WEBVTT - Smart Talks with IBM: How open source can democratize AI 0:00:04.519 --> 0:00:12.719 Welcome to Tech Stuff, a production from iHeartRadio. Today, we 0:00:12.760 --> 0:00:15.720 are witnessed to one of those rare moments in history, 0:00:16.079 --> 0:00:19.319 the rise of an innovative technology with the potential to 0:00:19.440 --> 0:00:24.160 radically transform business in society forever. That technology, of course, 0:00:24.640 --> 0:00:28.200 is artificial intelligence, and it's the central focus for this 0:00:28.360 --> 0:00:32.400 new season of Smart Talks with IBM. Join hosts from 0:00:32.400 --> 0:00:36.159 your favorite Pushkin podcasts as they talk with industry experts 0:00:36.159 --> 0:00:39.760 and leaders to explore how businesses can integrate AI into 0:00:39.800 --> 0:00:43.120 their workflows and help drive real change in this new 0:00:43.200 --> 0:00:46.879 era of AI, and of course, host Malcolm Gladwell will 0:00:46.920 --> 0:00:49.200 be there to guide you through the season and throw 0:00:49.320 --> 0:00:52.199 in his two cents as well. Look out for new 0:00:52.240 --> 0:00:55.120 episodes of Smart Talks with IBM every other week on 0:00:55.160 --> 0:00:59.400 the iHeartRadio app, Apple Podcasts, wherever you get your podcasts, 0:00:59.600 --> 0:01:13.000 and learn more at IBM dot com, slash smart Talks. 0:01:10.720 --> 0:01:19.880 Pushkin Hello, Hello, Welcome to Smart Talks with IBM, a 0:01:19.959 --> 0:01:25.040 podcast from Pushkin Industries, iHeartRadio and IBM. I'm Malcolm Gladwell. 0:01:25.560 --> 0:01:29.080 This season, we're diving back into the world of artificial intelligence, 0:01:29.360 --> 0:01:32.720 but with a focus on the powerful concept of open 0:01:33.120 --> 0:01:38.280 its possibilities, implications, and misconceptions. We'll look at openness from 0:01:38.360 --> 0:01:41.360 a variety of angles and explore how the concept is 0:01:41.440 --> 0:01:45.000 already reshaping industries, ways of doing business, and a very 0:01:45.120 --> 0:01:48.880 notion of what's possible. In today's episode, I sat down 0:01:48.880 --> 0:01:52.720 with Mo Duffy, software engineering manager at red Hat, who 0:01:52.760 --> 0:01:56.440 works on instruct Lab, a project co developed by red 0:01:56.440 --> 0:02:00.040 Hat and IBM. Most shared with me how this a 0:02:00.080 --> 0:02:04.680 new initiative, is revolutionizing AI training, making it not only 0:02:04.720 --> 0:02:09.400 more accessible, but also more inclusive. This project, unique in 0:02:09.440 --> 0:02:14.280 the industry, allows developers to submit incremental contributions to one 0:02:14.400 --> 0:02:18.760 base AI model, creating a continuous loop of development, much 0:02:18.880 --> 0:02:24.200 like normal open source software. By leveraging community contributions and 0:02:24.360 --> 0:02:27.560 IBM's cutting edge granite models, Mo in the team of 0:02:27.600 --> 0:02:30.680 ibmrs and red hatters are paving the way for a 0:02:30.720 --> 0:02:35.880 future where AI development is a communal endeavor. Our insights 0:02:35.919 --> 0:02:40.400 into open source software extend beyond technical proficiency to the 0:02:40.440 --> 0:02:44.800 profound impact of collaborative effort. At the heart of Moe's 0:02:44.840 --> 0:02:49.400 work is a belief in democratizing technology, ensuring that AI 0:02:49.480 --> 0:02:54.560 becomes a tool accessible to all. So let's explore how Moe, 0:02:54.720 --> 0:02:58.680 red Hat and IBM are empowering individuals and businesses alike 0:02:59.000 --> 0:03:03.400 to reshape the fuel future of technology through collaboration and innovation. 0:03:08.440 --> 0:03:11.320 We thank you for joining me today. Thank you so 0:03:11.400 --> 0:03:15.400 much for I have just about the most Irish name ever. 0:03:16.240 --> 0:03:21.079 I do very proud you weren't born in Ireland my grandparents, 0:03:21.160 --> 0:03:23.040 Oh your grandparents, I see? Where did you grow up? 0:03:23.440 --> 0:03:24.480 New York Queens? 