WEBVTT - L’Oréal and IBM: AI-Powered Beauty 0:00:00.120 --> 0:00:04.280 Welcome to Tech Stuff, a production from iHeartRadio. This season, 0:00:04.280 --> 0:00:07.880 non Smart Talks with IBM, Malcolm Glabwell is back, and 0:00:07.920 --> 0:00:10.600 this time he's taking the show on the road. Malcolm 0:00:10.640 --> 0:00:14.680 is stepping outside the studio to explore how IBM clients 0:00:14.760 --> 0:00:18.799 are using artificial intelligence to solve real world challenges and 0:00:18.880 --> 0:00:23.000 transform the way they do business, from accelerating scientific breakthroughs 0:00:23.079 --> 0:00:27.520 to reimagining education. It's a fresh look at innovation in action, 0:00:28.000 --> 0:00:31.720 where big ideas meet cutting edge solutions. You'll hear from 0:00:31.720 --> 0:00:36.000 industry leaders, creative thinkers, and of course, Malcolm Glabwell himself 0:00:36.280 --> 0:00:39.920 as he guides you through each story. New episodes of 0:00:39.960 --> 0:00:43.280 Smart Talks with IBM drop every month on the iHeartRadio app, 0:00:43.440 --> 0:00:47.240 Apple Podcasts, or wherever you get your podcasts. Learn more 0:00:47.320 --> 0:00:49.920 at IBM dot com slash smart Talks. 0:00:52.560 --> 0:00:56.560 To understand why the cosmetics supergiant Lorel Group is teaming 0:00:56.680 --> 0:00:59.600 up with IBM, you must first take a closer look 0:00:59.680 --> 0:01:03.640 at its products. Take lipstick, for example. It's one of 0:01:03.640 --> 0:01:06.840 those things that seems straightforward, a waxy cylinder that you 0:01:06.920 --> 0:01:09.920 rub on your lips to turn them a different color. Easy, right, 0:01:10.480 --> 0:01:14.240 Well maybe not. As my colleague Lucy Sullivan found out 0:01:14.440 --> 0:01:17.560 when I sent her an assignment to Loreel's North America 0:01:17.600 --> 0:01:19.319 Research and Innovation Center. 0:01:20.400 --> 0:01:26.360 All right, I'm reporting live from the Loreal visitor's parking lot. 0:01:27.280 --> 0:01:31.039 Malcolm told me that he would be sending me to Paris, 0:01:31.160 --> 0:01:36.560 France for this Looreal excursion, but instead I am in Clark, 0:01:36.800 --> 0:01:39.720 New Jersey. Pass a lot of strip malls on the 0:01:39.760 --> 0:01:43.160 way here. But to be fair to Clark, New Jersey 0:01:43.240 --> 0:01:46.440 and Lorel, this is a beautiful compound. It kind of 0:01:46.480 --> 0:01:48.760 looks like a spa. 0:01:49.480 --> 0:01:53.040 Lucy went into the center and was blown away. The 0:01:53.080 --> 0:01:58.320 facility houses about six hundred scientists and experts across skincare, makeup, fragrance, 0:01:58.360 --> 0:02:02.440 hair care, innovative packaging, and tech. It is one of 0:02:02.520 --> 0:02:05.800 the largest formulation lab spaces in the industry. It's the 0:02:05.840 --> 0:02:10.280 size of six basketball courts. The reason Loreel's facility is 0:02:10.280 --> 0:02:13.280 so big and has so many people is that everything 0:02:13.320 --> 0:02:16.160 Loriel does to bring a product to market happens here, 0:02:16.800 --> 0:02:21.400 from molecule discovery and product development to consumer testing. The 0:02:21.440 --> 0:02:25.000 center even has its own mini factory. My conception of 0:02:25.040 --> 0:02:28.360 lipstick that it's just a waxy stick was plain wrong. 0:02:29.160 --> 0:02:33.000 Lipstick is a high performance product born from years of research, 0:02:33.320 --> 0:02:39.239 consumer insights and precision science. Lipstick isn't simple. It's incredibly complex, 0:02:39.800 --> 0:02:42.280 and one of the main reasons it's so complex is 0:02:42.360 --> 0:02:45.160 just a nature of fashion trends. The kind of lipstick 0:02:45.200 --> 0:02:47.680 consumers want is constantly changing. 0:02:48.240 --> 0:02:50.520 A lot of our consumer insights with Loril is like, 0:02:50.720 --> 0:02:52.520 where are consumers going in the future. 0:02:52.960 --> 0:02:56.920 This is Nadine Gomez, She's vice president for Loreel's research 0:02:56.960 --> 0:02:58.600 and innovation development. 