Business Intelligence and BI Assistants at IBM TechXchange Conference

Keith Townsend: All right. We’re wrapping up our last interview of the day, IBM TechXchange. Bruno, we’re going to be-

Bruno Aziza: The most important. The last one, but the most important one.

Keith Townsend: The most important and the most challenging because we had a conversation earlier today and it already went over the allotted time.

Bruno Aziza: I tend to do that.

Keith Townsend: You do tend to do that, and I also tend to do that, so they’re going to keep us honest. You’re new to IBM. Talk to me about why you came to IBM and what you’ve learned so far.

Bruno Aziza: So I came here with the perspective of 25 years in the data and AI space. I’ve worked at very small start ups, mid-size companies, like Business Objects and very large companies, Microsoft Oracle, and most recently at Google. And when I look at the opportunity over the next 10 years, it’s very clear to me that customers listening to us they’re going to need more than just a technology provider, they’re going to need a business partner that understands business transformation. Because this AI of Gen AI truly is about embracing the business transformation opportunity. It’s about, of course, using technology, but understanding that the world is going to be hybrid. It’s going to need an open approach to it so we can understand and handle the multiplicity of data sources, data use cases, multiplicity of clouds, hybrid. And so when I thought about who is the company that these enterprises are going to want to work with the most and the one that’s most likely to help them get to the next level, IBM came to be the most natural choice in my mind.

Keith Townsend: So we talked about this at launch, this ability of IBM to have the technical prowess, the consulting arm, the just years of experience of digital transformation. And we came up with this idea that we’ll probably execute on at some point like the 10 Commandments of AI implementations and AI projects. But give me your top two now. What are the two things that businesses should do now in preparation for their AI journey?

Bruno Aziza: One thing I’ll also add is the history of innovation with IBM. This is the company that has invented Sequel. The company that in Silicon Valley Lab came up with the first deployment of DBMS. The observation that I have working with customers at scale is that there’s of course 10 things they’re going to look at, but the first two I can think of are performance to cost ratio. Of course you want high performance, but you also want to check that in with your cost. And so you want a partner that’s going to understand the complexity of the use cases, the complexity of the infrastructure, and deploy you towards what is going to be most cost-effective for the performance that you require.

The second one, that is the most basic one that’s very different from Enterprise Gen AI versus Consumer Gen AI is governance and security. Are you allowed to get access to that data? How should you attribute the content that you’re creating from the originated data? Was data that was using training, in fact data that was licensed? So there’s a lot of complexities you’re going to want to handle to make sure that your company is not going to be in the headlines because you’ve deployed that poorly. There’s a lot of excitement in Consumer Gen AI, and I’m very excited for that. I’m very thankful for that. The bit that CIOs in the enterprise have to think about is that very little of that actually transfers in the enterprise context.

Keith Townsend: So last question, IBM is known for the small launch models. Dispel the myth for me that organizations have to have, I don’t have a problem with a Llama 405b, but not every organization or use case needs that. Small models. What’s the thing with IBM and small models?

Bruno Aziza: So it’s ironic because I can say that size is not the only thing that matters, right? So you can have a lot of personality packed in this small amount. I think particularly in the Enterprise Gen AI space, you need to have a dual strategy for models. There’s going to be a lot of use cases where large language models make a lot of sense, but there’s also going to be a space where small models will make a lot of sense for domain specific. For the ability to return specific results that are very oriented towards this particular domain. So it’s not one versus the other, it’s most likely going to be both. And I think it’s very comfortable for the industry, and in general, our brain likes to pin one thing versus the other, but I think when it comes to models and when it comes to Gen AI, it’s an and strategy. You’re going to have to have a large language model strategy and a small model strategy.

Keith Townsend: So I really appreciate IBM taking the time out to try something new. We love these multiple micro videos in which we got a lot of information, not to slight you in these small packages. Comment below, do you enjoy this format? We’d love to hear your feedback. And visit IBM for more information around their models and their Gen AI strategy.