Claude Baudoin with cebe IT

SUMMARY KEYWORDS

omg, standards, ai, model, work, claude, ieee, years, called, organization, iso, developing, premature, including, member, images, rfi, language, platform, object management group

00:09

Karen, Hello and welcome. I'm Karen Quatromoni, Director of Public Relations for Object Management Group OMG. And welcome to our OMG podcast series at OMG. We're known for driving industry standards and building tech communities. Today, we're focusing on the Object Management Group, standards development organization, and we're happy to be here with Bill Hoffman, who is the OMG CEO and chairman who will lead today's podcast session.

00:40

Hi Claude, can you briefly introduce yourself and your role?

00:43

Yeah, sure. First. Thanks Bill for having me. My name is Claude Baudoin. I am the owner and principal at a small consulting company called cebe IT and knowledge management based in San Rafael, California. What we do is Enterprise Architecture, business process, modeling, managing the knowledge lifecycle and communities of practice for clients. My background is in software engineering, and IT management. I have experience in the semiconductor energy industries, and I've been associated with OMG for over 30 years now, and I've been hanging around OMG for years, and now I have, I am co-chairing three groups within OMG, the task force on business modeling and integration, also a working group called the cloud Working Group, which publishes guidelines for cloud users, and very importantly, for the topic you have in mind today, I'm also co chair of the AI platform Task Force for OMG, wow.

01:51

And we want to talk about AI today, so that's great, and congratulations, and thank you for being a member with us for over 30 years. Very much appreciate that. So, you know, AI's exploded over the last couple years. I mean, I wouldn't even say few, I'd say even the last couple years, it's just been crazy. And it's obviously pretty much in flux, whether it's the, you know, emergence of generative AI concerns about trustworthiness and bias. Does that mean it's too early to talk about AI standards?

02:20

Oh, that's an excellent question, which for decades we've been wondering, when standards are premature versus when they're timely versus when they're overdue. So I just would like to take a few seconds and tell you an anecdote. First in 1993 I was at an OMG meeting outside of Chicago, and there was a team of people I was part of, and we were writing a couple of booklets on comparative analysis of the many object oriented analysis and design methods that existed at the time. And one of the methodologists who had invented one such method excoriated us in an article in a well known publication in the Object Oriented world. And he wrote a piece called premature standardization considered harmful, which, of course, was a joke about go to considered harmful from the days of structured programming. And two years later, just two years later, in 1995 that same person was sitting in a room with his colleagues in San Jose, California, inventing the Unified Modeling Language UML, which was basically the standard for object oriented analysis and design. So at which point is a standard premature, versus at which point is it timely? That anecdote shows that sometimes it's a very short time interval to the point where people realize they need a standard. So I think the fundamental issue is this one, there's a point at which people who are applying or using a certain technology are wasting a lot of time manipulating data, handling issues of portability and interoperability and integration, because there is no standard and there is multiple proprietary formats and languages, etc. And instead of focusing on the innovation they want to do, they have to spend 80% of their time solving interoperability challenges. That's when we need standards. And there are some areas of AI that are reaching that point, which is why I think it's timely and not premature. Our job is to figure out where that border is and how to not stifle innovation through standards, but to allow people to focus more of their energy on that innovation by resolving for them through standards the portability and interoperability issues that

04:59

makes sense. Mean standards provide leverage. Leverage allows productivity increases that that all makes all makes good sense. So can you give me examples of areas where you think these standards would help AI, developers and also the users?

05:11

So one thing we're working on right now, we're just starting to work on it with the help of one of our members, a company in Austria called Zephyr solutions. We're looking at, can people port models of neural networks from one platform to another? So when you look at convolutional neural networks, CNN, today, when you train a network using some training data, generally speaking, you must execute then the model on the same platform with your real data in order to get the network to provide recommendations. So you cannot lift and shift the model of the network from one platform to another, because there is no standard representation of that, of that model. What do I mean by representation of a model? Well, a neural network has a certain architecture, has a certain topology. There's a number of layers, there's a number of nodes on each layer. Each node executes a certain activation function on its inputs to produce it outputs. These functions have parameters. Can we create a model of all this so that we can take a model and move it from one platform to another? So that's the most exciting thing we're looking into right now. There's another area that we've been looking at, but we haven't yet found, if you wish, a leader to tell us what model we could develop, it's image classifiers. So you know, when you want to train an AI to recognize images, think of a an autonomous car company that wants the camera to recognize cars versus pedestrians versus bikes. You train the model on 1000s of images. These images come with a description that says, this image contains a car, that image contains a pedestrian. There's no standard for this, or rather, there's a couple of different de facto standards that are not quite compatible with each other. So that's another area where we could develop a standard that would allow the developers of such models to ingest 1000s or even millions of images without having to spend months massaging the data so that it's it can be ingested by the model. And we have several other suggestions that have been made, for instance, metadata to describe the data sets that are used to train models in other areas than image classification. So actually, that goes with semantic tagging of information, which is a widespread need right now. There are other efforts in other OMG groups that have to do with standardizing metadata to handle the semantics aspect of this. So these are some of the things we want to

08:17

work on. It's going to be an amazing next few years. I can tell you that, how would you determine which AI standards, including some of the ones you just talked about, might be needed?