0:02:58.160 --> 0:03:00.880 Team Park Chemist are working out five six years down 0:03:00.919 --> 0:03:04.600 the line. We predicted that consumers wanted more of a 0:03:04.680 --> 0:03:06.240 softer look on their lips as well. 0:03:06.360 --> 0:03:07.880 So how do you predict something like that. 0:03:08.440 --> 0:03:12.799 We see slow signals from fashion houses and social media 0:03:12.840 --> 0:03:14.720 and things like that. We kind of see that trend 0:03:14.760 --> 0:03:17.880 evolving a little bit, and then we know at five 0:03:17.960 --> 0:03:18.800 six years it's going. 0:03:18.720 --> 0:03:19.280 To become big. 0:03:20.280 --> 0:03:23.040 Lucy talked with her about the origins of one of 0:03:23.080 --> 0:03:26.639 their products, Mabe Lene matt Inc Liquid lipstick. 0:03:27.240 --> 0:03:29.560 Our competitors had two steps. The first step is a 0:03:29.560 --> 0:03:32.200 base coat. It's super opaque. You get the color and 0:03:32.240 --> 0:03:35.720 you get the maddy, but it's very very drying ellips. 0:03:35.800 --> 0:03:37.800 You cannot wear that, honestly more than ten minutes. It 0:03:37.840 --> 0:03:40.600 feels like your lips are like aching at one point, 0:03:40.840 --> 0:03:42.440 So we had to develop a top coat, and you'll 0:03:42.440 --> 0:03:44.200 see many of our competitors did the same thing. It's 0:03:44.200 --> 0:03:46.240 like a bomb. You put it on top, it's super comfortable. 0:03:46.880 --> 0:03:50.160 But we also noticed that consumers kind of get tired 0:03:50.360 --> 0:03:52.760 of reapplying a bomb. So we're like, what can we 0:03:52.800 --> 0:03:55.160 do to create this two step into one step? 0:03:55.920 --> 0:03:58.760 So Loriel had a challenge, how do you make a 0:03:58.840 --> 0:04:03.400 comfortable liquid map lipstick that doesn't require consumers to reapply 0:04:03.560 --> 0:04:07.040 a top layer of bomb. Solving this type of problem 0:04:07.240 --> 0:04:10.080 takes a lot of resources and a lot of expertise, 0:04:10.440 --> 0:04:14.640 and crucially, it takes time. Remember Nadine said that working 0:04:14.680 --> 0:04:17.640 on a breakthrough product such as matt inc can take 0:04:17.880 --> 0:04:22.560 years before it comes out. But can this process be accelerated, 0:04:23.000 --> 0:04:26.760 taken further, be even more sustainable. That's what IBM and 0:04:26.839 --> 0:04:32.120 Laurel are hoping to find out. My name is Malcolm Gladwell. 0:04:32.240 --> 0:04:35.400 You're listening to the latest episode of Smart Talks with IBM, 0:04:35.680 --> 0:04:38.200 where we offer our listeners a glimpse behind the curtain 0:04:38.240 --> 0:04:42.200 of the world of technology. In our last episode, we 0:04:42.240 --> 0:04:45.640 talked about how an AI assistant created with IBM Watson 0:04:45.839 --> 0:04:50.599 X helps future teachers practice responsive teaching by simulating classroom 0:04:50.680 --> 0:04:55.359 interactions with students. In this episode, we take you on 0:04:55.400 --> 0:04:59.880 an even more unexpected journey into the world of cosmetics, 0:05:00.240 --> 0:05:04.719 hair care, skincare, fragrance, makeup, and how a custom AI 0:05:04.839 --> 0:05:08.920 model could help Lorel's researchers shape the future of what 0:05:08.960 --> 0:05:18.840 we put on our faces every morning. I want to 0:05:18.839 --> 0:05:21.680 stay on lipstick a moment longer to help illustrate what 0:05:21.760 --> 0:05:25.719 goes into Loriel's product development, and let's focus on matt 0:05:25.800 --> 0:05:30.040 inc lipstick. Loreel wanted to create something that was comfortable 0:05:30.480 --> 0:05:32.080 and could be applied in one step. 0:05:32.680 --> 0:05:34.720 So to go from two step to one step, we 0:05:34.800 --> 0:05:37.200 had to look cross functionally and try to figure out 0:05:37.240 --> 0:05:39.640 what can we bring into the product to make it 0:05:39.640 --> 0:05:42.880 more comfortable, And luckily we have many different types of 0:05:42.880 --> 0:05:43.799 products at Lorel. 