WEBVTT
00:00:04.480 -->
00:00:10.000All right, welcome folks to another amazing SNEA Experts on Data podcast.
00:00:10.240 -->
00:00:20.000My continuous pleasure to be a host of this fantastic community, and uh this is gonna be a great one, mostly because there's a plethora of Erics available on this call.
00:00:20.239 -->
00:00:22.239That isn't always a requirement.
00:00:22.320 -->
00:00:26.160I like to have a two Eric minimum, so it's even better that we got three plus.
00:00:26.399 -->
00:00:29.359And uh Cal, you'll be an honorary Eric today.
00:00:29.440 -->
00:00:43.280Uh so for folks that are brand new to what we do here, uh so SNEA is an amazing community that I love the fact that we're sharing so much information with our technology world, both in the community and beyond.
00:00:43.439 -->
00:00:45.039Uh so my name is Eric Wright.
00:00:45.119 -->
00:00:49.759I'm the co-founder of GTM Delta and also the host here of the SNEA Experts on Data Podcast.
00:00:49.920 -->
00:00:52.479And I'm welcomed by three other amazing humans.
00:00:52.719 -->
00:01:06.079We're gonna talk today about an amazing webinar that talked about vision AI, and this is a deep dive that I took a look at the content that is gonna be in here, and it's this is what we need.
00:01:06.239 -->
00:01:13.359You know, people always have questions and they want to be able to get to the good stuff and they want to go deep, but also understand the why.
00:01:13.519 -->
00:01:15.359This is such a beautiful merger of those two things.
00:01:15.439 -->
00:01:26.239So we're gonna have a bit of a chat about you know the subject matter, and of course, we've got links down below to point to you know getting in touch with the folks here as well as seeing the webinar itself.
00:01:26.480 -->
00:01:29.040So, with that, I'm gonna get onto the good stuff here and start.
00:01:29.120 -->
00:01:36.319Uh so Eric uh Smith, if you want to introduce yourself for folks that are new to you, and uh and then we'll make our way around.
00:01:36.799 -->
00:01:36.959Sure.
00:01:37.040 -->
00:01:37.599Thanks, Eric.
00:01:37.680 -->
00:01:44.560Uh Eric Smith, I am a distinguished engineer and I work for Dell uh as a member of their CTIO team.
00:01:44.799 -->
00:01:50.239Uh I am also the DSN uh Data Storage and Networking Communities Chair.
00:01:50.319 -->
00:01:56.879So that's a SNEA community that uh is sort of sponsoring uh the AI stack webinar series.
00:01:57.519 -->
00:01:58.159Thanks for having us.
00:01:59.120 -->
00:02:02.480And uh Eric Gamble, just want to keep the Eric flow going.
00:02:03.359 -->
00:02:04.480I'm Eric Gamble.
00:02:04.719 -->
00:02:06.159I work with IBM.
00:02:06.319 -->
00:02:14.159I am an AI technologist, and also I do technically it's data science, but it's an area of AI modeling and predictive modeling.
00:02:14.319 -->
00:02:15.680That's my specialty.
00:02:15.919 -->
00:02:25.280Um, I spent four years in addition to my day job as an adjunct professor with Wake Forest University doing graduate analytics course.
00:02:25.599 -->
00:02:26.879Fantastic.
00:02:27.919 -->
00:02:29.280And uh Cal.
00:02:30.719 -->
00:02:31.199Thanks.
00:02:31.280 -->
00:02:33.199Uh so my name's Cal Foshi.
00:02:33.439 -->
00:02:39.439Unfortunately, I'm not one of the three Eric's here, but I also, like Eric Gamble, work for IBM.
00:02:39.599 -->
00:02:43.680I uh my day job is the uh storage procurement engineering manager.
00:02:43.840 -->
00:02:55.840I manage our team of engineers that work with our suppliers, but I'm also one of IBM's subject matter experts in computer vision, and I do that both for internal use cases and with clients.
00:02:56.560 -->
00:02:57.120Amazing.
00:02:57.280 -->
00:02:57.680Excellent.
00:02:57.759 -->
00:02:58.080All right.
00:02:58.159 -->
00:03:00.560We got there's a reason why I call it experts on data.
00:03:00.639 -->
00:03:01.919We got experts in the room.
00:03:02.080 -->
00:03:10.000So uh Eric, I'd love oh, Eric, sorry, Eric Smith, I'd love to have you talk about the webinar series really quickly.
00:03:10.159 -->
00:03:18.240Like what what what is the value of the AI stack webinar series and what else have we already talked about and kind of what's coming up?
00:03:18.400 -->
00:03:22.479Because for me, I'm excited as heck, but uh I'm a bit of a captive audience here.
00:03:22.560 -->
00:03:24.159So let's tell the rest of the world about it.
00:03:24.479 -->
00:03:25.360Yeah, so thanks.
00:03:25.520 -->
00:03:26.240Thanks for the question.
00:03:26.319 -->
00:03:32.479Uh the AI stack webinar series is really about introducing people to the basic concepts that are involved with AI.
00:03:32.719 -->
00:03:43.840Uh, we started with an intro to AI ML concepts webinar that included a binary digit trainer and gave up gave people an overview of how um actually training a model works.
00:03:44.000 -->
00:03:50.879Uh we've we've subsequently followed that up with a webinar on inferencing, uh uh, which was uh a few weeks ago.
00:03:51.120 -->
00:04:00.000We've also done one uh just now on training a model, uh like the actual model used in production and one for vision uh vision at that.
00:04:00.080 -->
00:04:03.599And uh it was a phenomenal session, and I I think everybody should check it out.
