Bayer is using machine learning to predict the genetic makeup of crops, helping farmers speed up production.
Category
🤖
TechTranscript
00:00 [MUSIC PLAYING]
00:02 We used to cross existing seed varieties
00:05 to create millions of new progeny
00:08 and select for the trades that we care about and advance them
00:13 in the pipeline.
00:14 We used to call this era of breeding as selecting the best.
00:18 Over the last few years, we have transitioned
00:21 into designing the best.
00:23 We are using machine learning to predict
00:28 what kind of a genome or what kind of a seed variety
00:32 will give us the characteristics that we care about
00:36 in the environment they're going to be grown.
00:39 My name is Fani Chawale, and I lead the data sciences
00:43 within the plant breeding organization at Bayer.
00:47 We focus on developing new seed varieties for our row crops
00:52 as well as vegetable crops.
00:56 What we are trying to do is continue
00:58 to advance the trades that we care about the most.
01:03 Historically, that can be a very cumbersome process
01:05 and depend on your luck to find the ones
01:08 that you want to advance.
01:10 But with AI, you can get very targeted
01:12 and you can get very specific from very early on
01:15 in the process.
01:17 So that will save you both the resources,
01:19 and then it will also increase the probability of finding
01:22 the trade that you care about.
01:24 We're just trying to make sure that we are collecting
01:26 as much data across all of those different variables.
01:30 We use the genomic information of the seed varieties,
01:33 and we apply a machine learning model
01:36 to predict what their performance in the field
01:39 would look like.
01:41 So instead of testing these varieties in the field,
01:43 you're testing them by predicting.
01:47 What this means from a grower perspective,
01:49 ultimately, is they can get the seed varieties
01:52 that they care about one year in advance.
01:55 What it means from a bear perspective
01:57 is we are able to run this operation efficiently
02:00 because we haven't used one year of field testing
02:03 and the resources corresponding to that.
02:06 And if you think about it from a planet perspective,
02:08 I think we have been using this approach for about 10 years
02:11 now.
02:12 And in this 10 years, we have replaced close to 20,000 acres
02:17 of field testing just in North America.
02:20 So not having to run a field trial
02:22 and being able to use predicted information to advance material
02:27 in a breeding pipeline can lead to resource savings
02:31 across these different aspects, whether it
02:33 is land, water, nitrogen, or even crop protection.
02:37 If you look at our population, the numbers are increasing.
02:40 By 2050, we are looking at close to 10 billion people globally.
02:43 The arable land is decreasing.
02:45 In order to meet this demand and do it
02:48 in a way that is sustainable, we need
02:50 to continue to discover new seed products that
02:53 help us mitigate this challenge.
02:55 And that's what plant breeding is about.
02:58 (upbeat music)
03:00 [BLANK_AUDIO]