0:05:44.200 --> 0:05:47.440 That's Alex Good, a senior chemist who leads the lip 0:05:47.440 --> 0:05:50.640 products team in North America. She says the trick to 0:05:50.720 --> 0:05:55.040 making matt incwork was finding an elastomer, a substance they 0:05:55.040 --> 0:05:57.040 were already using in foundation. 0:05:57.960 --> 0:06:00.800 We have this elastomer that can give you like more 0:06:00.839 --> 0:06:03.520 comfortable and make it feel like there's like something on 0:06:03.520 --> 0:06:04.479 your lips like a cushion. 0:06:05.279 --> 0:06:08.800 She handed Lucy two jars. The first jar contained the 0:06:08.839 --> 0:06:11.320 former version of the product that was used in super 0:06:11.360 --> 0:06:14.760 State twenty four. By the way, this is exactly why 0:06:14.760 --> 0:06:18.799 I sent Lucy to the lab in my place the samples. 0:06:18.680 --> 0:06:20.480 And I actually have something for you to try here, 0:06:21.000 --> 0:06:25.080 so you can try this is what was the initial product. 0:06:25.400 --> 0:06:28.960 Okay, So this is like it sort of looks like Okay, 0:06:28.960 --> 0:06:31.520 it is clay. It looks like vacline that has like 0:06:31.560 --> 0:06:33.480 a more of a color. It's kind of a beige, 0:06:34.120 --> 0:06:36.880 looks like some skin. Okay. So this was from the 0:06:37.160 --> 0:06:43.040 two steps this would go on after oh okay, right violet. 0:06:44.040 --> 0:06:48.600 Okay, So it feels like very wet as you can see, 0:06:48.600 --> 0:06:51.400 it's kind of it's gonna absorb into your skin and 0:06:51.480 --> 0:06:54.000 leave and then you're gonna feel the dryness. 0:06:53.600 --> 0:06:56.080 Of the product once it's not okay. So we're gonna move. 0:06:56.040 --> 0:06:58.120 From the clay product that you have on your hand 0:06:58.160 --> 0:07:00.839 now to the last summer like you try half. 0:07:00.720 --> 0:07:04.320 One or This jar held the elastomer that Lareel has 0:07:04.360 --> 0:07:06.240 spent years developing in the lab. 0:07:07.279 --> 0:07:11.440 This one is a clear looks like aqua for a 0:07:11.560 --> 0:07:12.600 much player. 0:07:13.200 --> 0:07:16.720 And you can pay a physical layer that you're putting. 0:07:16.480 --> 0:07:16.920 On your aid. 0:07:17.160 --> 0:07:21.360 Yeah, so that's much thicker. It kind of like clumps together. Yeah, 0:07:22.360 --> 0:07:24.200 it was more of a cloudy it's less shimmery though 0:07:25.440 --> 0:07:26.120 that's intended. 0:07:26.360 --> 0:07:31.240 Yes, So this is a like a powder. This dispersed 0:07:31.360 --> 0:07:35.920 in dimethicone and it creates like a comfort on your lips. 0:07:35.920 --> 0:07:38.080 It feels like there's something there for a barrier to 0:07:38.160 --> 0:07:41.560 keep the film form on. And that's like the key 0:07:41.680 --> 0:07:44.880 ingredient that came from Foundation that we transferred into lipstick 0:07:44.960 --> 0:07:48.760 to give us this innovative product ahead of the market. Yeah, 0:07:48.960 --> 0:07:51.720 this is what gives it comfort. So the difference between 0:07:51.800 --> 0:07:54.720 super State twenty four and matt Inc is really the comfort. 0:07:54.760 --> 0:07:58.040 They both last a long time, but this matt Inc 0:07:58.120 --> 0:08:00.720 you don't have to apply the bomb over over again, 0:08:00.840 --> 0:08:02.800 so you can fly matting once for the day and 0:08:02.840 --> 0:08:03.240 you're good. 0:08:03.480 --> 0:08:07.840 For Alex Good is underselling it here, once for the 0:08:07.920 --> 0:08:12.120 day and you're good. That's a liquid lipstick revolution. Literally 0:08:12.360 --> 0:08:15.720 millions of loreal consumers around the world have worn matt ink. 0:08:15.960 --> 0:08:20.960 It's a blockbuster. It's also a marvel of science. The 0:08:21.000 --> 0:08:24.760 world's first liquid lipstick was developed in the nineteen thirties 0:08:24.800 --> 0:08:27.200 and it was actually just a stain for your lips, 0:08:27.800 --> 0:08:31.840 barely counts as lipstick. Then came another wave of liquid lipstick. 