00:04:03.840 -->
00:04:09.759Going on in the future, though, it's really becomes about how do you practically deploy uh AI ML solutions?
00:04:09.840 -->
00:04:11.680You know, what does it take to deploy a model?
00:04:11.840 -->
00:04:14.560What does it take to use it in inferencing?
00:04:14.639 -->
00:04:18.720Uh how do you how do you set up an environment such that you could run a model?
00:04:18.800 -->
00:04:22.560Um, you know, what are the what's the hardware requirements, all those things.
00:04:22.800 -->
00:04:31.040And uh, you know, there's going to be a practical uh webinar coming up that's related to uh you know creating a chatbot that's based on RAG.
00:04:31.279 -->
00:04:36.560So uh it's you know, we've got a bunch of interesting, interesting things coming up here uh over the next year.
00:04:37.279 -->
00:05:03.120And what I really adore about this, and I'm blessed to be a part of this community, is that it is such a very strong community focus around education and around engaging and connecting people because we are birds of a feather, and we are all trying to solve similar problems for our friends, clients, customers, etc., and really advance the world through you know shared standards and shared outcomes.
00:05:03.279 -->
00:05:10.319And and and what I love about this is even though we all come from different brands, when we're in SNEA, that's that's all that matters.
00:05:10.399 -->
00:05:14.079So very, very cool to see the the webinar and and many more like this.
00:05:14.240 -->
00:05:17.600So maybe I'll I'll have you start off, Eric Gamble.
00:05:17.839 -->
00:05:30.959Let's let's talk about the vision challenge and what is really sort of distinct about vision AI and where this is an important piece of this, isn't just like every other AI.
00:05:31.040 -->
00:05:36.480When people, unfortunately, when we say AI, the first thing they think is we won't even say it.
00:05:36.560 -->
00:05:38.240It's the iPhone of AI, right?
00:05:38.399 -->
00:05:41.759We think of generative AI and they go to immediately a brand.
00:05:42.240 -->
00:05:43.279But it's not that.
00:05:43.439 -->
00:05:48.720And especially beyond generative chat stuff, vision is fundamentally different.
00:05:48.879 -->
00:05:56.560So let's I'd love to get your sense on what are the challenges and and what are the kind of cool breakthroughs that we're sharing through the webinar.
00:05:57.199 -->
00:06:09.680So one of the unique things about vision, I like to characterize AI as not being a problem of ones and zeros, but it's a problem of getting, or I should say it's a challenge of trying to get ones and zeros to behave like people.
00:06:11.199 -->
00:06:18.399And in the human context of this, vision is something all people can relate to as far as that goes.
00:06:18.639 -->
00:06:26.480Um and as far as trying to understand how models work, how they how we would process algorithms, how we would do prediction.
00:06:26.639 -->
00:06:35.439Um we can do a lot in the education space just by talking about vision and and trying to put that into action.
00:06:36.000 -->
00:06:46.160So in in the in this in the in our specific case of computer vision, um one of the challenges we went through in the webinar is these algorithms operate at a low level.
00:06:46.720 -->
00:06:51.199Um, you know, we're talking in the ballpark of 600 by 600 pixels.
00:06:51.519 -->
00:07:00.079Um well when you go buy your high-resolution iPhone or Android, um what do you do with that?
00:07:01.040 -->
00:07:16.560And so Cal and I can talk about that more, but but the the reality is that just like a person, like when we're staring at a high-resolution image as a person, um these algorithms are going to downsample, up sample pixels just to fit their models.
00:07:16.720 -->
00:07:21.279But as people, we process it whatever we see, good or bad vision.
00:07:21.439 -->
00:07:24.879Um and so we're able to make all those decisions.
00:07:24.959 -->
00:07:35.920So the real challenge is how do we knowing that these algorithms work the way they do, knowing camera technology, knowing everything else, how do we really get it to work like a person and as good as a person?
00:07:36.240 -->
00:07:40.480And and one of the benefits of that is once you actually can do that and automate that.
00:07:40.800 -->
00:07:48.399Now we're an AI model um can operate it away in a way that a person can't.
00:07:48.560 -->
00:07:50.399Like I'm going to get tired.
00:07:50.560 -->
00:07:54.319If I did this for 16 hours, my eyes are gonna get really tired.
00:07:54.560 -->
00:07:58.959Um, the AI model doesn't get tired, it will perform consistently.
00:07:59.040 -->
00:08:03.279So so there's a benefit in a bridge once we figure out how to do this.
00:08:03.839 -->
00:08:05.920Well, that's a a very interesting thing.
00:08:06.000 -->
00:08:18.800And even when we talk about LLMs in general, it's tough for people to map like what it's trying to achieve and the way it behaves, and then we sort of complain, like, oh, it it behaves this particular way.
00:08:18.959 -->
00:08:34.720Like, that's we've told it to be it's it's representative of the training data, like it's there's a reason, and unfortunately, it's also the average of us, so it's not always the best, and it will take the average, but vision is such an important thing because we do this today.
00:08:34.799 -->
00:08:39.600So I've got a green screen, which no one can tell other than general lighting over my head.
00:08:39.919 -->
00:09:25.039And even when I do camera work, we use things like depth of field, and quite often a lot of podcasters and even cinematographers they lean on this idea of like we use depth of field to force focus into one area because we're trying to tell the person to this is the only thing that should matter to them in the image, and we like forcibly move people towards that as humans trying to create this, but meanwhile, the human eye takes in about a trillion objects in the scene at any moment in time, and like you said, Eric, like we can't even process that, but it's there, it's being consumed, and now so this is interesting, and I I'd love to have you sort of keep going on this idea, Cal.