0:03:24.800 --> 0:03:27.920 Oh you're l a see. So tell me a little 0:03:27.919 --> 0:03:30.600 bit about how how you got to red Hat. What 0:03:30.720 --> 0:03:31.320 was your path? 0:03:31.800 --> 0:03:34.320 When I was in high school? It was a chatty girl, 0:03:34.639 --> 0:03:37.160 teenage girl on the phone. We had one phone line. 0:03:37.440 --> 0:03:40.280 My older brother was studying at the local state college 0:03:40.280 --> 0:03:42.560 computer science, and he had to tell that end to 0:03:42.880 --> 0:03:45.760 compile his homework one phone line, and I'm on it 0:03:45.800 --> 0:03:48.760 all the time. He got very frustrated and he needed 0:03:48.760 --> 0:03:51.320 a compiler to do his homework. So he bought red 0:03:51.320 --> 0:03:55.760 Hat Linux from a CompUSA, brought it home and that 0:03:55.880 --> 0:03:57.840 was on the family computer. So I learned Linux and 0:03:57.840 --> 0:04:00.560 I started playing around with it. I really liked it 0:04:00.640 --> 0:04:04.520 because you could customize everything, like the entire user interface. 0:04:04.880 --> 0:04:07.040 You could actually modify the code of the programs you 0:04:07.080 --> 0:04:09.320 were using to do what you wanted. And for me, 0:04:09.400 --> 0:04:11.600 It was really cool because especially when you're a kid 0:04:11.600 --> 0:04:13.360 and like people tell you this is the way things 0:04:13.400 --> 0:04:15.120 are and you just have to deal with it. It's 0:04:15.240 --> 0:04:16.760 nice to be like I'm going to make things the 0:04:16.839 --> 0:04:20.120 way I want, modify the code and playing. Yeah, it 0:04:20.200 --> 0:04:22.040 was amazing and it was just such a time and 0:04:22.080 --> 0:04:25.080 like before it was cool, I was doing it and 0:04:25.880 --> 0:04:27.839 what I saw on that is sort of the potential 0:04:27.880 --> 0:04:30.920 like number one of like a community of people working together. 0:04:30.920 --> 0:04:34.000 And like the Internet existed, it was slow, it involved modems, 0:04:34.400 --> 0:04:36.320 but there were people that you could talk to who 0:04:36.360 --> 0:04:40.120 would give you tips and you'd share information, and this 0:04:40.240 --> 0:04:43.919 collaborative building something together is really something special. Right. I 0:04:43.920 --> 0:04:47.760 could file a complaint to whatever large software company made 0:04:47.760 --> 0:04:50.000 whatever software I was into, or I could go to 0:04:50.000 --> 0:04:52.120 an open source software community and be like, hey, guys, 0:04:52.160 --> 0:04:54.559 I think we should do this, and like, yeah, okay, 0:04:54.560 --> 0:04:56.719 I'll help, I'll pitch in. So you don't feel powerless, 0:04:56.720 --> 0:04:58.080 you feel like you can have an impact, and that 0:04:58.160 --> 0:05:02.040 was really exciting to me. However, open source software has 0:05:02.200 --> 0:05:05.440 a reputation for not having the best user interface, not 0:05:05.480 --> 0:05:10.000 the best user experience. So I ended up studying computer 0:05:10.040 --> 0:05:13.320 science and Electronic Media dual major, and then I did 0:05:13.360 --> 0:05:17.280 Human computeraction as my master's and my thought was, wouldn't 0:05:17.279 --> 0:05:20.960 it be nice if this free software accessible to anybody, 0:05:21.279 --> 0:05:23.320 if it was easier to use, some more people could 0:05:23.360 --> 0:05:26.000 use it and take advantage of it. And so, long 0:05:26.040 --> 0:05:29.320 story short, I ended up going to red Hat saying, Hey, 0:05:29.320 --> 0:05:30.720 I want to learn how you guys work. Let me 0:05:30.760 --> 0:05:33.480 embed in your team draft out of my graduate program. 0:05:33.600 --> 0:05:35.359 And I'm like, I want to do this for a living. 