1 00:00:00,040 --> 00:00:03,240 Speaker 1: For forecasting and improvement in the forecast. Earth Sciences they've 2 00:00:03,240 --> 00:00:06,360 Speaker 1: got a new supercomputer they've called it Cascade, uses AI 3 00:00:06,840 --> 00:00:08,840 Speaker 1: to predict the weather patterns. The claims it can produce 4 00:00:08,840 --> 00:00:11,120 Speaker 1: a five day forecast as good as a two day forecast. 5 00:00:11,119 --> 00:00:13,119 Speaker 1: But then the next question, obviously, is a two day 6 00:00:13,160 --> 00:00:16,479 Speaker 1: forecast really any good? Chris Brandolino as new as principal 7 00:00:16,520 --> 00:00:18,520 Speaker 1: scientist and as well us Chris, good. 8 00:00:18,320 --> 00:00:20,720 Speaker 2: Morning, Hey, good morning, Mike. 9 00:00:20,960 --> 00:00:22,360 Speaker 1: Is a two day any good? 10 00:00:24,600 --> 00:00:25,759 Speaker 2: I think? So? Why not? 11 00:00:26,280 --> 00:00:30,720 Speaker 1: Well, because sometimes it's wrong, that's a. 12 00:00:30,680 --> 00:00:33,240 Speaker 2: Negative way of looking at things. A lot of times 13 00:00:33,240 --> 00:00:33,639 Speaker 2: it's right. 14 00:00:34,640 --> 00:00:38,479 Speaker 1: So what do we base the ratio on? If it's 15 00:00:38,560 --> 00:00:40,839 Speaker 1: right more than fifty percent, then it's more right than 16 00:00:40,880 --> 00:00:42,360 Speaker 1: it's wrong. So that's fair enough as it. 17 00:00:43,360 --> 00:00:45,880 Speaker 2: Well, look, the bottom line is their skill. There is 18 00:00:45,960 --> 00:00:49,000 Speaker 2: skill in weather forecasting. It's not perfect. I mean, part 19 00:00:49,000 --> 00:00:50,640 Speaker 2: of it is. You know, if you ask a person, 20 00:00:50,720 --> 00:00:52,559 Speaker 2: you know, what do you define a perfect forecast? If 21 00:00:52,560 --> 00:00:54,080 Speaker 2: I say it's going to be twenty one degrees in 22 00:00:54,120 --> 00:00:56,360 Speaker 2: Auckland in the heads of twenty two, is that accurate? 23 00:00:56,440 --> 00:00:57,960 Speaker 2: I would think that's pretty dark accurate, you know what 24 00:00:58,000 --> 00:00:58,240 Speaker 2: I mean? 25 00:00:58,480 --> 00:01:01,480 Speaker 1: The scale up we've got here instication. Is this just 26 00:01:01,520 --> 00:01:04,640 Speaker 1: going to go exponential as II builds and grows? 27 00:01:05,640 --> 00:01:07,479 Speaker 2: Yeah, I think the big thing I'll guess I'll answer 28 00:01:07,480 --> 00:01:09,560 Speaker 2: it this one. I'm not a supercomputer nerd. I wish 29 00:01:09,600 --> 00:01:12,200 Speaker 2: it was, but anonymal weather nerd. I think one of 30 00:01:12,280 --> 00:01:14,560 Speaker 2: the main things to kind of tell you and your 31 00:01:14,600 --> 00:01:17,959 Speaker 2: listeners is that this supercomputer Cascade is about three times 32 00:01:18,000 --> 00:01:22,120 Speaker 2: more powerful than its predecessor. So it does computing speeds 33 00:01:22,160 --> 00:01:25,039 Speaker 2: of two point four ready for this petaflops And you're 34 00:01:25,040 --> 00:01:28,240 Speaker 2: probably wondering, what the heck is a petaflop? So imagine 35 00:01:28,319 --> 00:01:32,760 Speaker 2: if the entire Earth, about eight billion people, everyone was 36 00:01:32,800 --> 00:01:36,960 Speaker 2: doing three hundred three hundred thousand calculations per second at 37 00:01:36,959 --> 00:01:39,920 Speaker 2: the same time. That's the capability of the supercomputer. And 38 00:01:39,959 --> 00:01:42,400 Speaker 2: because and the reason we need that MIC is because 39 00:01:42,520 --> 00:01:45,600 Speaker 2: data is become king or queen I suppose, and with 40 00:01:45,760 --> 00:01:48,400 Speaker 2: so much more data coming in, we need to leverage 41 00:01:48,400 --> 00:01:50,800 Speaker 2: that data. I think the analogy I've been using, which 42 00:01:50,840 --> 00:01:53,520 Speaker 2: may resonate, is that imagine a restaurant getting a brand 43 00:01:53,520 --> 00:01:57,520 Speaker 2: new kitchen, and that kitchen really increases the capability. The 44 00:01:57,600 --> 00:02:00,000 Speaker 2: auven cooks faster, so you get the dishes out quick 45 00:02:00,120 --> 00:02:02,480 Speaker 2: or you have better tools things like that, so you're 46 00:02:02,520 --> 00:02:04,840 Speaker 2: able to operate more effectively and do different things that 47 00:02:04,920 --> 00:02:06,240 Speaker 2: maybe you wouldn't be able to do before. 