00:09:25.840 -->
00:09:32.960Or either you or Cal, and like what is the what are the limitations and why is this neat and how computer vision takes this on?
00:09:33.200 -->
00:09:36.399Well, so you just hit on something here as far as all the processing.
00:09:36.480 -->
00:09:38.879So a lot of it is signal and noise.
00:09:39.600 -->
00:09:41.200And and we don't think of it that way.
00:09:41.279 -->
00:09:43.600We're just we just look and we see and we process things.
00:09:43.840 -->
00:09:52.799But but the reality is that in our heads, um we are targeting signal and and extracting noise all the time at a very high rate.
00:09:53.039 -->
00:09:54.720I'll let Cal to jump in here.
00:09:54.960 -->
00:09:55.840Yeah, yeah.
00:09:55.919 -->
00:10:05.440So, you know, as people, and this is an aspect of humanity that we often don't talk about, we have our our senses, our sight, smell, hearing, taste, feel.
00:10:05.679 -->
00:10:12.480Oftentimes those senses are more effective at filtering out information than about gathering it, right?
00:10:12.639 -->
00:10:15.440You can see and hear and smell a million different things.
00:10:15.679 -->
00:10:20.559We have developed as people to understand what we need to be listening out for, right?
00:10:20.639 -->
00:10:22.559We need to hear that tiger sneaking up on us.
00:10:22.639 -->
00:10:25.919We don't really care about the birds flying in the trees.
00:10:26.399 -->
00:10:31.919The computer vision has none of that, that evolution, that background, that context, right?
00:10:32.080 -->
00:10:38.799So as modelers, we have to figure out how to how to either give the model that context or to simulate it.
00:10:38.960 -->
00:10:44.320Um, one thing strategy we use often is let's say, let's think of a door.
00:10:44.639 -->
00:10:47.759You might be wanting to see, is the door locked?
00:10:47.919 -->
00:10:52.639And to see if it's locked, you want to look at the handle and see if it's turned in the locked position.
00:10:53.120 -->
00:11:01.200Us as people, we can immediately look and see a door and we focus in on the doorknob, but the the computer vision doesn't know how to do that.
00:11:01.279 -->
00:11:03.679So we may first teach it what a door is.
00:11:03.840 -->
00:11:04.960Hey, this is a door.
00:11:05.279 -->
00:11:08.720So take a picture of the room, find just find the door.
00:11:08.799 -->
00:11:10.000So it finds the door.
00:11:10.159 -->
00:11:11.440Then we have another model.
00:11:11.519 -->
00:11:25.519It's like, okay, now that we've found the door and we've eliminated windows, we've eliminated um open uh entrances and exits, we've eliminated posters, all things that are rectangular that the model could get confused by.
00:11:25.600 -->
00:11:26.639We've eliminated those.
00:11:26.799 -->
00:11:28.000We just have a door now.
00:11:28.159 -->
00:11:31.519Now we have a model that says, go find a doorknob.
00:11:31.600 -->
00:11:34.480Um and that model specifically goes and finds the doorknob.
00:11:34.559 -->
00:11:35.440That's all it's doing.
00:11:35.600 -->
00:11:38.399And then we might have a third model, and that model is what's looking.
00:11:38.639 -->
00:11:41.360Is it in the locked position or the unlocked position?
00:11:41.600 -->
00:11:47.679Now, to a user who's taken that picture, they probably don't even realize all of that's happened.
00:11:47.840 -->
00:11:52.559All they know is that at the end of the day, their inspection told them the door is either locked or unlocked.
00:11:52.799 -->
00:12:02.960Um, we've used a series of simple models, a model that just finds a door, a model that just finds a doorknob, and then a model that looks for the position of the switch.
00:12:03.120 -->
00:12:11.279And as Eric talked about with signal and noise, we've eliminated all of that other noise in the room, all the windows, the posters, the entryways.
00:12:11.440 -->
00:12:14.480We've eliminated things that might be on the door, right?
00:12:14.559 -->
00:12:19.440The panels, the windows, um, the deadbolt, if that's what you're looking at.
00:12:19.519 -->
00:12:24.720And we've gone, we've you know, narrow focused it, but the model wouldn't do that itself.
00:12:24.879 -->
00:12:33.440So we've had to, we had to train three models to narrow in what your e would naturally do just because of all that his that context we have.
00:12:33.600 -->
00:12:36.000We know what a door is, we know what a doorknob is, right?
00:12:36.080 -->
00:12:39.279We don't get confused by the window, but the model does.
00:12:39.519 -->
00:12:45.120And um, that's a key part of filtering out all that extra noise for the models.
00:12:45.679 -->
00:12:47.679That was I just wanted to jump in.
00:12:47.759 -->
00:13:02.960That was actually one of the things that I walked away with the webinar, really having a healthy appreciation for was, you know, they introduced the concept of multiple models and how each model, you know, as Cal was just walking us through how each model does things uh progressively and uh and the value of doing that.
00:13:03.039 -->
00:13:11.039And I I didn't have a good appreciation for that with vision models specifically, but I think that's uh also there are many, many other applications of that same concept.
00:13:11.200 -->
00:13:32.960So well, and that's funny, Eric, when you say like I was thinking even from your side, like we've seen so much stuff that we've done where like at the very like sort of standards and like methods level that we're seeing that's happening, but then when we get to the like very specifics of inside how the outcome happens, what are the technologies and processes to get there?
00:13:33.200 -->
00:13:41.120And you know, it's not even about taking in more, but in fact, in this case, it's about eliminating the noise quickly.
00:13:41.360 -->
00:13:49.039And we're seeing you know, LLMs that are then being augmented with SLMs and other ways where we can get very purpose-built things.
