> 00:05:53,680 Speaker 1: Gpt to write emails and to write even marketing materials. 117 00:05:53,690 --> 00:05:55,500 Speaker 1: But what's interesting is that some of them have been 118 00:05:55,500 --> 00:05:57,750 Speaker 1: playing with it for about a month and they started 119 00:05:57,750 --> 00:06:01,930 Speaker 1: to realize, oh, the AI can potentially replace me of 120 00:06:01,930 --> 00:06:05,290 Speaker 1: course when we talk about the more sophisticated, higher level 121 00:06:05,290 --> 00:06:07,790 Speaker 1: kinds of critical thinking skills, 122 00:06:07,950 --> 00:06:10,440 Speaker 1: it's a bit of a hit and miss with GPT 123 00:06:10,450 --> 00:06:13,210 Speaker 1: because it's like a student that has memorized model answers. 124 00:06:13,210 --> 00:06:15,490 Speaker 1: So if it's a very niche discipline, it may not 125 00:06:15,490 --> 00:06:18,680 Speaker 1: have encountered those essays to memorize. So the answers will 126 00:06:18,680 --> 00:06:22,110 Speaker 1: not be very sophisticated just scratching the surface. But it's 127 00:06:22,110 --> 00:06:27,110 Speaker 1: basically only as smart as the database. It's drawing information from. Right? 128 00:06:27,120 --> 00:06:29,469 Speaker 1: That's right. It couldn't replace you as a philosopher in 129 00:06:29,470 --> 00:06:30,390 Speaker 1: that sense because 130 00:06:30,570 --> 00:06:33,290 Speaker 1: some of your thinking or thoughts well, may not even 131 00:06:33,290 --> 00:06:37,520 Speaker 1: exist online. Right? Yes, that's right. But the worry is 132 00:06:37,520 --> 00:06:41,330 Speaker 1: that now we're using GPT three or 3.5 right? With 133 00:06:41,330 --> 00:06:45,610 Speaker 1: newer versions there is a possibility it could get better. 134 00:06:45,620 --> 00:06:47,830 Speaker 1: Is this the point where we talk about sentient AI, 135 00:06:47,839 --> 00:06:49,540 Speaker 1: you know, in other words when it can start to 136 00:06:49,540 --> 00:06:52,779 Speaker 1: think perceive feel, you know, the kind of stuff we've 137 00:06:52,779 --> 00:06:54,849 Speaker 1: seen in movies where robots come to life and take 138 00:06:54,850 --> 00:06:55,700 Speaker 1: over our world. 139 00:06:55,873 --> 00:06:59,253 Speaker 1: Could that ever happen? I think at the moment, maybe 140 00:06:59,253 --> 00:07:02,013 Speaker 1: not one of the interesting experiments I did was to 141 00:07:02,023 --> 00:07:04,573 Speaker 1: get it to write a reflection article. You know, some 142 00:07:04,573 --> 00:07:08,013 Speaker 1: people say that GPT can't write reflections, but if you 143 00:07:08,013 --> 00:07:10,813 Speaker 1: play with the prompts enough, you can actually get it 144 00:07:10,813 --> 00:07:13,653 Speaker 1: to write something that looks like a reflection. But the 145 00:07:13,653 --> 00:07:16,203 Speaker 1: question is is it really capable doesn't have a soul 146 00:07:16,203 --> 00:07:19,113 Speaker 1: and it's actually reflecting on itself. The answer is no, 147 00:07:19,113 --> 00:07:21,163 Speaker 1: it's using a lot of men 148 00:07:21,176 --> 00:07:23,776 Speaker 1: mathematics to find words that are relevant to the prom. 149 00:07:23,786 --> 00:07:29,246 Speaker 1: So it's giving us sophisticated, impressive answers without real understanding 150 00:07:29,246 --> 00:07:31,406 Speaker 1: of what it's talking about. And that's why you can 151 00:07:31,416 --> 00:07:36,076 Speaker 1: make mistakes, factual errors quite regularly. The capability to discern, 152 00:07:36,076 --> 00:07:39,406 Speaker 1: which is something human beings especially skilled at Toby a 153 00:07:39,406 --> 00:07:42,676 Speaker 1: lot of people are worried, especially educators because they've seen 154 00:07:42,676 --> 00:07:45,536 Speaker 1: some sample answers for questions. And so this is pretty decent. 155 00:07:45,536 --> 00:07:46,466 Speaker 1: It sounds pretty good. 156 00:07:46,700 --> 00:07:50,200 Speaker 1: So will this then lead to students cheating handing in 157 00:07:50,200 --> 00:07:54,030 Speaker 1: papers which are literally plagiarized to a certain extent. It 158 00:07:54,030 --> 00:07:56,380 Speaker 1: wasn't unusual to get someone to write your essay for you, 159 00:07:56,380 --> 00:07:58,580 Speaker 1: but now you only have to do that. It's not 160 00:07:58,580 --> 00:08:01,750 Speaker 1: likely to happen. It's already happened. That's a student in 161 00:08:01,750 --> 00:08:04,660 Speaker 1: north Carolina has just been given a fail grade for 162 00:08:04,670 --> 00:08:07,770 Speaker 1: handing in an essay in this philosophy class that was 163 00:08:07,770 --> 00:08:11,610 Speaker 1: copied lifted from chat. Gpt. So yes, students are if 164 00:08:11,620 --> 00:08:13,100 Speaker 1: we're not careful going to pick this up 165 00:08:13,310 --> 00:08:16,360 Speaker 1: and misuse it, it's also worth pointing out. There are 166 00:08:16,370 --> 00:08:20,660 Speaker 1: plentiful positive uses of the same technology in education. You 167 00:08:20,660 --> 00:08:23,910 Speaker 1: can use it to give students immediate feedback. Students can 168 00:08:23,910 --> 00:08:26,760 Speaker 1: use it real time to work out whether they've polished 169 00:08:26,760 --> 00:08:28,840 Speaker 1: the essay enough and what great they're likely to get 170 00:08:28,840 --> 00:08:31,430 Speaker 