0:05:35.480 --> 0:05:38.080 This is cooler. So I thought this is the way 0:05:38.120 --> 0:05:39.719 to go, and I've been there ever since. They haven't 0:05:39.760 --> 0:05:40.599 been able to get rid of me. 0:05:42.640 --> 0:05:45.400 To backtrack this a little bit, you were talking about 0:05:45.400 --> 0:05:49.120 the sense of community that surrounds this way of thinking 0:05:49.160 --> 0:05:52.880 about software. Talk a little bit more about what that 0:05:52.920 --> 0:05:55.680 community is like, the benefits of that community, why it 0:05:55.760 --> 0:05:56.560 appeals to you. 0:05:57.040 --> 0:05:59.359 Sure, well, you know part of the reason I actually 0:05:59.440 --> 0:06:02.400 ended up going going to the graduate school track. Suddenly 0:06:02.440 --> 0:06:05.560 you're a peer of your professors and you're working side 0:06:05.600 --> 0:06:08.680 by side with them. At some point they retire and 0:06:08.720 --> 0:06:12.320 you're in the next generation. So it's sharing information, building 0:06:12.360 --> 0:06:14.920 on the work of others in sort of this cycle 0:06:15.040 --> 0:06:19.520 that extends past the human lifespan and in the same way, 0:06:19.640 --> 0:06:22.760 Like the open source model is very similar, but you're 0:06:22.800 --> 0:06:25.720 actually you're building something, and it's something in me. I'm 0:06:25.800 --> 0:06:28.640 just really attracted. Like I don't like talking about stuff. 0:06:28.680 --> 0:06:32.520 I like doing stuff with open source software. The software 0:06:32.520 --> 0:06:36.080 doesn't cost anything, the code is out there, generally uses 0:06:36.160 --> 0:06:39.480 open standards for the file formats. I can open up 0:06:39.480 --> 0:06:42.320 files that I created and open source tools as a 0:06:42.360 --> 0:06:46.040 high school student today because they were using open formats 0:06:46.040 --> 0:06:48.320 and that software still exists. I can still compile the 0:06:48.320 --> 0:06:51.679 code and it's an active community project. Like these things 0:06:51.720 --> 0:06:54.479 can outlast any single company in the same way that 0:06:54.560 --> 0:06:56.920 the academic community has been going on for so many 0:06:57.000 --> 0:06:59.680 years and hopefully we'll continue moving on. So it's sort 0:06:59.680 --> 0:07:02.599 of like not just the community around it, but just 0:07:02.680 --> 0:07:05.400 the knowledge sharing and also bringing up the next generation 0:07:05.480 --> 0:07:07.520 as well. Like all of that stuff really appealed to me. 0:07:07.560 --> 0:07:10.240 And also at the center of it, the fact that 0:07:10.720 --> 0:07:13.800 we could democratize it by following this open source process 0:07:14.160 --> 0:07:16.160 and feel like we have some control, We're not at 0:07:16.200 --> 0:07:19.240 the mercy of some faceless corporation making changes and we 0:07:19.280 --> 0:07:21.360 have no impact. Like that really appealed to me too. 0:07:21.440 --> 0:07:26.240 Yeah, for those of us who are not software phishonados, 0:07:27.040 --> 0:07:30.160 take a step backwards and give me a kind of 0:07:31.160 --> 0:07:34.640 description of terms. What's the opposite of open source proprietary? 0:07:34.720 --> 0:07:36.200 Proprietary is what we say, So. 0:07:37.040 --> 0:07:40.840 Specifically and practically, the difference would be what between something 0:07:40.840 --> 0:07:42.680 that was opens us in something that was proprietary. 0:07:42.760 --> 0:07:45.360 Sure, so there's a lot of difference. So with open 0:07:45.400 --> 0:07:49.360 source software you get these rights when you're given the software, 0:07:49.440 --> 0:07:51.840 you get the right to be able to share it. 0:07:51.920 --> 0:07:54.320 And depending on the lot, different licenses that are considered 0:07:54.360 --> 0:07:57.360 open source have different little things that you have to 0:07:57.360 --> 0:08:03.160 be aware of. With proprietary code, it's one copyright the company. 0:08:03.200 --> 0:08:05.560 Even a lot of times, when you sign your employment 0:08:05.560 --> 0:08:07.880 contract for a software company and you write code for them, 0:08:08.200 --> 0:08:09.840 you don't own it. You sign over your rights to 0:08:09.840 --> 0:08:11.760 the company. So if you leave the company, the code 0:08:11.800 --> 0:08:14.120 doesn't go with you. It stays in the ownership of 0:08:14.120 --> 0:08:16.760 that company. So then one like one company buys out 0:08:16.800 --> 0:08:18.920 another and kills a product, that code's gone. 0:08:19.040 --> 0:08:23.000 It's gone. For a business, why would a business want 0:08:23.080 --> 0:08:26.880 to be have open source code as opposed to proprietary. 