0:08:32.120 --> 0:08:34.520 When they were able to make it matt that was 0:08:34.520 --> 0:08:37.440 a two step version. It felt heavy on your lips. 0:08:37.840 --> 0:08:41.520 You had to keep reapplying the top coat. It was inconvenient. 0:08:42.679 --> 0:08:45.680 Loreel tackled that challenge in the lab with chemists like 0:08:45.800 --> 0:08:50.240 Alex and Nadine leaving the charge their breakthrough matt Inc. 0:08:51.080 --> 0:08:54.559 But creating matt Inc took a long time, trial and error, 0:08:54.800 --> 0:08:59.200 the hard work of scientific experimentation. As Nadean told Lucy, 0:08:59.520 --> 0:09:02.240 the lipstick team had to put the new product to 0:09:02.559 --> 0:09:03.600 extensive tests. 0:09:04.080 --> 0:09:07.000 We do a very robustability system here. You know, we 0:09:07.040 --> 0:09:10.400 have color odor appearance. We monitor this in extreme conditions. 0:09:10.760 --> 0:09:13.360 We simulate a forty five degrees celsius and that can 0:09:13.400 --> 0:09:15.559 be something like a three year shelf life. I'm saying 0:09:16.160 --> 0:09:19.079 we simulate your real life product. Like if you leave 0:09:19.120 --> 0:09:21.720 your lip gloss in the car in Arizona's one hundred 0:09:21.720 --> 0:09:23.520 and twelve degrees or three days, is it still going 0:09:23.559 --> 0:09:26.480 to perform? Is it gonna smell? Is it gonna look granted? 0:09:26.600 --> 0:09:28.160 Is it gonna change colors we. 0:09:28.080 --> 0:09:28.679 Do all that. 0:09:29.240 --> 0:09:33.120 See what I mean. Lipstick is complex. Most people would 0:09:33.120 --> 0:09:36.040 never consider it a piece of technology, but one lip 0:09:36.080 --> 0:09:39.160 product has millions of data points. 0:09:39.320 --> 0:09:41.080 So much science behind. And you can see here how 0:09:41.080 --> 0:09:43.400 many scientists we have. You know, some of them have PhDs, 0:09:43.400 --> 0:09:46.560 some of them have master's degrees to chemistry, biology, psychology. 0:09:46.640 --> 0:09:50.360 Also, when I first heard about this collaboration between LORI 0:09:50.559 --> 0:09:54.160 L and IBM, I was surprised. I thought, these are 0:09:54.200 --> 0:09:57.560 two very different companies. What do they really have in common? 0:09:57.920 --> 0:10:00.000 Pleasure, you guys, treasure. Yeah. 0:10:00.240 --> 0:10:02.439 To find out, I went to the IBM Research Center 0:10:02.440 --> 0:10:04.920 outside New York City, which I have to say is 0:10:04.960 --> 0:10:07.320 one of the coolest buildings I've ever been in, A 0:10:07.360 --> 0:10:11.960 semi circular modernist masterpiece with a long curving wall of windows, 0:10:12.160 --> 0:10:15.200 looks like something out of a Stanley Kubrick movie. I 0:10:15.280 --> 0:10:17.600 was there to talk with two experts from research and 0:10:17.640 --> 0:10:23.560 innovation at LOREL, Metheu Cassier and gabriel Bertoli. Matthew is 0:10:23.679 --> 0:10:27.920 VP for Digital and Transformation, Gabrielle is a Chief Digital 0:10:27.960 --> 0:10:32.439 Transformation Officer for Formulation. These are the people whose jobs 0:10:32.480 --> 0:10:36.160 are to oversee big changes within the company, and Methu 0:10:36.240 --> 0:10:38.000 told me to try on some lipstick. 0:10:38.440 --> 0:10:40.480 I'm gonna make you try this one. 0:10:40.679 --> 0:10:45.719 Okay, this is superstay vinitin final in. Yeah, so that's 0:10:45.720 --> 0:10:47.480 a glosse. Never in my life put on lipsey. I've 0:10:47.480 --> 0:10:48.840 no idea what I'm doing. You don't have to put it. 0:10:48.880 --> 0:10:49.800 You can try it virtually. 0:10:50.679 --> 0:10:54.120 Oh this may not be news to people who buy makeup, 0:10:54.360 --> 0:10:56.800 but it was news to me. You can try on 0:10:56.840 --> 0:11:00.120 loreal products virtually. They call it augmented beau. 0:11:01.240 --> 0:11:02.040 Oh my goodness. 0:11:02.720 --> 0:11:05.079 That is the strangest thing I've ever said. I look 0:11:05.160 --> 0:11:05.800 quite fetching. 