1: and what things the topics they haven't covered in their 171 00:08:31,430 --> 00:08:32,679 Speaker 1: current essay. It's very good 172 00:08:32,830 --> 00:08:35,730 Speaker 1: offering you that sort of feedback. And so there are 173 00:08:35,730 --> 00:08:40,070 Speaker 1: positive and like any technology negative uses. But in a society, 174 00:08:40,070 --> 00:08:42,380 Speaker 1: especially in a country like Singapore where rote learning has 175 00:08:42,380 --> 00:08:45,170 Speaker 1: been a large part of the way we learn academically 176 00:08:45,170 --> 00:08:48,240 Speaker 1: in Singapore. Which means if that's taken away, it almost 177 00:08:48,240 --> 00:08:51,490 Speaker 1: sounds like we've replaced road with robot. It sounds a 178 00:08:51,490 --> 00:08:54,329 Speaker 1: bit like an existential crisis just waiting to happen. 179 00:08:54,340 --> 00:08:57,830 Speaker 1: What do you think? It underlines some fundamental problems with 180 00:08:57,830 --> 00:09:02,040 Speaker 1: the education system if we're just testing things that require 181 00:09:02,050 --> 00:09:05,080 Speaker 1: only rote learning. Is that really the best way of 182 00:09:05,090 --> 00:09:07,620 Speaker 1: teaching these skills? I mean we're not asking people to 183 00:09:07,620 --> 00:09:09,820 Speaker 1: write essays because there's some shortage of essays in the 184 00:09:09,820 --> 00:09:12,500 Speaker 1: world were writing to answer essays so that they can 185 00:09:12,510 --> 00:09:16,590 Speaker 1: learn around the topic, read fully around a topic, learn 186 00:09:16,590 --> 00:09:18,820 Speaker 1: how to think critically and debate 187 00:09:19,070 --> 00:09:21,800 Speaker 1: two sides of an argument and those are things that 188 00:09:21,800 --> 00:09:25,050 Speaker 1: maybe we need to test in other ways than just 189 00:09:25,050 --> 00:09:28,439 Speaker 1: asking them to repeat essays. But in a way, learning 190 00:09:28,440 --> 00:09:31,739 Speaker 1: all those facts allowed me to then build the foundation 191 00:09:31,740 --> 00:09:35,510 Speaker 1: to write my essay or to build my thesis. So now, 192 00:09:35,510 --> 00:09:37,660 Speaker 1: if I don't even have to do that, does that 193 00:09:37,660 --> 00:09:40,370 Speaker 1: take away part of the learning experience? You put your 194 00:09:40,370 --> 00:09:42,620 Speaker 1: thumb on a really important point, which is that we 195 00:09:42,620 --> 00:09:44,660 Speaker 1: get people to write essays because 196 00:09:44,677 --> 00:09:46,527 Speaker 1: it is a good way to learn writing skills. It's 197 00:09:46,527 --> 00:09:49,487 Speaker 1: a good way to learn critical thinking. If we can 198 00:09:49,497 --> 00:09:52,136 Speaker 1: outsource a lot of that. Two machines, then how are 199 00:09:52,136 --> 00:09:54,477 Speaker 1: you going to learn those skills? I can imagine we 200 00:09:54,477 --> 00:09:57,367 Speaker 1: will pivot to a future like with calculators, I remember 201 00:09:57,367 --> 00:09:59,617 Speaker 1: the debate when I was back in school, but whether 202 00:09:59,617 --> 00:10:02,607 Speaker 1: we should use calculators in exams or not, and we 203 00:10:02,607 --> 00:10:05,577 Speaker 1: know eventually calculators win. And we do or get to 204 00:10:05,577 --> 00:10:09,097 Speaker 1: use calculators, at least not your first exams. But in 205 00:10:09,097 --> 00:10:10,380 Speaker 1: later exams universe 206 00:10:10,664 --> 00:10:13,424 Speaker 1: you can use a calculator anytime. But we did of course, 207 00:10:13,424 --> 00:10:15,594 Speaker 1: most of us had to sit exams when we were 208 00:10:15,604 --> 00:10:18,354 Speaker 1: in primary school and secondary school where we weren't allowed 209 00:10:18,364 --> 00:10:21,034 Speaker 1: to use calculators, we were taught those skills as a 210 00:10:21,034 --> 00:10:24,154 Speaker 1: basis for having a good mathematical grounding. And then once 211 00:10:24,154 --> 00:10:26,863 Speaker 1: we were capable of doing that, we were then allowed 212 00:10:26,864 --> 00:10:29,114 Speaker 1: to take calculators in so I can see the same 213 00:10:29,114 --> 00:10:32,134 Speaker 1: happening with these tours, which is that they won't be allowed. 214 00:10:32,144 --> 00:10:35,674 Speaker 1: Initially you'll have to learn those skills about writing and 215 00:10:35,674 --> 00:10:36,004 Speaker 1: about 216 00:10:36,111 --> 00:10:38,641 Speaker 1: critical thinking that you get from having to write an 217 00:10:38,641 --> 00:10:41,711 Speaker 1: essay yourselves and then at some later point you'll be 218 00:10:41,711 --> 00:10:43,981 Speaker 1: able to use these doors because it's worth pointing out 219 00:10:43,981 --> 00:10:45,251 Speaker 1: these tools are going to be out there in the 220 00:10:45,251 --> 00:10:48,601 Speaker 1: real world. Microsoft is one of the big backers between 221 00:10:48,601 --> 00:10:51,931 Speaker 1: open ai the company behind chat DBT, they're looking at 222 00:10:51,931 --> 00:10:56,011 Speaker 1: investing another 10 billion. They've already invest several billions in 223 00:10:56,011 --> 00:10:58,341 Speaker 1: chat GPT and therefore it's going to turn up in 224 00:10:58,341 --> 00:11:01,821 Speaker 1: all the Microsoft office products. So every time you open word, 225 00:11:02,000 --> 00:11:04,229 