0:08:26.240 --> 0:08:28.320 Well, for the same reasons, Like, say you're a business, 0:08:28.800 --> 0:08:32.680 You've invested all this money into this software platform right, 0:08:33.160 --> 0:08:36.280 and you've upskilled your employees on it and it's a 0:08:36.320 --> 0:08:39.120 core part of your business, and then a few years 0:08:39.200 --> 0:08:42.520 later that company goes out of business or something happens, 0:08:42.679 --> 0:08:46.040 or even something less drastic. You really need this future, 0:08:46.360 --> 0:08:49.480 but for the company that makes the software, it's not 0:08:49.480 --> 0:08:52.320 in their best interests. It's not worth the investment. They're 0:08:52.360 --> 0:08:54.280 not going to do it. How do you get that future? 0:08:54.600 --> 0:08:57.160 You either have to completely migrate to another solution, and 0:08:57.200 --> 0:08:59.120 this is something that's core at your business that's going 0:08:59.160 --> 0:09:02.040 to be a big deal to But if it's open source, 0:09:02.559 --> 0:09:05.640 you could either hire a team of experts. You could 0:09:05.720 --> 0:09:08.160 hire software engineers who are able to go do this 0:09:08.360 --> 0:09:12.200 for you. Go in the upstream software community, implement the 0:09:12.240 --> 0:09:15.080 feature that you want, and it'll be rolled into the 0:09:15.120 --> 0:09:17.520 next version of that company software. So even if that 0:09:17.600 --> 0:09:20.800 company didn't want to implement the future, if they did 0:09:20.800 --> 0:09:23.760 it open source, they would inherit that feature from the 0:09:23.880 --> 0:09:26.400 upstream community, is what we call it, So you have 0:09:26.480 --> 0:09:28.880 some control over the situation. If it's open source, you 0:09:28.920 --> 0:09:32.960 have an opportunity to actually affect change in the product, 0:09:33.200 --> 0:09:35.240 and you could then pick it up or pay somebody 0:09:35.240 --> 0:09:37.360 else to pick it up, or another company could form 0:09:37.400 --> 0:09:39.960 and pick it up and keep it going, So there's 0:09:40.000 --> 0:09:42.440 more possibilities. If it's open source, it's more like it's 0:09:42.480 --> 0:09:44.079 like an insurance policy almost. 0:09:44.120 --> 0:09:47.680 So innovation from the standpoint of the customer, innovation is 0:09:47.679 --> 0:09:50.679 a lot easier when you're working in an open source environment. 0:09:50.880 --> 0:09:51.520 Absolutely. 0:09:51.920 --> 0:09:55.680 Yeah. So now at RedHat, you're working with something called 0:09:55.679 --> 0:09:58.440 instruct lab. Tell us a little bit about what that is. 0:09:58.880 --> 0:10:01.040 So the thing that really excites to me about getting 0:10:01.040 --> 0:10:03.400 to work on this project is AI is sort of 0:10:03.440 --> 0:10:06.600 that has been this scary thing for me because it's 0:10:06.600 --> 0:10:09.760 one of those things like in order to be able 0:10:09.800 --> 0:10:15.040 to pre train a model, you have to have unobtainium GPUs, 0:10:15.760 --> 0:10:20.040 you have to have rich resources. It takes months, it 0:10:20.080 --> 0:10:25.040 takes expertise. There's a small handful of companies that can 0:10:25.080 --> 0:10:29.079 build a model from pre train to something usable, and 0:10:29.320 --> 0:10:31.640 it kind of feels like those early days when I 0:10:31.760 --> 0:10:34.480 was kind of delving in software and the same way. 0:10:34.520 --> 0:10:37.840 I think if more people could contribute to AI models, 0:10:38.640 --> 0:10:42.240 then it wouldn't be just influenced by whichever company had 0:10:42.280 --> 0:10:45.480 the resources to build it. And there's been a lot 0:10:45.520 --> 0:10:48.880 of emphasis on pre training models, so taking massive terabytes 0:10:49.000 --> 0:10:52.959 data sets, throwing them through masses of GPUs over months 0:10:52.960 --> 0:10:56.240 of time, spending hundreds of millions of dollars to build 0:10:56.320 --> 0:10:59.160 a base model. But when instruct lab does is say, okay, 0:10:59.360 --> 0:11:02.280 you have a base model. We're