00:13:49.200 -->
00:13:50.799And like I said, humans are interesting.
00:13:50.960 -->
00:13:54.960Like I said, when we see things, it's it's obvious, and it's completely obvious.
00:13:55.200 -->
00:14:03.600We know exactly why things behave a certain way, but even when that's the case, we still we're still wondered by it.
00:14:04.159 -->
00:14:15.279And what we miss sometimes is the LLM and these models are are dealing with that wonder, you know, and they have to also then present it.
00:14:15.440 -->
00:14:20.879And we have to, as model builders, and how do we build these systems?
00:14:21.120 -->
00:14:29.600We have to extract the very human required elements from them because we can't just you know take it all in and ship it all back.
00:14:29.759 -->
00:14:38.799And we're we're a long way from you know, blueberry muffin or Pomeranian, which is where we uh kind of all started with that vision of like, you know, can these vision models actually work?
00:14:39.039 -->
00:14:45.840Like, I know every time I accidentally lose my security key and I have to do a captcha, I'm like, really?
00:14:46.159 -->
00:14:47.759We don't know what a fire hydrant is yet?
00:14:47.919 -->
00:14:57.120We're still doing this thing, but it's because we have to continue to tune, train, and like that stuff all has a purpose.
00:14:57.279 -->
00:15:04.080People often forget that those CAPTCHA models are not just for us, they're they're doing a neat thing behind the scenes.
00:15:05.840 -->
00:15:14.960So on you know, Eric, uh I'd love to hear, you know, what is it that we see that we do now?
00:15:15.200 -->
00:15:16.639What are some of the methods?
00:15:16.799 -->
00:15:21.919Things like understanding what what is the the bounding box, how do we manage for for labeling?
00:15:22.000 -->
00:15:31.200How like there's so many intricate activities that are happening in how we achieve the end result with Vision AI.
00:15:31.279 -->
00:15:35.600So maybe walk through a couple of the technical hoops that we're jumping through.
00:15:36.000 -->
00:15:52.000Yeah, so so um just as a background, one of the things we have to stay focused on with AI is that it really at its fundamental core, it's about patterns and and detecting changes in patterns.
00:15:52.639 -->
00:15:56.639And all the other stuff is built around that.
00:15:57.120 -->
00:16:02.000So on bounding boxes, um, they're simply rectangular.
00:16:02.240 -->
00:16:14.799Um and most of what we do in the world of I'll say object detection, which is which is the bulk of this, and even in classification, it's based on um rectangular boxes.
00:16:15.279 -->
00:16:22.799There is an area of detection known as segmentation where you can draw polygons with lots of sophisticated shape cutouts and stuff like that.
00:16:22.960 -->
00:16:33.360Um, what happens is that if if you actually do all that sophisticated polygon work and feed it into an object detection model, like I like I'll use YOLO, I use that a lot in the presentation.
00:16:33.600 -->
00:16:35.919Um, it's just gonna make it a box.
00:16:36.480 -->
00:16:39.759It's gonna take all your time drawing and just box it.
00:16:40.639 -->
00:16:45.440So you have to be aware of tools like that because some of the tools out there don't advertise that.
00:16:45.600 -->
00:16:50.639But but the the key is that think of this box as a signal and noise thing.
00:16:50.879 -->
00:17:05.839Um in a way, one of the strange metaphors I use is that if you are a baseball player and you're a professional batter, um you work with a batting box.
00:17:06.720 -->
00:17:13.519If you had the ability to force that pitch where you want it every time, and inside that box, you would do that.
00:17:14.079 -->
00:17:15.759Um that's our goal.
00:17:15.839 -->
00:17:17.359That's what we're trying to do.
00:17:17.759 -->
00:17:23.039We're trying to create these models that can really fight inside this box and do really well.
00:17:23.200 -->
00:17:31.839So so, in some ways, like if we sit down with a client and a client tells us, I want this model to perform with an error rate of I'll say one in 10,000.
00:17:32.799 -->
00:17:36.720A lot of people would stop and walk away from the table at that, especially with vision.
00:17:36.960 -->
00:17:42.799Um the reality is you you you have to assess, well, how could we do that?
00:17:43.119 -->
00:17:55.680And this is why we go back to the systems of small models is as far as um we can very effectively get these models to do a task really well and then pass the next task off to the next model.
00:17:56.079 -->
00:18:01.039Um this is how we this is how we target those error rates and work that in that range.
00:18:01.519 -->
00:18:09.599Um I've kind of stepped away from the the question on the bounty box, but um that's that's essentially what we're trying to think of in the bounty box.
00:18:09.759 -->
00:18:12.160It goes back to signal noise and patterns.
00:18:12.400 -->
00:18:14.319So we want that box to be signal.
00:18:14.400 -->
00:18:15.599We don't want a lot of noise in there.
00:18:15.680 -->
00:18:20.319We put a little bit of noise just to keep margin on it, um, but we want to box in the pattern.
00:18:20.400 -->
00:18:31.759Now, when we say a pattern, um we have to think in patterns, and sometimes those patterns are like if I have a I use this in the example there, the Coke bottle.
00:18:31.839 -->
00:18:35.359Well, well, maybe the only thing I care about in my decision is the bottle cap and the label.
00:18:35.440 -->
00:18:39.359Well, maybe my pattern just needs to tie on that and leave everything else behind.
00:18:39.519 -->
00:18:42.000Sorry, I'm not trying to advertise for Copricot.
00:18:42.799 -->
00:18:45.039Um just just an example there.
00:18:45.279 -->
00:18:53.599But the the the point being is that um you think in patterns and you want to think about signal and noise and just get rid of noise all the time.