Speaker 1: it will be their offering advice. So in a way 226 00:11:04,230 --> 00:11:06,180 Speaker 1: we can't run away from it. It's here to stay, 227 00:11:06,179 --> 00:11:08,370 Speaker 1: it's not going to disappear just because we don't like it. 228 00:11:08,590 --> 00:11:10,860 Speaker 1: And in a way you're saying the foundation needs to 229 00:11:10,860 --> 00:11:12,930 Speaker 1: be there. So yeah, we need to know that two 230 00:11:12,929 --> 00:11:15,110 Speaker 1: plus two is four, so that if I go to 231 00:11:15,110 --> 00:11:17,620 Speaker 1: the shop and buy something for four bucks and I 232 00:11:17,620 --> 00:11:19,490 Speaker 1: give him 10, I know I should get six back. 233 00:11:19,490 --> 00:11:22,620 Speaker 1: But anything beyond that, why stress your brain that says 234 00:11:22,620 --> 00:11:24,600 Speaker 1: when you can just use a tool to help you 235 00:11:24,600 --> 00:11:27,839 Speaker 1: achieve the same outcome. As long as we don't become lazy, 236 00:11:27,850 --> 00:11:31,220 Speaker 1: but we are we have become lazy. Alright. It has 237 00:11:31,220 --> 00:11:33,250 Speaker 1: made us lazy for sure. Well I do have 238 00:11:33,260 --> 00:11:36,860 Speaker 1: an analogy for us to consider like for example, nowadays food, 239 00:11:36,860 --> 00:11:39,240 Speaker 1: we have vending machines, we have apps that can send 240 00:11:39,240 --> 00:11:41,800 Speaker 1: people to deliver food for us. Right? Why are we 241 00:11:41,800 --> 00:11:44,930 Speaker 1: still learning how to cook? We buy machines to make coffee, 242 00:11:44,929 --> 00:11:47,040 Speaker 1: but why we some of us we spend hundreds of 243 00:11:47,040 --> 00:11:50,790 Speaker 1: dollars to brew coffee ourselves. Why? Because there is something 244 00:11:50,790 --> 00:11:53,750 Speaker 1: about the creative aspects we want to create. We want 245 00:11:53,750 --> 00:11:56,679 Speaker 1: to consume things that are also, in a certain sense 246 00:11:56,679 --> 00:11:57,920 Speaker 1: our artisanal handmade 247 00:11:58,059 --> 00:12:01,520 Speaker 1: and in many ways learning this basic skills helps us 248 00:12:01,530 --> 00:12:04,570 Speaker 1: to create. And if we don't learn this, if we 249 00:12:04,570 --> 00:12:08,189 Speaker 1: just rely purely on chat GPT, we lose the ability 250 00:12:08,190 --> 00:12:11,280 Speaker 1: to create these things and or even take pride and 251 00:12:11,280 --> 00:12:13,550 Speaker 1: ownership in the things that we create. So that's one 252 00:12:13,550 --> 00:12:15,400 Speaker 1: of the things that's quite important. And 253 00:12:15,550 --> 00:12:18,050 Speaker 1: going back to the analogy of food, why do we 254 00:12:18,050 --> 00:12:20,309 Speaker 1: want to cook our own food? Because after a while 255 00:12:20,309 --> 00:12:24,590 Speaker 1: we get tired of store bought things. Mass produced fast food. However, right? 256 00:12:24,590 --> 00:12:27,390 Speaker 1: So sometimes we just want something a little homemade, so 257 00:12:27,390 --> 00:12:30,300 Speaker 1: to speak. And in many ways I teach 1000 plus 258 00:12:30,300 --> 00:12:33,090 Speaker 1: every year, Right? And one of the things that I've noticed, 259 00:12:33,100 --> 00:12:37,300 Speaker 1: for example, corporate professional text on websites, students already rejecting that, 260 00:12:37,300 --> 00:12:39,330 Speaker 1: they say, oh it's very inauthentic. It's very insincere, we 261 00:12:39,330 --> 00:12:39,850 Speaker 1: don't like that. 262 00:12:40,030 --> 00:12:43,209 Speaker 1: Likewise, chat GPT has a certain writing style and if 263 00:12:43,210 --> 00:12:46,309 Speaker 1: we expose people to this more and more people are 264 00:12:46,309 --> 00:12:48,809 Speaker 1: going to say, hey, you know, we don't really want 265 00:12:48,820 --> 00:12:52,069 Speaker 1: this mass produced kind of writing, we want something that 266 00:12:52,080 --> 00:12:55,699 Speaker 1: inspires that intrigues us something unique as well. So I 267 00:12:55,700 --> 00:12:59,100 Speaker 1: don't think that GPT is going to like completely wipe 268 00:12:59,100 --> 00:13:03,170 Speaker 1: out human writing completely. Okay, Jonathan, I have bad news 269 00:13:03,170 --> 00:13:06,270 Speaker 1: for you. There's going to be an aftermarket in different 270 00:13:06,270 --> 00:13:07,720 Speaker 1: skins for chat. Gpt, 271 00:13:07,740 --> 00:13:09,750 Speaker 1: you can ask it already to write in the style 272 00:13:09,750 --> 00:13:11,740 Speaker 1: of someone else. You can ask it to write in 273 00:13:11,740 --> 00:13:15,569 Speaker 1: the style of a Chicago gangster or or William Shakespeare 274 00:13:15,570 --> 00:13:19,120 Speaker 1: or whoever you ask and it does quite a convincing way. 275 00:13:19,130 --> 00:13:22,309 Speaker 1: So there'll be an aftermarket to write in your style, 276 00:13:22,320 --> 00:13:24,640 Speaker 1: your fine tune it with all the texts that you've 277 00:13:24,640 --> 00:13:28,530 Speaker 1: written in the past and now start speaking just like 278 00:13:28,530 --> 00:13:31,359 Speaker 1: you do writing just like it's almost like I create 279 00:13:31,360 --> 00:13:34,690 Speaker 1: my avatar, these are my characteristics, this is my writing 280 00:13:34,690 --> 00:13:35,450 Speaker 1: style 281 00:13:35,700 --> 00:13:38,940 Speaker 1: and I can make it mine, so to speak. Yeah. 282 00:13:38,950 --> 00:13:41,600 Speaker 1: And I imagine a lot of people won't, won't put 283 00:13:41,600 --> 00:13:44,190 Speaker 1: a little disclaimer at the bottom saying this reply to 284 00:13:44,190 --> 00:13:47,240 Speaker 1: your email was actually generated by Toby's avatar, Not by 285 00:13:47,240 --> 00:13:50,130 Speaker 1: Toby if you wish to speak to a real person, 286 00:13:50,130 --> 00:13:53,229 Speaker 1: press one. So how will I know if the real 287 00:13:53,230 --> 00:13:56,660 Speaker 1: Toby Walsh has written to me or not? You won't, 288 00:13:56,670 --> 00:14:00,000 Speaker 1: I mean they are tools already exist that try and 289 00:14:00,000 --> 00:14:04,880 Speaker 1: recognize computer synthetic chair text written by chatbots like chat, 290 00:14:05,120 --> 00:14:09,400 Speaker 1: but they're very easily fooled. There are statistical regularities in 291 00:14:09,400 --> 00:14:12,010 Speaker 1: what it writes as Jonathan said, it does sound a 292 00:14:12,010 --> 00:14:14,490 Speaker 1: little artificial if you if you know what to look 293 00:14:14,490 --> 00:14:16,790 Speaker 1: for but you could just ask Jack deputy to write 294 00:14:16,790 --> 00:14:20,320 Speaker 1: the text in some other voice. That's pretty convincing. Open 295 00:14:20,320 --> 00:14:25,930 Speaker 1: Ai are talking about adding some cryptographic watermarks to disguise 296 00:14:25,940 --> 00:14:28,440 Speaker 1: in the choice of words which synonyms are used in 297 00:14:28,440 --> 00:14:31,110 Speaker 1: the length of sentences in the length of words that 298 00:14:31,110 --> 00:14:31,670 Speaker 1: it's ca 299 00:14:31,690 --> 00:14:35,280 Speaker 1: computer generated in fact generated by chat. DBT. But again, 300 00:14:35,290 --> 00:14:38,500 Speaker 1: once we know what tools they've used cryptographic tools. Again, 301 00:14:38,500 --> 00:14:40,430 Speaker 1: it's not going to be hard to defeat those to 302 00:14:40,440 --> 00:14:43,120 Speaker 1: replace words with their synonyms with some other program or 303 00:14:43,120 --> 00:14:45,270 Speaker 1: to run it through another chapel. And if I may 304 00:14:45,270 --> 00:14:47,460 Speaker 1: add there's also the problem of what do we do 305 00:14:47,460 --> 00:14:49,820 Speaker 1: if we find false positives because like it or not 306 00:14:49,820 --> 00:14:52,450 Speaker 1: there are people who happen to write like an A. 307 00:14:52,450 --> 00:14:55,400 Speaker 1: I actually know as a friend of a colleague, a 308 00:14:55,400 --> 00:14:58,240 Speaker 1: scientist actually he tried playing with this detection tools and 309 00:14:58,260 --> 00:15:01,270 Speaker 1: one of the popular Ai detection tools actually flagged him 310 00:15:01,270 --> 00:15:05,350 Speaker 1: as 99%. Yes Jonathan, I bet you is not a 311 00:15:05,350 --> 00:15:08,470 Speaker 1: native english speaker. That may be the case. But there 312 00:15:08,480 --> 00:15:12,820 Speaker 1: is the possibility that this is the usual case of 313 00:15:12,830 --> 00:15:15,130 Speaker 1: Ai tools is that they will have biases and these 314 00:15:15,130 --> 00:15:17,490 Speaker 1: tools are probably going to be biased against non native 315 00:15:17,490 --> 00:15:20,680 Speaker 1: english speakers because they're having to be much more cognitive 316 00:15:20,680 --> 00:15:24,830 Speaker 1: about what they're saying. Probably saying it in more stilted style. 317 00:15:29,710 --> 00:15:32,950 Speaker 1: Hello everyone. My name is Christina and I'm Adrienne and 318 00:15:32,950 --> 00:15:35,350 Speaker 1: we're the host of a podcast called work it if 319 00:15:35,350 --> 00:15:37,330 Speaker 1: you've never heard of it. Well, it's a good time 320 00:15:37,330 --> 00:15:40,140 Speaker 1: to tap in. In the last 20 episodes. We've discussed 321 00:15:40,140 --> 00:15:43,330 Speaker 1: topics like how to negotiate for a salary increase or 322 00:15:43,330 --> 00:15:45,940 Speaker 1: how to get along with younger colleagues who have different 323 00:15:45,940 --> 00:15:47,070 Speaker 1: values from you, which 324 00:15:47,080 --> 00:15:50,840 Speaker 1: incidentally is our top performing episode. If what consumes your 325 00:15:50,840 --> 00:15:54,570 Speaker 1: life and you want some perspective on issues like management, stress, 326 00:15:54,580 --> 00:15:58,310 Speaker 1: even office romance, this podcast should be on your list. 327 00:15:58,320 --> 00:16:01,320 Speaker 1: A new episode drops every monday. Catch us on the 328 00:16:01,320 --> 00:16:04,450 Speaker 1: Sienna app or wherever you get your podcast, 329 00:16:06,810 --> 00:16:09,480 Speaker 1: the way it's evolving, Is it not just a matter 330 00:16:09,480 --> 00:16:12,980 Speaker 1: of time before these kinks will be worked out? Yes, 331 00:16:12,980 --> 00:16:16,210 Speaker 1: we will get to working out these kinks more and more. 332 00:16:16,220 --> 00:16:18,180 Speaker 1: I'm not sure if they will always go away completely, 333 00:16:18,180 --> 00:16:22,239 Speaker 1: but we certainly do a much better job. Certainly today, 334 00:16:22,250 --> 00:16:23,890 Speaker 1: these people are going to be discriminated against If we're 335 00:16:23,890 --> 00:16:28,670 Speaker 1: not careful, they're going to fail the computer human body