going to fine tune in. 0:11:02.320 --> 0:11:05.760 On the other end, it takes less compute resources. The 0:11:05.800 --> 0:11:08.040 way we've built in struck lab, you can play around 0:11:08.040 --> 0:11:10.600 with the technology and learn it on it off the 0:11:10.600 --> 0:11:14.040 shelf laptop that you can actually buy. So in this 0:11:14.160 --> 0:11:17.880 way we're enabling a much broader set of people to 0:11:18.000 --> 0:11:20.679 play with AI, to contribute it, to modify it. And 0:11:20.720 --> 0:11:23.680 I'll tell you one story from red Hat. Succi, who 0:11:23.760 --> 0:11:28.199 is our chief diversity officer, very interested in inclusive language 0:11:28.200 --> 0:11:31.200 and open source software, doesn't have any experience with AI. 0:11:31.880 --> 0:11:33.920 We have a community model that we have an upstream 0:11:33.920 --> 0:11:36.520 project around for people to contribute knowledge and skills to 0:11:36.559 --> 0:11:38.680 the model. She's like, I want to teach the model 0:11:39.000 --> 0:11:41.920 how to use inclusive language, like replace this word with 0:11:42.000 --> 0:11:43.600 this word or this word with this word. Oh my, 0:11:43.640 --> 0:11:46.520 oh that's so cool. So she paired up with Nicholas 0:11:46.559 --> 0:11:49.079 who is a technical guy at red Hat, and they 0:11:49.240 --> 0:11:52.560 built and submitted a skill to the model that you 0:11:52.600 --> 0:11:54.440 can just tell the model, can you please take this 0:11:54.559 --> 0:11:57.440 document and translate this language to more inclusive language, and 0:11:57.480 --> 0:11:59.920 it will do it. And they submitted it to the community. 0:12:00.080 --> 0:12:01.760 They were so proud. It was like, that's the kind 0:12:01.800 --> 0:12:04.200 of thing that, like, you know, maybe a company would 0:12:04.240 --> 0:12:06.680 be incentivized to do that, but if you have some 0:12:06.800 --> 0:12:10.400 tooling that's open source and something that anybody could access, 0:12:10.559 --> 0:12:13.120 then those communities could actually get together and build that 0:12:13.240 --> 0:12:14.480 knowledge into AI models. 0:12:14.720 --> 0:12:19.239 Just so understand, what you guys have is the structure 0:12:19.320 --> 0:12:23.600 for an AI system, and in other cases, individual companies 0:12:23.760 --> 0:12:27.720 own and train their own AI systems. It takes enormous 0:12:27.720 --> 0:12:30.720 amount of resources. They hoover up all kinds of information, 0:12:31.320 --> 0:12:34.480 train it according to their own hidden set of rules, 0:12:34.559 --> 0:12:39.000 and then a customer might use that for some price. 0:12:39.360 --> 0:12:41.280 What you're saying is, in the same way that we 0:12:41.360 --> 0:12:45.200 democratize the writing of software before, let's democratize the training 0:12:45.240 --> 0:12:48.880 of an AI system. So anyone can contribute here and 0:12:49.320 --> 0:12:53.000 teach the model the things that they're interested in teaching 0:12:53.000 --> 0:12:55.920 the model. I'm guessing correct me. On the one hand, 0:12:56.400 --> 0:12:58.520 this model, at least in the beginning, is going to 0:12:58.520 --> 0:13:01.680 have a lot fewer resources available to it. But on 0:13:01.720 --> 0:13:03.480 the other hand, it's going to have a much more 0:13:03.840 --> 0:13:05.760 diverse set of inputs. 0:13:06.280 --> 0:13:09.360 That's right. And the other thing is that IBM, basically 0:13:09.440 --> 0:13:12.120 is part of this project, has something called the Granite 0:13:12.160 --> 0:13:15.320 Model family, and they've donated some granite models. So these 0:13:15.360 --> 0:13:18.040 are the ones that take the months and terabytes of 0:13:18.120 --> 0:13:21.280 data and all the GPUs to train. So IBM has 0:13:21.320 --> 0:13:24.560 created one of those, and they have listed out and 0:13:24.640 --> 0:13:26.800 linked to the data sets that they used, and they 0:13:26.840 --> 0:13:29.760 talk about the relative proportions they used when pre training, 0:13:30.120 --> 0:13:32.160 so it's not just a black box. You know where 0:13:32.160 --> 0:13:35.000 the data came