00:18:53.759 -->
00:18:55.519All the time get rid of noise.
00:18:56.079 -->
00:19:00.880And you're trying to make it like as you said, like we want these things to behave in a human way.
00:19:01.039 -->
00:19:17.200And the part of it is we're learning how humans behave as part of this because we don't understand fully, like we're building neural networks, and we still have to figure out where the the end result is because we don't quite know up here exactly how it works.
00:19:17.279 -->
00:19:20.880We're we're very far along, thanks to a ton of really amazing research.
00:19:21.119 -->
00:19:24.319But isn't like with vision, even simple things.
00:19:24.480 -->
00:19:28.480I told people there's assumptive things that we do when we look at a scene.
00:19:29.359 -->
00:19:35.119And people it's a small thing that people don't realize, but you know, we have to get rid of those assumptions.
00:19:35.279 -->
00:19:38.319How do you train around that human problem?
00:19:40.559 -->
00:19:42.240You want to take that, Cal, or I'll take it?
00:19:42.480 -->
00:19:43.359It's up to you.
00:19:45.440 -->
00:19:57.759So training around a human problem is always uh one of the biggest things that we have to do, and a big part of what Eric and I do is also teach people how to use vision tools.
00:19:58.079 -->
00:20:07.200So he and I are are actually Experts in it, but our um our overall goal when we work with clients is that they end up being able to do it themselves without us.
00:20:07.359 -->
00:20:14.720Uh so to teach how to overcome the the human assumptions is a challenging one.
00:20:15.279 -->
00:20:23.759A big thing we run into is, and this is especially with with engineers that we work with, are um good and bad.
00:20:23.920 -->
00:20:28.319They want us, they want the model to find it's that it's good or it's bad.
00:20:28.480 -->
00:20:31.519And we have to remind them it's it's about patterns, right?
00:20:31.680 -->
00:20:32.799We can find a pattern.
00:20:32.960 -->
00:20:35.440We're gonna teach you how to find a pattern, a distinct pattern.
00:20:35.599 -->
00:20:42.880Now, this pattern may represent a defect in your manufacturing process, but it's in itself not good or bad.
00:20:42.960 -->
00:20:44.319It's just the pattern.
00:20:44.720 -->
00:21:06.559And once we can it's just a pattern, and once we get the the the engineers or the modelers uh to start thinking in patterns and get that kind of mind shift from a you know go, no, go, good, bad to a pattern A and pattern B, then you can really get some cool things going with your inspections because you're not limited to that binary thinking.
00:21:06.880 -->
00:21:11.039Because maybe you have three patterns for good and four patterns for bad.
00:21:11.279 -->
00:21:23.039And um, so you may end up with seven objects where you're really just trying to say this is a good piece of equipment, this is a bad piece of equipment, but you end up with seven different objects um that you have decisions to make on.
00:21:23.200 -->
00:21:25.279Um, and that's that's the key one.
00:21:25.359 -->
00:21:28.720Uh, you know, as Eric said, it's it's all about the patterns.
00:21:29.119 -->
00:21:42.000Yeah, and and what Cal mentioned there, this is another strength in how you approach the modeling, is that if I have three different patterns of bad, I don't I don't care if they get confused.
00:21:42.640 -->
00:21:44.720I'm trying to reject on bad.
00:21:45.039 -->
00:22:03.839So so so there's actually modeling techniques where we strive to drop the accuracy down because all the models are going to have error, but if we can bucket that error, like buckets the bucket a lot of the confusion in the bad among the bads, and bucket the confusion of the goods among the goods, that's the game.
00:22:03.920 -->
00:22:05.039That's what we're trying to do.
00:22:05.279 -->
00:22:10.880So some of this is gaming the algorithms and trying to figure out how to do that and and pocket the error.
00:22:11.359 -->
00:22:40.480That's an interesting one too, because you think the we're we've we were a long way from hot dog or not hot dog to like being able to say, is this a manufacturing deviation that could lead to a rocket exploding when it's pressed by heat, like and being able to use that vision, but part of not just the capability to do it, but to optimize the compute required and the time required to do so.
00:22:40.559 -->
00:22:57.440And I think this is one of the most amazing innovations we're seeing, and and certainly in in the SNEA organization and and far beyond, where people are coming together to we've already got the thing, but now how do we make the thing better, accurate, faster, cheaper, ethical, and safe?
00:22:57.680 -->
00:23:05.359Like there's a lot of other criteria now, because as you said, we could just if I take a vision model, do I need to see the whole scene?
00:23:05.519 -->
00:23:11.200Or should I can I narrow the focus on setting that bounding box to a particular area or whatever?
00:23:11.359 -->
00:23:24.319There's minor things or using multiple SLMs and some kind of agentic handoff between that could in the end mean that we use less compute, less power, let where we get higher accuracy.
00:23:24.480 -->
00:23:33.839There's many, many ways now that we are testing because the fundamentals I think are pretty sound, but now we can really experiment.
00:23:33.920 -->
00:23:37.680So what's the the fun you're seeing around some of that stuff, Eric?
00:23:38.160 -->
00:23:44.240Well, so so in the vision space, um, I mentioned the the yellow models and they keep evolving.
00:23:44.480 -->
00:23:49.839And and um I'll use an example like yellow version three.
00:23:49.920 -->
00:23:57.039I mentioned in the in the webinar that there's no licensing fees to use it in production, but if you use version eight or version 11, there are.
00:23:57.920 -->
00:24:05.599The gain from that is these as these algorithms evolve, some of them are faster, some of them are more accurate, so depending on the use case.