test. 336 00:16:28,680 --> 00:16:31,650 Speaker 1: And one of the worries that we have as educators is, what, 337 00:16:31,665 --> 00:16:33,665 Speaker 1: what do we do if we get these kinds of 338 00:16:33,675 --> 00:16:36,545 Speaker 1: false positive cases because we can't tell them apart from 339 00:16:36,545 --> 00:16:39,455 Speaker 1: the people who actually cheat using GPT. So then what 340 00:16:39,455 --> 00:16:42,320 Speaker 1: kinds of disciplinary protocols or how do we even handle 341 00:16:42,320 --> 00:16:44,900 Speaker 1: these kind of situations? We don't want non native speakers 342 00:16:44,900 --> 00:16:48,410 Speaker 1: to get into trouble or under reputation just because they 343 00:16:48,410 --> 00:16:51,690 Speaker 1: happen to write like an Ai well, Jonathan, ultimately the 344 00:16:51,690 --> 00:16:54,800 Speaker 1: only answer, the only foolproof way of working out whether 345 00:16:54,800 --> 00:16:56,510 Speaker 1: people are using these tools to cheat 346 00:16:56,750 --> 00:16:59,930 Speaker 1: is to give them proper exam conditions, put them in 347 00:16:59,930 --> 00:17:02,440 Speaker 1: a vigil ated room, make sure they don't have access 348 00:17:02,440 --> 00:17:04,960 Speaker 1: to a paper and pencil and don't have access to 349 00:17:04,960 --> 00:17:07,430 Speaker 1: any computing devices. That's the only way, you know, one 350 00:17:07,430 --> 00:17:09,330 Speaker 1: of our worries that we have with going back to 351 00:17:09,340 --> 00:17:12,410 Speaker 1: close book final exams is that we lose a lot 352 00:17:12,410 --> 00:17:15,520 Speaker 1: of the educational developments and progress that we actually made 353 00:17:15,520 --> 00:17:16,270 Speaker 1: over the past 354 00:17:16,280 --> 00:17:18,620 Speaker 1: few years, especially with covid a lot of our E 355 00:17:18,619 --> 00:17:21,670 Speaker 1: learning blended learning. We've actually were able to make good 356 00:17:21,670 --> 00:17:24,590 Speaker 1: use of continuous assessments to push our students further. So, 357 00:17:24,590 --> 00:17:28,310 Speaker 1: imagine all that progress just set aside just because of 358 00:17:28,310 --> 00:17:32,219 Speaker 1: fears of cheating by AI is ironic to be taking 359 00:17:32,220 --> 00:17:35,130 Speaker 1: a computer science or AI related exam and using pen 360 00:17:35,130 --> 00:17:35,810 Speaker 1: and paper. 361 00:17:35,990 --> 00:17:38,590 Speaker 1: Okay, so we have this dilemma and that's where we're stuck. 362 00:17:38,590 --> 00:17:43,060 Speaker 1: So it's educators, what then how do we approach this situation? 363 00:17:43,070 --> 00:17:46,419 Speaker 1: I think we have to ultimately consider what exactly are 364 00:17:46,420 --> 00:17:49,220 Speaker 1: we teaching and how should we examine it? If it's 365 00:17:49,220 --> 00:17:52,130 Speaker 1: the sorts of skills that can only be examined by 366 00:17:52,140 --> 00:17:55,350 Speaker 1: an essay? I'm not sure it's the right thing because 367 00:17:55,359 --> 00:17:57,070 Speaker 1: in the real world, you will have access to these 368 00:17:57,070 --> 00:18:00,720 Speaker 1: doors equally closed book exams. Individual waiting rooms. Again, I'm 369 00:18:00,720 --> 00:18:01,310 Speaker 1: not sure you're 370 00:18:01,580 --> 00:18:04,090 Speaker 1: the right thing. I mean, that's testing always remember exams, 371 00:18:04,090 --> 00:18:06,889 Speaker 1: testing my ability to remember things well, I don't have 372 00:18:06,890 --> 00:18:08,600 Speaker 1: to remember so much that I can always look it 373 00:18:08,600 --> 00:18:11,139 Speaker 1: up as long as I know how to apply that knowledge. 374 00:18:11,150 --> 00:18:13,359 Speaker 1: But I think we need to think ultimately, are we 375 00:18:13,359 --> 00:18:16,100 Speaker 1: teaching the right skills and are we measuring them in 376 00:18:16,100 --> 00:18:18,960 Speaker 1: the appropriate way? One of the concerns that we, as 377 00:18:18,960 --> 00:18:23,020 Speaker 1: educators need to us is what kinds of learning outcomes 378 00:18:23,020 --> 00:18:26,450 Speaker 1: do we lose? If students use GPT or if we 379 00:18:26,450 --> 00:18:26,860 Speaker 1: embrace 380 00:18:26,960 --> 00:18:29,359 Speaker 1: GPT as a norm in education, what do they lose? 381 00:18:29,359 --> 00:18:31,660 Speaker 1: And the other question is how can we add value 382 00:18:31,660 --> 00:18:33,600 Speaker 1: to their learning? Because one of the things that we 383 00:18:33,600 --> 00:18:37,140 Speaker 1: recognize is that students are going out for internships in 384 00:18:37,150 --> 00:18:39,070 Speaker 1: the working world, They're going to see that people are 385 00:18:39,070 --> 00:18:41,660 Speaker 1: using GPT and then they're going to ask, why am 386 00:18:41,660 --> 00:18:44,380 Speaker 1: I coming to university to learn all these things if 387 00:18:44,390 --> 00:18:47,490 Speaker 1: people in the working world just using GPT to do work? Right. 