from, which is a pretty open position 0:13:35.040 --> 0:13:37.600 to take. That is what we recommend as the base. 0:13:37.679 --> 0:13:40.280 So you use the instruct lab tuning. You take this 0:13:40.360 --> 0:13:43.280 base granite model that IBM has provided, and you use 0:13:43.320 --> 0:13:45.760 the instruct lab tooling that red Hat works on, and 0:13:45.800 --> 0:13:48.120 you use that to fine tune the model to make 0:13:48.160 --> 0:13:49.800 it whatever you want. 0:13:50.320 --> 0:13:53.240 I want to go back to the partnership between IBM 0:13:53.320 --> 0:13:57.319 and red Hat here with them providing the granite model 0:13:57.920 --> 0:14:00.480 to your instruct lab. Is this the first ti I'm 0:14:00.960 --> 0:14:03.200 red Hat and IBM have collaborated like this. 0:14:04.000 --> 0:14:06.559 I think it's something that's been going on, Like another 0:14:06.840 --> 0:14:09.480 a product within the red Hat family would be open 0:14:09.520 --> 0:14:12.840 Shift AI, where they collaborate a lot with IBM Research team, 0:14:13.080 --> 0:14:15.679 like BLM is one of the components of that product 0:14:15.679 --> 0:14:19.680 that there's a nice kind of exchange and collaboration between 0:14:19.720 --> 0:14:21.000 the two companies. 0:14:21.760 --> 0:14:24.400 How large is the potential community of people who might 0:14:24.480 --> 0:14:27.560 contribute to instruct lab It could. 0:14:27.320 --> 0:14:30.640 Be thousands of people. I mean, we'll see, it's early days. 0:14:31.000 --> 0:14:34.320 This is early technology that was invented at IBM Research 0:14:34.440 --> 0:14:36.480 that they partnered with us at red Hat to kind 0:14:36.480 --> 0:14:39.480 of build the software around it. There's still more to go, 0:14:39.600 --> 0:14:41.840 Like right now, we have a team in the community 0:14:41.920 --> 0:14:44.240 that's actually trying to build a web interface to make 0:14:44.280 --> 0:14:47.280 it easier for anybody to contribute. So we have a 0:14:47.320 --> 0:14:50.440 lot of those sort of user experience for the contributor 0:14:50.480 --> 0:14:52.560 to the model stuff to work out that we're still 0:14:52.600 --> 0:14:55.800 actively building on. But like my vision for it even 0:14:55.960 --> 0:14:58.840 is I like going back to that academic model of 0:14:58.920 --> 0:15:01.440 learning from what others and building upon it over time. 0:15:02.000 --> 0:15:04.280 It would be very good for us to sort of 0:15:04.320 --> 0:15:08.640 go out and try to collaborate with academics of all fields, like, hey, 0:15:08.680 --> 0:15:11.320 you know, the model doesn't know about your field, would 0:15:11.400 --> 0:15:14.360 you like to put something into the model about your 0:15:14.400 --> 0:15:17.000 field so it knows about it, or even you know, 0:15:17.520 --> 0:15:20.400 talk to the model it got it wrong, let's correct it. 0:15:20.440 --> 0:15:22.680 Can we lean on your expertise to correct it and 0:15:22.760 --> 0:15:24.840 make sure it gets it right and sort of use 0:15:24.960 --> 0:15:28.040 that community model as a way for everybody to collaborate 0:15:28.080 --> 0:15:33.400 because before instruct Lab, my understanding is if you wanted 0:15:33.440 --> 0:15:35.920 to take a model that's open source license and play 0:15:35.960 --> 0:15:37.520 with it, you could do that. You could take a 0:15:37.560 --> 0:15:40.200 model kind of off the shelf from Hugging Face and 0:15:40.280 --> 0:15:42.360 fine tune it yourself. But it's a bit of a 0:15:42.400 --> 0:15:45.000 dead end because you made your contributions, but there's no 0:15:45.040 --> 0:15:48.400 way for other people to collaborate with you. So the 0:15:48.400 --> 0:15:50.920 way that we've built this is based on how the 0:15:50.960 --> 0:15:55.280 technology works. Everybody can contribute to it. This is something 0:15:55.320 --> 0:15:57.320 that you can keep growing and growing and growing over time. 0:15:57.680 --> 0:16:01.000 Yeah. Yeah, what's the level of expert te is necessary 0:16:01.040 --> 0:16:02.000 to be a contributor. 