1
00:00:00,320 --> 00:00:03,880
Speaker 1: Education Minister Erica Stanford announcing that school exam marking will

2
00:00:03,920 --> 00:00:07,960
Speaker 1: be aided by the use of AI, taking the responsibility

3
00:00:07,960 --> 00:00:10,680
Speaker 1: away from teachers. Are we opening a can of worms here?

4
00:00:10,840 --> 00:00:14,080
Speaker 1: Danny Beddingfield is a learning development an AI specialist at

5
00:00:14,160 --> 00:00:16,320
Speaker 1: Cyclone Computer Company, joining us this morning.

6
00:00:16,400 --> 00:00:19,200
Speaker 2: Danny, good morning, Oh, good morning, How are you good?

7
00:00:19,239 --> 00:00:23,320
Speaker 1: Thank you? Would you trust AI to mark your kids' exams?

8
00:00:25,440 --> 00:00:26,119
Speaker 2: In a nutshell?

9
00:00:26,239 --> 00:00:26,599
Speaker 1: No?

10
00:00:26,600 --> 00:00:29,160
Speaker 2: No. Look, I do a lot of work with alongside

11
00:00:29,200 --> 00:00:31,920
Speaker 2: teachers and businesses around the use of AI, and there's

12
00:00:31,920 --> 00:00:33,800
Speaker 2: a lot of really great things that it can do

13
00:00:33,880 --> 00:00:36,800
Speaker 2: to save us time. But I've got real concerns around

14
00:00:37,120 --> 00:00:40,920
Speaker 2: transparency and around ethical use of AI, particularly the privacy

15
00:00:41,400 --> 00:00:44,479
Speaker 2: area as well, And I just feel like using it

16
00:00:44,520 --> 00:00:47,720
Speaker 2: to mark student work is really crossing a barrier that

17
00:00:47,760 --> 00:00:48,479
Speaker 2: we don't want to go to.

18
00:00:48,800 --> 00:00:51,159
Speaker 1: What's wrong with that? Because in my mind, I mean,

19
00:00:51,159 --> 00:00:53,920
Speaker 1: there's things like maths and science presumably be pretty straight

20
00:00:54,000 --> 00:00:57,480
Speaker 1: up and down English. Maybe if there's an essay writing competition,

21
00:00:57,560 --> 00:01:00,680
Speaker 1: it might not be as as workable, But what are

22
00:01:01,200 --> 00:01:03,440
Speaker 1: what's wrong with with AI doing it?

23
00:01:04,520 --> 00:01:06,240
Speaker 2: Well? The main the main problem I've got at the

24
00:01:06,240 --> 00:01:09,120
Speaker 2: moment is that all the AI sort of tools that

25
00:01:09,160 --> 00:01:13,280
Speaker 2: we're using are generally based on overseas data or overseas

26
00:01:13,360 --> 00:01:15,720
Speaker 2: large language models, and the problem of that, of course

27
00:01:15,800 --> 00:01:17,679
Speaker 2: is they're not based on anything to do with New Zealand,

28
00:01:17,720 --> 00:01:22,240
Speaker 2: and New Zealand's a very unique country to buy multicultural situation.

29
00:01:22,560 --> 00:01:24,360
Speaker 2: You know, obviously we've got the Treaty of Whitangui and

30
00:01:24,400 --> 00:01:27,000
Speaker 2: so on, and all these things are not taken into

31
00:01:27,040 --> 00:01:30,440
Speaker 2: consideration in the AI models that we're currently using, so

32
00:01:30,480 --> 00:01:32,480
Speaker 2: you know, we'd have to look at developing our own

33
00:01:32,520 --> 00:01:35,480
Speaker 2: models in here in Altieroa to make them more suitable

34
00:01:35,520 --> 00:01:38,880
Speaker 2: to the data that we're hoping to mark student work.

35
00:01:38,920 --> 00:01:42,360
Speaker 2: And then there's the whole idea around transparency with students

36
00:01:42,360 --> 00:01:44,640
Speaker 2: and things as well, which is just really really concerned

37
00:01:44,680 --> 00:01:44,959
Speaker 2: for me.

38
00:01:45,200 --> 00:01:48,440
Speaker 1: Things like maths and science though nothing to do with

39
00:01:48,480 --> 00:01:52,200
Speaker 1: the treaty, quite universal and same everywhere.

40
00:01:52,720 --> 00:01:55,720
Speaker 2: No, yeah, no, I agree there. But you know we're

41
00:01:55,800 --> 00:01:58,440
Speaker 2: using tools right now. You know, you can create forms

42
00:01:58,480 --> 00:02:01,480
Speaker 2: and things on most you know, Google and Microsoft tools

43
00:02:01,480 --> 00:02:04,200
Speaker 2: that we use in schools that will just simply mark

44
00:02:04,320 --> 00:02:06,520
Speaker 2: you know, kind of multi choice questions and things. That's fine,

45
00:02:06,520 --> 00:02:08,760
Speaker 2: we don't need AI for that, you know, where they're

46
00:02:08,760 --> 00:02:11,799
Speaker 2: going to be using AI is for judging tools or

47
00:02:11,880 --> 00:02:14,960
Speaker 2: judging assessments that, as you say, like essays and things

48
00:02:15,000 --> 00:02:18,919
Speaker 2: that have a lot more complications within how you judge

49
00:02:18,919 --> 00:02:20,960
Speaker 2: whether they are achieved or a mirror or a past

50
00:02:21,080 --> 00:02:22,480
Speaker 2: or whatever they're going to come up with with their

51
00:02:22,480 --> 00:02:23,880
Speaker 2: new marketing systems.

52
00:02:24,080 --> 00:02:27,040
Speaker 1: Okay, Danny, interesting, appreciate your time this morning. Danny Beddingfield

53
00:02:27,160 --> 00:02:30,280
Speaker 1: Learning Development and AI special ast a Cyclone computer company.

54
00:02:31,240 --> 00:02:34,200
Speaker 2: For more from Early Edition with Ryan Bridge. Listen live

55
00:02:34,360 --> 00:02:37,360
Speaker 2: to news Talks it Be from five am weekdays, or

56
00:02:37,400 --> 00:02:39,320
Speaker 2: follow the podcast on iHeartRadio