388 00:18:47,500 --> 00:18:51,470 Speaker 1: So we also have to say, okay, this learning outcomes 389 00:18:51,470 --> 00:18:52,300 Speaker 1: that they will lose 390 00:18:52,310 --> 00:18:55,050 Speaker 1: if they embrace it, how else can we add value 391 00:18:55,060 --> 00:18:58,070 Speaker 1: so that they will not be replaced by an AI 392 00:18:58,080 --> 00:19:00,140 Speaker 1: or how can we add value so that they can 393 00:19:00,140 --> 00:19:02,970 Speaker 1: see the value of university education. So you're talking about 394 00:19:02,970 --> 00:19:04,899 Speaker 1: the value add, That's what it is. So I agree 395 00:19:04,910 --> 00:19:08,510 Speaker 1: these are tools that will exist. So is it not important? 396 00:19:08,510 --> 00:19:10,830 Speaker 1: We teach our young ones how to use these tools 397 00:19:10,830 --> 00:19:15,510 Speaker 1: effectively to do more, be more productive, get better answers 398 00:19:15,520 --> 00:19:17,750 Speaker 1: so that they don't waste time doing the legwork. 399 00:19:17,950 --> 00:19:20,060 Speaker 1: Can we do that? We can do that. But what 400 00:19:20,060 --> 00:19:21,460 Speaker 1: it means is that there's going to be a big 401 00:19:21,460 --> 00:19:23,810 Speaker 1: shift in the kinds of learning objectives that we want 402 00:19:23,810 --> 00:19:26,550 Speaker 1: for our students. Here's an example, it's not foolproof example, 403 00:19:26,550 --> 00:19:28,379 Speaker 1: but say for example, you want to teach them to 404 00:19:28,380 --> 00:19:31,220 Speaker 1: embrace that GPT and do better, right? We can get 405 00:19:31,220 --> 00:19:34,139 Speaker 1: them to generate an essay copy that into Microsoft word, 406 00:19:34,150 --> 00:19:36,830 Speaker 1: enable track changes and then we grade them on the 407 00:19:36,830 --> 00:19:39,889 Speaker 1: kinds of edits that they make. But you notice that's 408 00:19:39,900 --> 00:19:43,650 Speaker 1: editorial skills, that's my job. That's not essay writing skills already. 409 00:19:43,970 --> 00:19:47,480 Speaker 1: So the kinds of learning objectives will change will shift. 410 00:19:47,490 --> 00:19:50,190 Speaker 1: And it also means that the kinds of learning exercises 411 00:19:50,190 --> 00:19:52,830 Speaker 1: and activities will change. And that's why I say where 412 00:19:52,830 --> 00:19:55,770 Speaker 1: the start of a learning revolution the moment we say yes, 413 00:19:55,770 --> 00:19:58,140 Speaker 1: we're going to embrace their GPT, we need to teach 414 00:19:58,140 --> 00:20:01,150 Speaker 1: them higher skills because this means my degree. You know, 415 00:20:01,150 --> 00:20:04,580 Speaker 1: if I was based on a thesis, my 50 page 416 00:20:04,580 --> 00:20:06,860 Speaker 1: thesis for whatever my Master's program now, 417 00:20:06,880 --> 00:20:09,790 Speaker 1: it's like, is it no longer as valuable as what 418 00:20:09,790 --> 00:20:12,189 Speaker 1: it once was. It sounds like there's going to be 419 00:20:12,190 --> 00:20:16,120 Speaker 1: a fundamental shift in the way we embrace education and outcomes. 420 00:20:16,140 --> 00:20:19,150 Speaker 1: One of the comments I'm seeing quite regularly online now, 421 00:20:19,160 --> 00:20:21,970 Speaker 1: many students saying, hey, you know, I can learn better 422 00:20:21,980 --> 00:20:24,459 Speaker 1: by talking to chat GPT. In fact, some of them 423 00:20:24,460 --> 00:20:28,120 Speaker 1: are saying chat GPT teachers better than my teachers. I hope, 424 00:20:28,119 --> 00:20:29,770 Speaker 1: you know what I 425 00:20:30,320 --> 00:20:33,790 Speaker 1: so there is that concern, how can we as educators 426 00:20:33,800 --> 00:20:36,379 Speaker 1: add value to their learning so that they won't say 427 00:20:36,380 --> 00:20:37,810 Speaker 1: I can learn from a I don't need to go 428 00:20:37,810 --> 00:20:39,980 Speaker 1: to university or I just need to do is learn 429 00:20:39,980 --> 00:20:42,800 Speaker 1: how to write prompts. But the question that we must 430 00:20:42,810 --> 00:20:45,650 Speaker 1: examine is what can we do as educators to add 431 00:20:45,650 --> 00:20:49,210 Speaker 1: value to their learning so that we ourselves cannot be 432 00:20:49,210 --> 00:20:52,230 Speaker 1: replaced by an AI. Do you agree with that Toby Yeah, 433 00:20:52,230 --> 00:20:54,869 Speaker 1: I was reading someone who taught themselves python over a 434 00:20:54,869 --> 00:20:58,719 Speaker 1: couple of days by using chat GPT as their personalized tutor. 435 00:20:58,730 --> 00:21:01,300 Speaker 1: You could ask it questions doesn't matter how dumb or 436 00:21:01,310 --> 00:21:05,639 Speaker 1: how repetitive the questions were. It will patiently and quite 437 00:21:05,640 --> 00:21:09,560 Speaker 1: professionally answer those questions. So that's a fantastic example of 438 00:21:09,560 --> 00:21:12,640 Speaker 1: what you could use chat GPT give everyone their own 439 00:21:12,640 --> 00:21:15,900 Speaker 1: personalized tutor. Now, even a wealthy country like Singapore can't 440 00:21:15,900 --> 00:21:19,220 Speaker 1: afford to give every student their own full time personalized 441 00:21:19,220 --> 00:21:23,530 Speaker 1: tutor who can answer every question 24 7. So a 442 00:21:23,530 --> 00:21:26,880 Speaker 1: great opportunity that we should embrace. I guess the concern 443 00:21:26,880 --> 00:21:27,130 Speaker 1: is that 444 00:21:27,140 --> 00:21:29,520 Speaker 1: especially here in Singapore, having good grades is such a 445 00:21:29,520 --> 00:21:31,860 Speaker 1: big deal now. It's almost like people are a bit 446 00:21:31,859 --> 00:21:34,220 Speaker 1: worried house is going to affect the way people look 447 