0:11:06.559 --> 0:11:09.319 That's amazing. And I can just. 0:11:09.320 --> 0:11:11.600 Hit you can choose your color absolutely. 0:11:12.320 --> 0:11:13.680 So I'm on a little app. 0:11:13.880 --> 0:11:16.720 It's looking at me and it's just showing me exactly 0:11:16.720 --> 0:11:19.160 how I would look with different shades of lipstick. So 0:11:19.160 --> 0:11:20.880 the odd idea of going into a store and trying 0:11:20.920 --> 0:11:22.920 on each one, you cannot do that from home, if 0:11:22.920 --> 0:11:23.800 you're not even at the store. 0:11:24.320 --> 0:11:27.320 Yeah, absolutely, that's all purpose. If you want to manage 0:11:27.360 --> 0:11:29.440 a trend, I would go for something more like pitch. 0:11:30.480 --> 0:11:31.720 You think I'm a peach person. 0:11:32.679 --> 0:11:35.040 I don't know that looks I have to say, that 0:11:35.040 --> 0:11:37.680 looks kind of natural. It just is enhanced. It's given 0:11:37.720 --> 0:11:43.200 me a boys share I would not otherwise have. This 0:11:43.360 --> 0:11:47.640 is why Loreel says it creates beauty products and beauty experiences. 0:11:48.200 --> 0:11:52.000 Loriel is a beauty tech company. Over the last decade, 0:11:52.240 --> 0:11:55.120 Loreel has seized the power of AI and more recently, 0:11:55.480 --> 0:12:00.480 generative AI technology has become a driving force alongside science 0:12:00.559 --> 0:12:04.160 and creativity. And while some of this digital technology is 0:12:04.280 --> 0:12:08.360 relatively new, Matthew helped me see that IBM and Lorel 0:12:08.720 --> 0:12:10.720 have always had a lot in common. 0:12:11.440 --> 0:12:14.840 So the original creator of Loyal Legentulier was a chemist 0:12:14.840 --> 0:12:17.840 in nineteen or nine, so one hundred and sixteen years ago, 0:12:18.520 --> 0:12:21.680 and he created this new air color type for the 0:12:21.760 --> 0:12:25.120 market in France, and then little by little, it has 0:12:25.160 --> 0:12:27.400 been always a very scientific company. So if you look 0:12:27.400 --> 0:12:30.400 a little bit at key facts, we invented sun filters 0:12:30.720 --> 0:12:33.960 in the nineteen thirties, there was a very very big 0:12:34.000 --> 0:12:36.439 milestone where we also invented not only product, but a 0:12:36.520 --> 0:12:41.080 reconstructed skin. So if you look at nineteen seventeen nine, 0:12:41.679 --> 0:12:44.400 we've been the created this reconstructed skin that helped us 0:12:44.480 --> 0:12:46.600 to go out of animal testing very fast, and by 0:12:46.640 --> 0:12:49.600 the way, before the law even asked it to cosmetic companies, 0:12:50.200 --> 0:12:53.680 and then more recently, because it's a history of innovation, 0:12:53.760 --> 0:12:56.280 we'll launch on new molecules like one that you can 0:12:56.280 --> 0:12:59.400 find in laroche pose Melabi three, which is really helping 0:12:59.600 --> 0:13:03.320 people to find again some you know, spots they could 0:13:03.320 --> 0:13:05.600 have on their skin. It's all about like big mountation, 0:13:05.800 --> 0:13:06.640 how to regulate it. 0:13:07.360 --> 0:13:10.880 Loreel and IBM were both started in the early twentieth century, 0:13:11.280 --> 0:13:14.840 Loreel in nineteen oh nine and IBM in nineteen eleven. 0:13:15.360 --> 0:13:19.200 Both companies have long standing histories of innovation, of using 0:13:19.280 --> 0:13:22.360 trial and error to improve everything they do. The two 0:13:22.440 --> 0:13:25.240 companies have been doing that in parallel for more than 0:13:25.240 --> 0:13:29.439 a century until recently. When does it start? When do 0:13:29.600 --> 0:13:31.160 Lorel and IBM start working together? 0:13:32.040 --> 0:13:34.960 So we started in twenty twenty three, at the end 0:13:34.960 --> 0:13:37.080 of the year. But you know, really the discussion is 0:13:37.080 --> 0:13:41.680 really recent, absolutely, absolutely, it's really recent in reality. You know, 0:13:41.800 --> 0:13:45.120 I would say the first really interaction happened at the 0:13:45.120 --> 0:13:46.520 beginning of twenty twenty four. 0:13:47.320 --> 0:13:50.760 This is Gabriel Bertoli, who I spoke to alongside Matthew. 