08:28

So that's an that's a good question, because OMG works very differently from some other standards organization. Instead of having a committee of OMG members, you know, or employees of OMG, if you wish, you know, smoked in the proverbial smoked field filled room deciding by themselves what standards are needed. We have a very open process at OMG, as you know. So one of the things we can do is issue a request for information and RFI to find out what organizations need, and that's going to drive our priorities. And we can issues. When we issue those RFIs, they go to the general public, not just to the members of OMG. So we can collect input from the entire world. In fact, in 2019 five years ago, as we speak, the National Institute for Standards and Technology, NIST in the US had issued such an RFI, and they had 98 responses, including one from OMG. The problem with that RFI is that it's five years old, and in the meantime, things happen, including the explosion of large language models, llms or generative AI. So the landscape has changed, and the other thing is NIST mission is standards for the US government. So the way some of the questions were worded did not necessarily address or incite responses from an international. Audience, and OMG, is very much a global organization, so we have a process to gather more input, and we're actually thinking of issuing a new RFI to answer your question, which is to determine which standards might be needed. Excellent.

10:16

So, you know, I know there's a lot of other organizations that are looking at developing AI standards. How would you think our OMG efforts differ from those?

10:26

So when we speak specifically about AI, there's a lot of stuff going on right now both at ISO, the International Organization for standards, which has a an entire subcommittee devoted to AI. It's called SC 42 and then the IEEE standards association is also doing some work on on AI. So when you look at ISO they have a large number of subgroups within SC 42 which is working on various standards. But when you look at ISO standards, a lot of them are basically guidelines. They tell people this is what you should do. They're not, they're not very precise standards, if you wish. And OMG, works much more on formal standards that include a UML model, maybe an owl ontology, maybe a specific language, data formats and a graphical notation for those models, etc. So OMG is work is much more technically precise and gives a lot more concrete direction to users and developers on how they should do things, as opposed to general advice on thou shalt do X. IEEE standards Association, we have a liaison with them too. I should have mentioned that we have liaisons with many parts of ISO. So we work together. We collaborate. We feed a lot of OMG specifications to ISO so that they become international standards. So we are working together. IEEE, I'm also co chair of a project group on AI, terminology and data formats. It's going pretty slowly. And again, it's going to be a glossary. It's not a technical model of an area of AI, which is what OMG does,

12:30

very good. Bill, you would think after 35 years, we figured out how to build standards, and I think we do a good job at it. So how can our listeners get involved?

12:41

Well, I mean Bill, thanks for asking the question. You would be probably even more qualified than I am to tell the listeners that OMG is a member driven organization. So it's it's not a couch potato opportunity. It's an opportunity to get skin in the game and get involved. So of course, people can subscribe to, you know, the LinkedIn page or the mailing list and read periodic updates. But really, people who understand the importance of standards need to get involved in creating those standards. And the best way to do that is to become an OMG member, and to start attending our quarterly meetings, and the meetings we have in between, in order to generate the RFPs and then the standards. So obviously the OMG website explains all that to you. You can find information on how become, to become a member. There is a cost associated with that, and there is obviously a cost to attend meetings, including travel. So you need to start by understanding the benefits to your organization to justify those costs and that membership. And to me, it's it boils down to this, do not wait for others to create a standard that might constrain what you do, then you suddenly discover that maybe your customers are asking you whether you comply with that standard, and you haven't seen it coming. If instead, you get a front seat and you're part of the performance, so to speak, and you are part of the team that solicits and creates and selects a standard, then you're shaping the solution, and then you are you have advanced knowledge of the way the standard is shaping up, and that'll give you an advantage in the market Once the standard is adopted because you will have started developing your tools or developing your practice as a user, as a function of the standard that you see being developed because you're participating in it. So gotta be an actor, not a passive spectator,

14:59

right? A. Leader, not a follower. Thank you, Claude. I appreciate your time.

15:04

You're welcome that you have been able to discuss this with you.

15:09

Thank you both. And so today you've been listening to Claude Baudoin speaking about AI standards. Has the time come, and Claude is representing the Object Management Group, standards development organization, thank you. Applause.