48 00:02:06,440 --> 00:02:09,799 Speaker 1: Yeah, what's maximum value from a commercial scalable point of view? 49 00:02:09,840 --> 00:02:11,520 Speaker 1: Given what AI is going to do is a five 50 00:02:11,560 --> 00:02:13,800 Speaker 1: day forecast, seven to eighteen day, one month, team. 51 00:02:13,680 --> 00:02:16,400 Speaker 2: Year, what I think it's all of that, So I 52 00:02:16,440 --> 00:02:17,840 Speaker 2: guess to take you on a bit of a journey. 53 00:02:17,840 --> 00:02:19,680 Speaker 2: So one of the things we can do now with 54 00:02:19,800 --> 00:02:22,320 Speaker 2: our high res model, which goes out to your favorite 55 00:02:22,320 --> 00:02:25,519 Speaker 2: time span. It sounds like two days. That used to 56 00:02:25,600 --> 00:02:28,240 Speaker 2: run on the old supercomputer, it will take one hundred 57 00:02:28,280 --> 00:02:31,959 Speaker 2: and eighty minutes. Now it takes eighty minutes, so that's better. Okay, 58 00:02:32,000 --> 00:02:36,320 Speaker 2: what about longer range predicting? So most times when people 59 00:02:36,360 --> 00:02:37,880 Speaker 2: go to an app, they you know, look at their 60 00:02:37,880 --> 00:02:40,400 Speaker 2: favorite weather app. That's what we call deterministic. That's simply 61 00:02:40,400 --> 00:02:43,440 Speaker 2: one outcome and that's okay, But in reality, there's a 62 00:02:43,440 --> 00:02:45,280 Speaker 2: lot of outcomes that could happen. So we have an 63 00:02:45,360 --> 00:02:49,120 Speaker 2: ensemble model that has eighteen different outcomes. We're going to 64 00:02:49,200 --> 00:02:52,960 Speaker 2: expand the rectangle it looks at over New Zealand to 65 00:02:53,040 --> 00:02:55,360 Speaker 2: go up to the Pacific to New Caledoni, out to 66 00:02:55,400 --> 00:02:58,520 Speaker 2: Aussie Melbourne and down to the southern Ocean, so better, 67 00:02:58,720 --> 00:03:01,400 Speaker 2: I guess, larger area where monitoring in terms of modeling 68 00:03:01,440 --> 00:03:03,840 Speaker 2: with our ensemble, and then over the next year or so, 69 00:03:03,840 --> 00:03:05,799 Speaker 2: we're going to be looking to increase that from five 70 00:03:05,880 --> 00:03:08,840 Speaker 2: days to ten days, so we're gaining not only areas 71 00:03:08,840 --> 00:03:12,320 Speaker 2: to get insight on, but also the time ten days 72 00:03:12,320 --> 00:03:16,320 Speaker 2: and then there's longer longer rains forecasting five weeks, six weeks, 73 00:03:16,480 --> 00:03:18,760 Speaker 2: and getting a better understanding of what themes may be 74 00:03:18,840 --> 00:03:21,240 Speaker 2: come in their way for I think we're planning and 75 00:03:21,360 --> 00:03:23,880 Speaker 2: getting some heads up as to hey, in three or 76 00:03:23,919 --> 00:03:26,000 Speaker 2: four weeks there's a signal that it could be really 77 00:03:26,120 --> 00:03:28,960 Speaker 2: dry or on the other end of the spectrum, really wet, 78 00:03:29,120 --> 00:03:33,280 Speaker 2: and then monitoring that and getting people prepared for potentially 79 00:03:33,320 --> 00:03:36,480 Speaker 2: some large scale big weather events or things like dryness 80 00:03:36,520 --> 00:03:36,960 Speaker 2: or drought. 81 00:03:37,120 --> 00:03:39,800 Speaker 1: Fantastic like your passion. Chris Brandelina, who's the need of 82 00:03:39,880 --> 00:03:41,480 Speaker 1: principal scientists with us this morning. 83 00:03:41,960 --> 00:03:44,840 Speaker 2: For more from the Mic Hosking Breakfast, listen live to 84 00:03:44,960 --> 00:03:48,040 Speaker 2: news talks it'd be from six am weekdays, or follow 85 00:03:48,080 --> 00:03:49,600 Speaker 2: the podcast on iHeartRadio.