1
00:00:00,080 --> 00:00:02,040
Speaker 1: A little bit of help coming to the health system's way.

2
00:00:02,080 --> 00:00:04,480
Speaker 1: We think it's a first of a can't study scientists

3
00:00:04,640 --> 00:00:07,240
Speaker 1: work out how to predict surges in demand for hospitals,

4
00:00:07,280 --> 00:00:10,039
Speaker 1: think winter flu spikes and stuff like that. Stefan Albrush

5
00:00:10,320 --> 00:00:12,720
Speaker 1: from the School of Computer Science at Auckland University with

6
00:00:12,800 --> 00:00:16,919
Speaker 1: us on the Stephan morning to you. Hello, how generative

7
00:00:17,000 --> 00:00:18,720
Speaker 1: is the AI that you're using in this?

8
00:00:21,239 --> 00:00:23,560
Speaker 2: Oh? Pardon, can you please a repeat the question?

9
00:00:23,880 --> 00:00:26,440
Speaker 1: How much AI is used in the prediction of this?

10
00:00:27,920 --> 00:00:31,360
Speaker 2: Ah? So, in the current study, we are using the

11
00:00:31,400 --> 00:00:36,000
Speaker 2: state of the art AI algorithms, you could say. So

12
00:00:36,680 --> 00:00:40,920
Speaker 2: the two algorithms we focused a bit about are based

13
00:00:40,920 --> 00:00:46,400
Speaker 2: on artivision. You're a network landing. So this is quite

14
00:00:46,800 --> 00:00:50,400
Speaker 2: the state of the art concept that is used at

15
00:00:50,440 --> 00:00:51,080
Speaker 2: the moment here.

16
00:00:51,360 --> 00:00:54,400
Speaker 1: Okay, So how much of it is information fed in

17
00:00:55,080 --> 00:00:58,560
Speaker 1: from past patterns that help predict future patterns or is

18
00:00:58,600 --> 00:01:00,680
Speaker 1: it all literally generate from nothing?

19
00:01:02,320 --> 00:01:06,520
Speaker 2: I know, it's really data driven. It's really data based.

20
00:01:08,200 --> 00:01:12,960
Speaker 2: We're using data that goes back to twenty twelve and

21
00:01:14,240 --> 00:01:17,160
Speaker 2: the main job of these algorithms is to, you know,

22
00:01:17,520 --> 00:01:21,360
Speaker 2: take this data and try to find patterns, recurring patterns

23
00:01:22,600 --> 00:01:26,679
Speaker 2: that can be used again to match them to what

24
00:01:26,800 --> 00:01:30,000
Speaker 2: happens now to get a prediction for the near future.

25
00:01:30,480 --> 00:01:32,959
Speaker 1: Is there COVID in the numbers? And if there's COVID

26
00:01:32,959 --> 00:01:35,039
Speaker 1: in the numbers, that'll skew the numbers, won't it.

27
00:01:37,080 --> 00:01:41,080
Speaker 2: That's a really interesting question I'm also very curious about.

28
00:01:41,200 --> 00:01:45,000
Speaker 2: So in the current study, we focused on data from

29
00:01:45,680 --> 00:01:49,600
Speaker 2: we always say pre pandemic, So we focused on everything

30
00:01:49,600 --> 00:01:54,880
Speaker 2: we got from twenty twelve to twenty nineteen, and we

31
00:01:54,960 --> 00:02:01,160
Speaker 2: are currently working on integrating data from post pandemic era.

32
00:02:03,240 --> 00:02:08,040
Speaker 2: And so first tests will run with having COVID numbers

33
00:02:08,280 --> 00:02:11,320
Speaker 2: in data, but we can then also test what happens

34
00:02:11,320 --> 00:02:14,000
Speaker 2: when we leave them out. And it's actually an interesting

35
00:02:14,200 --> 00:02:17,040
Speaker 2: question we are wanting to answer with the currents leg

36
00:02:17,120 --> 00:02:22,800
Speaker 2: with ongoing studies is yeah, how how will it change

37
00:02:24,240 --> 00:02:28,440
Speaker 2: when we find off keep the numbers in or what

38
00:02:28,520 --> 00:02:32,360
Speaker 2: would happen when we remove the COVID cases? Yeah, yeah, exactly,

39
00:02:32,400 --> 00:02:35,960
Speaker 2: You're right, that's a that's a huge change when you

40
00:02:36,160 --> 00:02:40,600
Speaker 2: think about the data collected from before that these two years,

41
00:02:40,639 --> 00:02:42,959
Speaker 2: right and after these two years exactly percistly.

42
00:02:43,000 --> 00:02:44,600
Speaker 1: That's why we'll stay in touch with Stephan see how

43
00:02:44,639 --> 00:02:47,520
Speaker 1: it goes for your Stephanol Bridge our University of Oakland

44
00:02:47,520 --> 00:02:48,600
Speaker 1: School of Computer Science.

45
00:02:49,240 --> 00:02:52,120
Speaker 2: For more from the mic Asking Breakfast, listen live to

46
00:02:52,240 --> 00:02:52,800
Speaker 2: news talks.

47
00:02:52,800 --> 00:02:56,000
Speaker 1: It'd be from six am weekdays, or follow the podcast

48
00:02:56,040 --> 00:02:56,919
Speaker 1: on iHeartRadio