0:16:02.600 --> 0:16:04.480 You don't need to be a data scientist, and you 0:16:04.520 --> 0:16:07.480 don't need to have exotic hardware. Honestly, if you don't 0:16:07.480 --> 0:16:10.320 even have laptop hardware that meets SUSPEC for doing instruct 0:16:10.400 --> 0:16:13.560 Labs laptop version, you can submit it to the community 0:16:13.640 --> 0:16:16.160 and then we'll actually build it for you. We have 0:16:16.240 --> 0:16:18.600 bots and stuff that do that, and we're hoping over 0:16:18.640 --> 0:16:21.160 time to make that more accessible, first by having a 0:16:21.280 --> 0:16:23.880 user interface and then maybe later on having a web service. 0:16:24.200 --> 0:16:27.400 Yeah, so give me an example of how a business 0:16:27.480 --> 0:16:29.560 might make use of instruct lab. 0:16:30.120 --> 0:16:32.640 One of the things that businesses are doing with AI 0:16:32.760 --> 0:16:36.440 right now is using hosted API services. They're quite expensive, 0:16:36.720 --> 0:16:39.520 but they're finding value, but it's hard given the amount 0:16:39.560 --> 0:16:41.840 of money they're spending. And one of the things that's 0:16:41.840 --> 0:16:43.680 a little scary about it too, is like you have 0:16:44.000 --> 0:16:48.440 very sensitive internal documents and you have employees maybe not 0:16:48.600 --> 0:16:51.480 understanding what they're actually doing because you know, how would 0:16:51.520 --> 0:16:54.920 you if you're not technical enough when you're asking said 0:16:55.560 --> 0:17:00.800 public web service AI model information about out you're copy 0:17:00.800 --> 0:17:05.120 pasting internal company documents. It's going across the Internet into 0:17:05.160 --> 0:17:08.360 another company's hands, and that company probably shouldn't have access 0:17:08.400 --> 0:17:11.600 to that. So what both RedHat and IBM and the 0:17:11.640 --> 0:17:14.720 space are looking at, like the instruct lab model is 0:17:14.840 --> 0:17:18.320 very modest. It's seven billion parameter model, very small. It's 0:17:18.440 --> 0:17:21.560 very cheap to serve inference on a seven billion parameter model. 0:17:22.240 --> 0:17:25.639 It's competing with trillion parameter models that are hosted. You 0:17:25.760 --> 0:17:28.840 take this small model that is cheap to run inference on, 0:17:29.480 --> 0:17:33.400 you train it with your own company's proprietary data inside 0:17:33.400 --> 0:17:36.000 the walls of your company, on your own hardware. You 0:17:36.040 --> 0:17:39.320 can do all sorts of actual data analysis on your 0:17:39.320 --> 0:17:41.919 most sensitive data and have the confidence that has not 0:17:42.000 --> 0:17:42.960 left the premises. 0:17:43.760 --> 0:17:46.800 In that use case, you're not actually training the model 0:17:46.840 --> 0:17:50.080 for everyone. You're just taking it and doing some private 0:17:50.119 --> 0:17:52.800 stuff on it. Exactly doesn't leave the building. But that's 0:17:52.840 --> 0:17:59.040 separate from an interaction where you're doing something that contributes overall. 0:17:59.440 --> 0:18:02.280 Right, That's something maybe that I should be more clear 0:18:02.280 --> 0:18:04.600 about is there's sort of two tracks here, and this 0:18:04.680 --> 0:18:08.320 is very red hat classic. You have your upstream community 0:18:08.359 --> 0:18:10.960 track and you have your business product track. So the 0:18:11.040 --> 0:18:14.960 upstream community track is just enabling anybody to contribute to 0:18:15.000 --> 0:18:16.840 a model in a collaborative way and play with it. 0:18:17.280 --> 0:18:21.119 The downstream product business oriented track is now take that 0:18:21.240 --> 0:18:25.840 tech that we've honed and developed in the open community 0:18:26.520 --> 0:18:29.000 and apply it to your business knowledge and skills. 0:18:30.040 --> 0:18:33.880 This community driven approach marks a pivotal shift towards more 0:18:33.880 --> 0:18:39.679 accessible AI solutions. The contrast between externally hosted AI services, 0:18:39.960 --> 0:18:43.159 and the open model enhanced by instruct lab underscores the 0:18:43.160 --> 0:18:48.240 potential for broader adoption of AI in diverse business contexts. 