00:24:05.759 -->
00:24:17.440Um, and so that whole idea of total cost to do the inferencing or total total cost as far as time and and compute to do things, it does make sense to play with, I'll say the shiny new golf club.
00:24:17.680 -->
00:24:20.480Go, go, go test the new tool, see what they can do, right?
00:24:20.640 -->
00:24:25.200Take take those use cases you know, test them again, and see if you can improve that.
00:24:25.359 -->
00:24:28.960Um a lot of that can ripple out into benefit and savings.
00:24:30.880 -->
00:24:39.440And from that, maybe Cal, how does uh that evolution of a model, you know, play out?
00:24:39.599 -->
00:24:49.039And where do you how do you define the start and finish of uh you know some type of tuning evolution and it is now better than it was?
00:24:49.200 -->
00:24:58.559You know, what when do we see this goal achieved when it is a very there's a lot of places where you could go is tough to measure in my mind?
00:24:59.759 -->
00:25:03.680Yeah, so uh the evolution of a model is always a fun one to talk about.
00:25:03.920 -->
00:25:14.640So to evolve a model that's the primary job of a modeler, uh the one who creates the models, and you might be evolving that model in a few different ways.
00:25:14.720 -->
00:25:31.279Um, a common one we have is we we have it running, uh testing it, and we see that it may work really well on pictures captured during the day on the manufacturing line, but at night it still gets it right, but the confidence score is lower, right?
00:25:31.359 -->
00:25:37.200It was 97, 98, 99 during the day, and it might be 88, 89, 90 during the evening.
00:25:37.519 -->
00:25:38.319Still correct.
00:25:38.480 -->
00:25:41.279Everything's going, and if you weren't looking at it, you wouldn't know.
00:25:41.440 -->
00:25:44.640But that lower score suggests there's a weakness in the model.
00:25:44.799 -->
00:25:54.960So the next evolution would be you take those pictures that were captured at night, you correctly label them with the bounding boxes like Eric talked about, and you retrain your model with those new images.
00:25:55.119 -->
00:26:00.559Now your model is better trained on what it was previously weak at, right?
00:26:00.640 -->
00:26:02.160Those nighttime images.
00:26:02.319 -->
00:26:09.759Um you test that model if it's good, then you deploy it, and that's now your new main model for your inspection.
00:26:10.000 -->
00:26:20.799Now, um I like to caution people here: a lot of the applications that create models will have um a feature that may allow it to tune itself.
00:26:21.440 -->
00:26:24.880Uh, this is something that we tend that Eric and I will warn against.
00:26:24.960 -->
00:26:26.960And I like to quote uh Dr.
00:26:27.039 -->
00:26:33.759Malcolm from Jurassic Park you know, you were so preoccupied on whether or not you could, you didn't stop to think if you should.
00:26:34.079 -->
00:26:39.599Now you can let these models update themselves and evolve themselves.
00:26:39.920 -->
00:26:46.640Our warning on that is um your original model may have had an innate error in it.
00:26:46.799 -->
00:26:51.039And that error is gonna propagate every time it iterates itself.
00:26:51.200 -->
00:26:57.440And if you're not paying attention to it, if you're not observing it or ever seeing it, that error is gonna get worse and worse and worse.
00:26:57.599 -->
00:27:03.440I'm sure we've all seen it, you know, playing with some of the different AI models online, right?
00:27:03.519 -->
00:27:14.240You you feed it a picture of a duck and say, redraw this, and then you keep feeding it the redraw, redraw, and before long you don't have a duck at all.
00:27:14.400 -->
00:27:19.440Um that's because that little error that was in its first one gets bigger and bigger and bigger.
00:27:19.680 -->
00:27:33.200The second part of a self-evolving model is one of the key jobs of a modeler is understanding what the model does and how it works, where it messes up and where it works well.
00:27:33.440 -->
00:27:46.559If the model's changing itself, you're gonna lose that knowledge on how this model functions, and you may not be able to deploy it properly to do what you want in your inspection, right?
00:27:46.640 -->
00:27:56.559We have our model that finds the objects, but then we have the inspection that takes the findings, the objects that you found and makes real life determinations, right?
00:27:56.720 -->
00:28:04.799Sometimes it's shut the line down, and um the quickest way to get your inspection turned off is to shut a line down when it didn't need to get shut down.
00:28:05.039 -->
00:28:16.000So uh we advise, at least in our area of expertise in manufacturing, um don't usually let your models change themselves.
00:28:16.160 -->
00:28:20.559It's it's much better to have oversight from somebody who can see what's happening.
00:28:20.720 -->
00:28:23.200Um, and then you asked about deploying a model.
00:28:23.359 -->
00:28:25.440I'm gonna keep it fairly simple on that.
00:28:25.599 -->
00:28:27.119It's gonna depend on your use case.
00:28:27.279 -->
00:28:31.440Sometimes you have use cases that need 99.99% accuracy, right?
00:28:31.599 -->
00:28:34.079You're spitting out a thousand vehicles a day.
00:28:34.400 -->
00:28:37.519You are 99.9% accurate.
00:28:37.680 -->
00:28:42.880Well, that means you're losing one vehicle a day, and that vehicle might be worth$70,000.
00:28:43.599 -->
00:28:46.640Um, so 99.9 is not good enough.
00:28:46.799 -->
00:28:52.799You have other use cases where the the client has really no way to find these defects.
00:28:52.880 -->
00:28:57.519They might be dents, scratches, corrosion, it could be partial seeding of connectors.
00:28:57.599 -->
00:29:07.759And if you can find 80% of the most common ones, you're helping them out a ton because those are 80, those 80% of errors wouldn't have been found any other way.
00:29:08.079 -->
00:29:12.880Um, so for that, you're a little loosier and a little goosier on what you allow.