0:13:51.440 --> 0:13:54.640 What really played a key role here is we wanted 0:13:54.679 --> 0:13:58.880 to bring from a logic perspective to R and D together, 0:14:00.120 --> 0:14:03.560 which Normally, you know companies like us, you just go 0:14:03.679 --> 0:14:06.480 to a provider. You know, it's a customer and a 0:14:06.559 --> 0:14:09.120 supplier and new work they delivered to you. Here, the 0:14:09.160 --> 0:14:10.520 concept was totally different. 0:14:11.360 --> 0:14:14.840 Mid two said that the collaboration began with simple conversations. 0:14:15.160 --> 0:14:18.040 So if you look at the way IBM entered into 0:14:18.640 --> 0:14:22.880 Loreal Labs, it's started by interviewing people, what would help 0:14:22.920 --> 0:14:25.120 you to do your job? What is your business need? 0:14:25.680 --> 0:14:28.160 So it was, by the way, two months ago, a 0:14:28.320 --> 0:14:32.160 long series of interviews and from all the people around 0:14:32.200 --> 0:14:35.200 the world we have in research in Brazil, in India, 0:14:35.480 --> 0:14:40.160 in China, Japan, US, France of course, So we really 0:14:40.200 --> 0:14:41.760 want to make sure that at the end of the day, 0:14:42.160 --> 0:14:44.240 this new model, this new tool that we will give 0:14:44.280 --> 0:14:46.920 to people is really people c trick in the way 0:14:46.920 --> 0:14:48.520 that it selves their daily need. 0:14:49.120 --> 0:14:53.160 More the point, Lorel has leveraged technology for decades and 0:14:53.240 --> 0:14:58.720 accumulated amounted of scientific knowledge, everything from consumer aspirations and 0:14:58.760 --> 0:15:02.080 market trends, to the results of all the experiments conducted 0:15:02.160 --> 0:15:06.200 during product development, to which formulations melt in a hot car. 0:15:07.000 --> 0:15:10.040 It's hard to get your head around. Loreal isn't just 0:15:10.080 --> 0:15:14.080 a cosmetics company. It's a beauty data powerhouse. 0:15:15.040 --> 0:15:19.880 If we have sixteen thousand terabat of data coming from 0:15:20.200 --> 0:15:27.800 consumer insights, coming from market research, coming from sales, well 0:15:27.920 --> 0:15:32.760 with the new technology, maybe by aligning those two and 0:15:32.880 --> 0:15:35.880 using best in class technology you can solve that problem. 0:15:35.920 --> 0:15:38.680 So you say you have sixteen terabytes of data. Put 0:15:38.680 --> 0:15:40.600 that in perspective. How much data is that? 0:15:41.200 --> 0:15:41.520 Give me? 0:15:42.840 --> 0:15:46.800 This is one hundred year of Looreal data based on 0:15:47.000 --> 0:15:51.200 the last you know, forty years of data in the systems. 0:15:51.360 --> 0:15:53.280 So this is really I mean, we're talking about one 0:15:53.360 --> 0:15:56.280 hundred year of data that only Lorial have. Let's take 0:15:56.320 --> 0:15:58.520 the example of the ellipsis. I mean, you know, if 0:15:58.560 --> 0:16:02.240 ellipsex can be between twenty and thirty year ow material, 0:16:03.040 --> 0:16:05.800 each raw material will have I would say ten or 0:16:05.880 --> 0:16:10.440 fifteen way of doing things. 0:16:11.800 --> 0:16:14.240 Gabrielle is talking about how things used to be done. 0:16:14.560 --> 0:16:18.480 Researchers at Loreel needed roughly twenty five ingredients for a 0:16:18.520 --> 0:16:21.640 new lipstick formulation, but they have to choose from a 0:16:21.640 --> 0:16:25.960 pool of hundreds, if not thousands, of raw materials, and 0:16:26.040 --> 0:16:28.280 even after they settle on the ones they want, they 0:16:28.280 --> 0:16:30.680 have to figure out how much of each ingredient they 0:16:30.720 --> 0:16:35.080 need and in what form, what molecular weight, what combination. 