0:18:48.760 --> 0:18:52.160 She envisions a future in which technological innovation is more 0:18:52.200 --> 0:18:56.520 tailored to individual business needs, guided by principles of openness 0:18:56.640 --> 0:19:02.640 and security. Seer imaginary case study. Sure, I'm a law firm, 0:19:03.200 --> 0:19:06.960 I'm an entertainment law I have one hundred clients who 0:19:06.960 --> 0:19:11.800 are big stars. They all have incredibly complicated contracts. I 0:19:11.880 --> 0:19:16.520 feed a thousand of my company's contracts from the last 0:19:16.920 --> 0:19:20.119 ten years into the model, and then every time I 0:19:20.160 --> 0:19:22.760 have a new contract, I ask the model, am I 0:19:22.800 --> 0:19:25.479 missing something? Can you go back and look through all 0:19:25.480 --> 0:19:28.119 our own contracts and show me a contract that is 0:19:28.200 --> 0:19:32.240 missing key components or exposes us to some liability. In 0:19:32.280 --> 0:19:37.240 that case, the model would know my law firm contracts really, 0:19:37.440 --> 0:19:40.199 really well. It's as if they've been working out my 0:19:40.280 --> 0:19:44.360 law firm. They're not distracted by other people's particular styles 0:19:45.400 --> 0:19:49.320 or a bunch of contracts from the utility industry, or 0:19:49.400 --> 0:19:53.880 they know entertainment law contracts exactly. 0:19:54.000 --> 0:19:55.840 Yeah, And you can train it in your own image, 0:19:55.880 --> 0:19:59.600 your style of doing things. It's something that your company 0:19:59.680 --> 0:20:03.399 can that is uniquely helpful to you. No third party 0:20:03.400 --> 0:20:05.639 could do that because no third party understands how you 0:20:05.720 --> 0:20:09.439 do business and understands your history and your documents. So 0:20:09.440 --> 0:20:12.080 it's sort of a way of getting value out of 0:20:12.119 --> 0:20:14.679 the stuff you already have sitting in a file cabinet somewhere. 0:20:14.800 --> 0:20:16.040 It's it's very cool. 0:20:16.320 --> 0:20:19.320 Yeah, give me a sort of a real world case 0:20:19.320 --> 0:20:22.560 study where you think the business use case would be 0:20:22.560 --> 0:20:27.159 really powerful. What's a business that really could see an 0:20:27.200 --> 0:20:31.200 advantage to using instruct lab in its way. 0:20:31.680 --> 0:20:33.959 The demo that I've given a couple of times at 0:20:33.960 --> 0:20:37.520 different events used an imaginary insurance company. So you say, 0:20:37.560 --> 0:20:41.480 you have this company, you have to recommend repairs for 0:20:41.560 --> 0:20:45.000 various types of claims. You've been doing this for years, 0:20:45.040 --> 0:20:47.920 you know. If you know the windshield's broken and you've 0:20:47.920 --> 0:20:50.760 gotten this type of accident and it's this model car, 0:20:50.960 --> 0:20:52.720 these are the kinds of things you want to look at. 0:20:53.400 --> 0:20:56.199 So you could talk to any insurance agent in the 0:20:56.200 --> 0:20:58.920 field and be like, oh, you know, it's a Tesla. 0:20:59.000 --> 0:21:01.480 You might want to look at the or something. They'll 0:21:01.520 --> 0:21:04.840 have some latent knowledge just so you can take that 0:21:04.960 --> 0:21:07.560 and train it into a model. Honestly, I think these 0:21:07.640 --> 0:21:10.680 kind of new technologies are better when they're less visible. 0:21:11.280 --> 0:21:13.719 So say you have the claims agents in the field 0:21:13.760 --> 0:21:15.760 and they have this tool and they're kind of entering 0:21:15.800 --> 0:21:18.760 the claim data. They're on the scene at the car, 0:21:19.320 --> 0:21:22.000 and it might say, oh, look, I see this is 0:21:22.040 --> 0:21:24.560 a Ford fiesta. These are things you want to look 0:21:24.600 --> 0:21:27.800 at for this type of accident. As you're entering the data, 0:21:28.200 --> 0:21:30.080 it could be going through the knowledge you had loaded 0:21:30.119 --> 0:21:32.600 into the model and be making these suggestions based on 0:21:32.600 --> 0:21:35.560 your company's background, and hey, you know, let's not make 0:21:35.600 --> 0:21:38.120