00:29:12.960 -->
00:29:18.880Um, you're gonna focus on your your middle 80%, 80% of the most common defects.
00:29:19.039 -->
00:29:26.079For that 99.99%, you're really stress testing it with those edge cases because those are what's gonna trip up your model.
00:29:26.640 -->
00:29:38.720Um, once you feel like you tested enough, you you let it run and you know you monitor it, as I said, you got to watch it, see where it's performing well, see where it's struggling.
00:29:40.960 -->
00:29:42.000I'm gonna add to that.
00:29:42.160 -->
00:29:48.720So there's a ton of iteration in this, and and Cal is kind of explaining how we do that.
00:29:48.880 -->
00:29:52.240Um there's also diminishing return.
00:29:52.559 -->
00:30:02.079And so as we're doing these iterations, um we want to try to do them quickly and and you know, take big chunks off the table.
00:30:02.160 -->
00:30:11.440But it once we get that in a range where the demand the return is diminishing, um we we push that more to maintenance type of effort.
00:30:11.680 -->
00:30:14.960Um, but but versioning is a key thing too.
00:30:15.279 -->
00:30:25.519Um so as we're doing these iterations or even changing models over time, uh, if I need to get back to where I was six months ago, that version needs to be there.
00:30:25.599 -->
00:30:27.839That that ability to go back to that version.
00:30:28.000 -->
00:30:35.759Um people who jump into this and just keep you know iterating and changing and and and don't do the versioning work, um, eventually you'll regret it.
00:30:35.920 -->
00:30:39.200Eventually you need to cling to that and go back to that to reset.
00:30:40.079 -->
00:30:45.839Yeah, it even in like code generation, this is one thing that people often have the problem with.
00:30:46.000 -->
00:30:53.200Eventually the the good code gets rolled off the back, and you you no longer have your sort of safe reference.
00:30:53.359 -->
00:31:05.920And with vision, it's so different, especially because the real-time aspect of what most vision stuff is doing, it's a very like gener generative AI in the tech sense.
00:31:06.079 -->
00:31:13.920Yeah, we want it to be fast, but it's crapping out words at a reasonable rate to keep up with human consumption is not hard.
00:31:14.079 -->
00:31:34.880But being able to consume, process, and then inform somebody based on vision that's happening in real time, especially in a manufacturing sense, like the chance to suddenly like pull the line, you know, without having a human have to see the thing, that has to be something that happens in real time.
00:31:35.119 -->
00:31:40.559And there's so much difference in the efforts to train and the challenges.
00:31:40.799 -->
00:31:47.519So when you have clients and folks that are like saying, Hey, don't I just take a foundation model and just like just keep going?
00:31:47.680 -->
00:31:55.839Like, what is the surprise that you hear when you kind of tell people this is what training will look like as we begin to implement?
00:31:59.599 -->
00:32:15.759Um so uh the the biggest thing, especially when we're training clients for the first time, that whole bridge of thinking and patterns, we we go back to that because uh again, they make a lot of assumptions in their head.
00:32:16.000 -->
00:32:22.319Um we do, you know, so it's interesting.
00:32:23.200 -->
00:32:26.880When we're training, I mean we make a lot of mistakes.
00:32:27.039 -->
00:32:29.200We make a ton of mistakes creating models.
00:32:29.359 -->
00:32:33.440Um the the key is is not to cling to bad assumptions.
00:32:33.680 -->
00:32:36.160Um, don't if you're wrong, you're wrong.
00:32:36.319 -->
00:32:36.960Pivot.
00:32:37.200 -->
00:32:38.160Pivot fast.
00:32:39.920 -->
00:32:47.200Um and and and don't it's not a fight to be right, it's a fight to get it to work and have good results.
00:32:48.799 -->
00:32:57.039It is important that we keep that objective, and then now connecting the technology and the methods just to go, Eric Smith.
00:32:57.119 -->
00:33:06.079I'd you've probably seen a lot of people even taking this content in and learning from Eric and Cal and folks in the community.
00:33:06.480 -->
00:33:18.640You probably are seeing people that are like, oh, like they may be hopefully surprised and educated and easing their journey, you know, who are maybe not working directly on foundation or they just want to consume the stuff.
00:33:18.720 -->
00:33:27.680But as other technologists in the community, what are some of the fun things you're hearing back when people are hearing, you know, these type of tips and and tricks?
00:33:29.599 -->
00:33:32.880I think that's a great question.
00:33:33.119 -->
00:33:45.920Um, let me start with sort of what I what I was surprised with, uh, what I've been surprised with with the you know, the feedback that we've gotten from the sessions in general, and then I'll go to this one specifically.
00:33:46.240 -->
00:33:56.079Um, I think people really appreciate understanding how the technology kind of works first, uh, getting it getting a really good hands-on sense for what they can do with it.
00:33:56.319 -->
00:33:58.640And this dump this session was perfect for that.
00:33:58.799 -->
00:34:11.920You we we spent a lot of time looking at you know rectangles and and talking about uh you know what some of the challenges with working with each with each sort of uh picture uh granularity, you know, scale up, scale down of images.
00:34:12.079 -->
00:34:24.320I mean, it was really interesting, but uh the big one of the biggest surprises for me came when we were talking about planning the session, and we had started off with this idea that we would have a training webinar.
00:34:24.480 -->
00:34:26.719Uh and you know, let's train a model.
00:34:26.800 -->
00:34:41.840And so I trained a simple uh binary digit classification thing in our very first uh session, and I said, well, let's just, you know, we'll have on an expert that's from from uh an expert with LLM and then Eric and Cal experts in vision, and we'll just do one session on training.