00:21:34,220 --> 00:21:36,670 Speaker 1: at me because they don't know how legit migrate is 448 00:21:36,670 --> 00:21:39,290 Speaker 1: whether it's based on something that I wrote or someone 449 00:21:39,290 --> 00:21:41,720 Speaker 1: else wrote, can this ever be controlled or do we 450 00:21:41,720 --> 00:21:43,830 Speaker 1: just have to embrace it and say let's go with 451 00:21:43,830 --> 00:21:46,879 Speaker 1: the flow and let's adapt to it instead of making 452 00:21:46,880 --> 00:21:49,780 Speaker 1: it adapt to us. I think we can try to 453 00:21:49,780 --> 00:21:52,960 Speaker 1: adapt to it and this is me speaking optimistically, it's 454 00:21:52,960 --> 00:21:55,540 Speaker 1: a great opportunity in education for us to move away 455 00:21:55,550 --> 00:21:59,690 Speaker 1: from purely the academics, purely looking at grades because if 456 00:21:59,690 --> 00:22:01,709 Speaker 1: we say, okay, check Gpt can do a lot of 457 00:22:01,710 --> 00:22:04,860 Speaker 1: these things, let's go for maybe like a more holistic 458 00:22:04,859 --> 00:22:08,200 Speaker 1: development portfolio rather than just focusing on grades. How can 459 00:22:08,200 --> 00:22:11,600 Speaker 1: we help our students to develop their portfolios so that 460 00:22:11,600 --> 00:22:14,220 Speaker 1: we can market them better in the workforce, that could 461 00:22:14,220 --> 00:22:17,330 Speaker 1: be one area which for now, eh I can't, can't 462 00:22:17,330 --> 00:22:20,690 Speaker 1: do too well here in Singapore, especially the private tuition 463 00:22:20,690 --> 00:22:23,950 Speaker 1: industry is huge. Talking about lots and lots of money. 464 00:22:23,960 --> 00:22:26,520 Speaker 1: This could kill that because if I no longer need 465 00:22:26,520 --> 00:22:30,920 Speaker 1: to memorize facts for exams. If every exam is kind 466 00:22:30,920 --> 00:22:33,590 Speaker 1: of like an open book exam, then you end up 467 00:22:33,590 --> 00:22:37,250 Speaker 1: trying to teach kids to think, which is way harder 468 00:22:37,250 --> 00:22:41,859 Speaker 1: than just saying, memorize these 50 answers. Maybe the question 469 00:22:41,859 --> 00:22:45,780 Speaker 1: also is our educators equipped and ready to be able 470 00:22:45,780 --> 00:22:47,750 Speaker 1: to do that as well because if we don't have 471 00:22:47,750 --> 00:22:51,629 Speaker 1: the right people teaching our youth these methods. Then again 472 00:22:51,640 --> 00:22:52,360 Speaker 1: we're all kind of 473 00:22:52,380 --> 00:22:54,680 Speaker 1: a little bit in limbo. This is why we need 474 00:22:54,680 --> 00:22:58,399 Speaker 1: to teach them more critical thinking skills because as we said, 475 00:22:58,410 --> 00:23:01,930 Speaker 1: there are biases in the training data and sometimes the 476 00:23:01,940 --> 00:23:04,550 Speaker 1: Ai can produce factual errors. So we need to train 477 00:23:04,550 --> 00:23:07,139 Speaker 1: our students to be able to spot these things to 478 00:23:07,140 --> 00:23:10,429 Speaker 1: be able to critically assess. Is the information presented to 479 00:23:10,430 --> 00:23:13,410 Speaker 1: me accurate? Is it sound so things like this? We 480 00:23:13,410 --> 00:23:15,700 Speaker 1: definitely need to teach them. I remember when I first 481 00:23:15,700 --> 00:23:17,700 Speaker 1: got this job, I had to do a current affairs 482 00:23:17,700 --> 00:23:19,649 Speaker 1: test to find out what was happening in the world. 483 00:23:19,650 --> 00:23:20,790 Speaker 1: The questions were like 484 00:23:20,940 --> 00:23:24,920 Speaker 1: In 1985, this event happened in Indonesia. What was it, 485 00:23:24,920 --> 00:23:27,149 Speaker 1: who was the president of this country at this time? 486 00:23:27,160 --> 00:23:30,000 Speaker 1: Oh my gosh, I did so badly, but I was 487 00:23:30,000 --> 00:23:31,370 Speaker 1: a bit of a rebel and I wrote I don't 488 00:23:31,369 --> 00:23:33,609 Speaker 1: know the answers, but I can find them out on 489 00:23:33,609 --> 00:23:36,540 Speaker 1: the job, fortunately I did get the job and I 490 00:23:36,540 --> 00:23:38,659 Speaker 1: hope my boss is not listening to that right now. 491 00:23:38,670 --> 00:23:41,090 Speaker 1: So based on all we've discussed and all we know 492 00:23:41,090 --> 00:23:44,030 Speaker 1: right now, Ai is evolving at a tremendous speed and 493 00:23:44,030 --> 00:23:45,200 Speaker 1: it is here to stay. 494 00:23:45,480 --> 00:23:48,840 Speaker 1: What would you say to people listening in now, what 495 00:23:48,840 --> 00:23:50,889 Speaker 1: kind of words of advice do you have for them 496 00:23:50,900 --> 00:23:54,109 Speaker 1: for our students, for our young workers? Well, I think 497 00:23:54,109 --> 00:23:56,180 Speaker 1: this is obviously going to have a big impact, not 498 00:23:56,180 --> 00:23:58,760 Speaker 1: just on education, but on work in general and what 499 00:23:58,760 --> 00:24:02,060 Speaker 1: jobs stay and what jobs get displaced by machines. If 500 00:24:02,060 --> 00:24:04,790 Speaker 1: you want advice as to where to focus your attention, 501 00:24:04,790 --> 00:24:06,030 Speaker 1: focus on those things 502 00:24:06,200 --> 00:24:11,850 Speaker 1: that distinguish you from machines, creativity, adaptability, flexibility, those critical 503 00:24:11,850 --> 00:24:14,889 Speaker 1: thinking skills that machines are not as good at and 504 00:24:14,890 --> 00:24:18,250 Speaker 1: that humans still have a unique edge over the machines. 