0:16:35.800 --> 0:16:38.760 It's not just a math problem. It's a problem that 0:16:38.880 --> 0:16:46.400 requires balancing multiple perspectives safety, performance, quality, compliance standards, sustainability, 0:16:46.480 --> 0:16:50.280 and more. It can take years. But what if you 0:16:50.280 --> 0:16:54.160 could simulate hundreds of cars parked in a sweltering heat. 0:16:54.400 --> 0:16:56.960 What if you could do all those trials and errors 0:16:57.280 --> 0:17:01.280 virtually over and over and over again. What if instead 0:17:01.320 --> 0:17:05.080 of mixing materials together by hand, you could ask AI 0:17:05.200 --> 0:17:09.160 to predict what combinations might work best and then try 0:17:09.240 --> 0:17:10.159 those out first. 0:17:10.760 --> 0:17:14.920 This is ten on the power of twenty five. This 0:17:15.000 --> 0:17:19.480 is one hundred billion of years for a human to 0:17:19.600 --> 0:17:23.679 do a change in the formula or the possibility they have. 0:17:24.480 --> 0:17:29.480 You can only do this by using technology, power of 0:17:29.600 --> 0:17:31.080 technology and data that you have. 0:17:31.760 --> 0:17:35.000 This, Matthew says, is where IBM can come in to 0:17:35.119 --> 0:17:39.800 help take things further. Using artificial intelligence, IBM can help 0:17:39.840 --> 0:17:43.520 Lorio create a custom AI model that helps to crunch 0:17:43.560 --> 0:17:47.520 those numbers, to be a companion to the researchers, to 0:17:47.560 --> 0:17:48.640 give them superpowers. 0:17:48.880 --> 0:17:50.879 We don't want to replace the intuition of the sentis. 0:17:51.000 --> 0:17:53.959 We just want to make sure that this intuition is 0:17:54.000 --> 0:17:58.320 really augmented by some calculation poor that as Gabrielle said, 0:17:58.320 --> 0:18:00.880 then do those ten and the poor of twenty five 0:18:01.080 --> 0:18:04.360 solution and probably try this one, this one, this one, 0:18:04.560 --> 0:18:07.520 it looks like a better solution and Thentimately that's really 0:18:07.600 --> 0:18:09.680 the decision of the chemist to make it happen. 0:18:12.359 --> 0:18:12.719 Well. 0:18:13.000 --> 0:18:15.960 To make a predictive AI model that can give Lorel 0:18:16.040 --> 0:18:20.480 researchers those superpowers, you'd need that mountain of data, years 0:18:20.640 --> 0:18:24.480 worth of laboratory testing and all Loreal's data digitized and 0:18:24.560 --> 0:18:29.160 AI ready. You'd need to train artificial intelligence on everything 0:18:29.200 --> 0:18:31.679 the company has already done in order for it to 0:18:31.680 --> 0:18:33.120 predict what it could do. 0:18:33.600 --> 0:18:37.920 Loriial has one hundred years of course of data, fifty 0:18:38.040 --> 0:18:39.760 years of digitized EGDA. 0:18:40.440 --> 0:18:44.080 This is Mariam Ashuri, Senior director of Product Management for 0:18:44.160 --> 0:18:47.639 IBM Watson X. Loreal has the data and part of 0:18:47.680 --> 0:18:50.480 IBM's job is to help put that data to work, 0:18:50.960 --> 0:18:55.280 which involves ensuring data quality. Mariam talked about the concept 0:18:55.520 --> 0:18:56.960 of AI ready data. 0:18:57.640 --> 0:19:00.480 The sole purpose of this data engineering cuyper is to 0:19:00.560 --> 0:19:05.000 clean the data, and we call them AI ready data 0:19:05.280 --> 0:19:09.080 makes them ready to be consumed by AI. So basically 0:19:09.119 --> 0:19:12.840 looking into biases in the data to fix the distribution, 0:19:13.080 --> 0:19:16.400 looking into guard brains that we are putting into place 0:19:16.480 --> 0:19:19.040 in terms of removing personal information. 0:19:19.920 --> 0:19:22.960 Variam that explained that a custom model like the one 0:19:23.040 --> 0:19:26.240 IBM is creating with Lorel can be more efficient and 0:19:26.359 --> 0:19:29.640 targeted than the larger general purpose AI models. 0:19:29.880 --> 0:19:33.240 You've heard about large language models. The reason that they 0:19:33.280 --> 0:19:36.840 call them large language model is they are exposed into 0:19:38.200 --> 0:19:42.280 really large amount of data. So the larger the model, 0:19:42.320 --> 0:19:45.280 the more cake of all the models are, but also 0:19:45.720 --> 0:19:49.920 the larger computed requires that translate stand increase carbon footprint 0:19:50.000 --> 0:19:54.439 and energy consumption that translates stand, increase latency that's your 0:19:54.480 --> 0:19:58.719 response time that translatestand increase costs. So we started seeing 0:19:58.800 --> 0:20:04.679 that interprise started grabbing a much smaller model customize it 0:20:04.720 --> 0:20:09.160 on their proprietary data that's the data their DOMAINO specific data, 0:20:09.280 --> 0:20:13.080 or the data about their users to create something differentiated 0:20:13.000 --> 0:20:16.840 that is applicable to a real word use case but 0:20:17.000 --> 0:20:20.639 also delivers the performance that they needed for a fraction 0:20:20.720 --> 0:20:23.520 of the costs. And that's why there's been a lot 0:20:23.560 --> 0:20:27.879 of push around using custom models versus very large general 0:20:27.960 --> 0:20:29.040 purpose models. 0:20:29.600 --> 0:20:33.040 So how is a custom model created? Miriam says, you 0:20:33.080 --> 0:20:36.200 start with a base model. Imagine you're buying a car, 0:20:36.680 --> 0:20:38.840 You could get a minivan or a sedan or a 0:20:38.840 --> 0:20:41.720 sports car, and then you get to customize it. You 0:20:41.720 --> 0:20:45.000 could add a sunroof, leather seats, or a rearview camera. 0:20:45.480 --> 0:20:47.560 Turns out you could do the same thing with your 0:20:47.600 --> 0:20:51.119 AI model. You pick a base and then you customize it. 0:20:51.480 --> 0:20:54.560 You tune it on the data unique to your organization. 0:20:55.000 --> 0:20:58.880 We do believe that one model doesn't fit all use cases. 0:20:59.480 --> 0:21:03.480 You want to truly have access to any model anywhere, 0:21:03.560 --> 0:21:07.960 and by any model anywhere, I really mean any model anywhere, 0:21:08.040 --> 0:21:12.919 open source, proprietary, low call out your machine wherever the 0:21:12.960 --> 0:21:16.240 model is. You want to host it yourself, because then 0:21:16.600 --> 0:21:19.439 you would be able to take advantage of the best 0:21:19.480 --> 0:21:22.320 of the technology at any point and pick the right 0:21:22.359 --> 0:21:23.879 model for the target use case. 0:21:24.240 --> 0:21:27.639 So a custom model tuned on Lorel's data would be 0:21:27.680 --> 0:21:31.840 more targeted and efficient than a general purpose model. It 0:21:31.880 --> 0:21:37.040 would understand the researchers world and provide transparency into its workings. 0:21:37.440 --> 0:21:40.280 That's part of the magic. And what could a custom 0:21:40.400 --> 0:21:44.880 AI foundation model do for a company like Lorel if. 0:21:44.800 --> 0:21:48.520 You accord it was just moder is contain the complexity 0:21:49.240 --> 0:21:50.320 of the formulation. 0:21:50.960 --> 0:21:51.560 That's game. 0:21:51.680 --> 0:21:55.240 La Moline an IBM distinguished engineer and one of the 0:21:55.280 --> 0:21:57.280 people working on the AI model. 0:21:57.520 --> 0:22:03.159 And to hyperli the formulate or to go not only faster, 0:22:03.600 --> 0:22:08.000 but also I would say, be able to include more 0:22:08.040 --> 0:22:13.080 complexity or so in the formulation, more personalization, more certain ability, 0:22:13.640 --> 0:22:17.600 better selected ingredient. So it's really a tool to help 0:22:17.680 --> 0:22:22.040 them and to also help them to unniche the creativity. 0:22:25.119 --> 0:22:29.040 THEOMI saying that with its custom AI model, Lorel could 0:22:29.040 --> 0:22:32.639 improve every step of its product development pipeline, make the 0:22:32.680 --> 0:22:36.600 process faster and more sustainable. But he's also saying that 0:22:36.640 --> 0:22:39.560 the model could help Lorel create something that's never been 0:22:39.600 --> 0:22:40.280 done before. 0:22:40.880 --> 0:22:47.159 What could that product be? So I'm mourning you with that. 0:22:47.280 --> 0:22:49.240 All my questions are going to be really dumb. 0:22:50.040 --> 0:22:52.040