505 00:24:18,260 --> 00:24:21,160 Speaker 1: One of the interesting trends that I see in the 506 00:24:21,160 --> 00:24:23,590 Speaker 1: young people these days is the way they experience the 507 00:24:23,590 --> 00:24:26,359 Speaker 1: internet is very different from the way we experience the internet. 508 00:24:26,369 --> 00:24:29,070 Speaker 1: Say for example, tell your crazy fact. Like bananas are 509 00:24:29,070 --> 00:24:32,330 Speaker 1: berries with large freakish seats. If you don't agree with me, 510 00:24:32,340 --> 00:24:34,700 Speaker 1: our instinct is to go check it on google, right? 511 00:24:34,880 --> 00:24:37,560 Speaker 1: But for many of the young people, their instinct is 512 00:24:37,560 --> 00:24:40,310 Speaker 1: to check it on Tiktok. A mother of a child. 513 00:24:40,310 --> 00:24:43,450 Speaker 1: Just just told me this today. Her daughter checks Tiktok 514 00:24:43,460 --> 00:24:45,970 Speaker 1: first rather than google. And a lot of our young 515 00:24:45,970 --> 00:24:47,969 Speaker 1: people are doing this and now many of them are 516 00:24:47,970 --> 00:24:51,570 Speaker 1: embracing chat Gpt now, why is this a problem? Because 517 00:24:51,580 --> 00:24:53,830 Speaker 1: at least with google we can see if I don't 518 00:24:53,830 --> 00:24:56,200 Speaker 1: like the first result. I'm not sure about the first result. 519 00:24:56,210 --> 00:24:59,070 Speaker 1: I can look at page two. Page three with GPT 520 00:24:59,070 --> 00:25:01,760 Speaker 1: were only given one answer and it is answered in 521 00:25:01,760 --> 00:25:01,810 Speaker 1: the 522 00:25:01,810 --> 00:25:06,629 Speaker 1: very authoritative manner. The danger is that we become very unquestioning. 523 00:25:06,640 --> 00:25:09,570 Speaker 1: We just accept the single narrative that's given to us. 524 00:25:09,580 --> 00:25:11,760 Speaker 1: So we need to learn critical thinking so that we 525 00:25:11,760 --> 00:25:14,090 Speaker 1: can be more discerning with the kinds of information presented 526 00:25:14,090 --> 00:25:16,649 Speaker 1: to us. Yeah, I like that idea. The sermon is 527 00:25:16,650 --> 00:25:18,530 Speaker 1: very important. I mean the same thing with the news 528 00:25:18,530 --> 00:25:21,680 Speaker 1: very often we tell people there's so many avenues to 529 00:25:21,680 --> 00:25:24,430 Speaker 1: hear about how something has happened. But it can be 530 00:25:24,430 --> 00:25:26,609 Speaker 1: fake news. And if you don't double check and cross 531 00:25:26,609 --> 00:25:28,750 Speaker 1: check with others, you can be misled. 532 00:25:29,150 --> 00:25:32,120 Speaker 1: So it sounds like we are entering or are already 533 00:25:32,119 --> 00:25:34,350 Speaker 1: in a world where that kind of critical thinking as 534 00:25:34,350 --> 00:25:37,630 Speaker 1: you're mentioned is increasingly important. Well gentlemen, thank you so 535 00:25:37,630 --> 00:25:40,580 Speaker 1: much for sharing your insights here with us. I mean 536 00:25:40,580 --> 00:25:43,650 Speaker 1: it sounds like science fiction but a world where robots 537 00:25:43,650 --> 00:25:46,600 Speaker 1: take over basic functions is already here. We see them 538 00:25:46,609 --> 00:25:50,100 Speaker 1: serving us our food and drinks, even playing chess and 539 00:25:50,100 --> 00:25:53,879 Speaker 1: in many respects technology and ai are largely improving the 540 00:25:53,880 --> 00:25:56,180 Speaker 1: quality and productivity of our lives. So 541 00:25:56,369 --> 00:25:59,580 Speaker 1: perhaps it's time to recalibrate the way we do things 542 00:25:59,580 --> 00:26:01,760 Speaker 1: and the way we look at technology gone are the 543 00:26:01,760 --> 00:26:05,510 Speaker 1: days of memorizing details because it is just a google 544 00:26:05,510 --> 00:26:07,920 Speaker 1: search away. Should we instead be teaching them how to 545 00:26:07,920 --> 00:26:10,359 Speaker 1: manage such tools and to find new ways to use 546 00:26:10,359 --> 00:26:12,580 Speaker 1: them more effectively. But there's no turning back and if 547 00:26:12,580 --> 00:26:12,689 Speaker 1: you 548 00:26:12,690 --> 00:26:15,020 Speaker 1: you want to write with that pen and paper, even 549 00:26:15,020 --> 00:26:17,520 Speaker 1: if you want to sometimes you can't because I've been 550 00:26:17,520 --> 00:26:19,320 Speaker 1: to places where there is no pen and paper and 551 00:26:19,320 --> 00:26:21,380 Speaker 1: the only option was to tap on the screen. Either 552 00:26:21,380 --> 00:26:23,429 Speaker 1: you do that or you don't get your food. So 553 00:26:23,440 --> 00:26:25,740 Speaker 1: I think we've given something for our folks to talk 554 00:26:25,740 --> 00:26:29,140 Speaker 1: about over the chinese new Year season this coming week. 555 00:26:29,150 --> 00:26:31,560 Speaker 1: On that note, we wish our listeners are very happy 556 00:26:31,560 --> 00:26:34,740 Speaker 1: lunar New year, enjoy your steamboat dinners. The team behind 557 00:26:34,760 --> 00:26:39,240 Speaker 1: this episode is Jacqueline, chan, Joanne, chan and Christina robert, 558 00:26:39,260 --> 00:26:41,180 Speaker 1: and I'm Stephen signing off.