00:34:41.920 -->
00:34:50.400And and it was clear, excuse me, it was clear that that was not going to work almost immediately because of how wide the topic was.
00:34:50.559 -->
00:34:52.800Um and so we focused on vision.
00:34:52.960 -->
00:35:04.800And what the other thing that I took away from this session and and in general these sessions are that a lot of the concepts are transferable, not exactly, but a lot of the principles are are similar.
00:35:04.960 -->
00:35:07.519Um I like the idea of failing fast.
00:35:07.599 -->
00:35:17.760Uh, I like the idea of not getting hung up on a specific assumption, uh, being very flexible and willing to pivot, all those things are are very important, I think, and they're common.
00:35:18.000 -->
00:35:34.559Um uh uh yeah, so that I mean I think that's really, really just sort of giving people a basis to sort of build upon and understand the differences between between training a large language model and and and a vision model has been uh sort of was interesting just from that perspective.
00:35:35.039 -->
00:35:39.519And and maybe Cal, I'm gonna call on you, I'm gonna take you into a deep dark topic.
00:35:39.679 -->
00:35:51.920Because one thing we hate is when text shows up in images, holy heck, it's the most brutal thing to consume because there's a lot of risk in how it's text is treated not as text.
00:35:52.079 -->
00:36:01.840We see it like I, as a human, I see that's a word and I treat it like a word, but we then have to train vision models to be able to do that amongst graphical things.
00:36:01.920 -->
00:36:03.360And this kind of touches on OCR.
00:36:03.440 -->
00:36:12.320I'll say, as an example, like even if you do OCR like this or like this, it will fundamentally change the accuracy.
00:36:12.480 -->
00:36:29.519And so I've even had folks that they say putting in documents and making sure that they are in the directional format will speed and also increase the accuracy when consuming and and rendering text out of out of data with OCR.
00:36:29.599 -->
00:36:37.519So I love to how does the text in images and OCR kind of come into play when we're looking at at the vision challenge?
00:36:37.920 -->
00:36:42.239Yeah, it it is a deep can be a deep dark topic.
00:36:42.400 -->
00:36:52.480Um, Eric and I have uh banged our heads against many of walls over the years uh when it's come to OCR and text-based use cases.
00:36:52.800 -->
00:37:01.599Um one thing I'll throw out there to begin with is don't be hesitant to use existing OCR technology.
00:37:01.840 -->
00:37:08.159Um it's it's a form of vision, but it's it's a honed-in, focused form of vision, right?
00:37:08.320 -->
00:37:18.800It it can recognize digits and numbers, letters versus maybe a more generic uh vision algorithm can be trained to really see anything.
00:37:19.039 -->
00:37:28.079Now, that's not to say you can train almost any vision model to learn what letters and numbers and characters are.
00:37:28.239 -->
00:37:34.719Um but is it gonna be better than an existing OCR that you can that's open source that you can use?
00:37:34.960 -->
00:37:35.679Probably not.
00:37:35.840 -->
00:37:39.039It really depends on the characters, I suppose.
00:37:39.280 -->
00:37:42.639Um you talked about you know the rotation.
00:37:42.719 -->
00:37:44.079That's that's a big one.
00:37:44.239 -->
00:37:56.079So we've actually um had models specifically to to see is this rotated the right way, and if it's not, you have code to rotate it before it even gets fed to an OCR.
00:37:56.320 -->
00:38:00.800It goes back to what Eric talked about, like a baseball player, right?
00:38:00.960 -->
00:38:12.559If I could stand up with my bat and say I want a 70 mile an hour ball right down the middle, I'm probably gonna hit it almost every time because I know exactly how fast it's coming in, I know where it's gonna come in.
00:38:12.639 -->
00:38:23.119And that's what our goal as modelers always is, no matter what the use case is, is to get that consistent pitch, same speed, same place, same time every time.
00:38:23.280 -->
00:38:26.159Um, then the model doesn't have to work very hard.
00:38:26.480 -->
00:38:36.000Um when it comes to uh accuracy of OCR, um, there's a lot of tricks in there.
00:38:36.239 -->
00:38:44.239Eric and I had to become experts in regular expressions when dealing with OCR because again, we wanted to tee it up for the model.
00:38:44.320 -->
00:38:51.119We knew we were looking at a serial number and we knew it followed a pretty specific alphanumeric pattern.
00:38:51.280 -->
00:39:00.880Um, it could be any amount of letter or numbers, or any any letters or numbers, but it went a specific pattern, you know, number, number, letter, letter, whatever it is.
00:39:01.119 -->
00:39:08.960So we we use regular expressions or reg X is what you might see in in the industry to tee it up for the model.
00:39:09.119 -->
00:39:13.760Hey, you're gonna see three letters, and then you're gonna see four numbers.
00:39:13.920 -->
00:39:14.719And guess what?
00:39:14.880 -->
00:39:23.119The three letters aren't gonna none of the letter will never be O or A or whatever the the rules were there.
00:39:23.360 -->
00:39:25.679Um, and those are all just little things.
00:39:25.760 -->
00:39:35.360You have to become experts in a lot of random areas to really get these effective and functional models because it's all about reducing the noise, increasing the signal.
00:39:36.000 -->
00:39:45.519Yeah, just to add to that, when when Cal and I get called in to do modeling work and help teach people, they are focused on making these work right, not just good.
00:39:47.840 -->
00:39:50.800They want them working really, really well.
00:39:51.360 -->
00:39:55.599So some of the techniques he's talking about are what we do.
00:39:56.000 -->
00:40:19.440Yeah, that and I think this is the there's so many people who Their exposure to this stuff is basically one-shotting off of a foundation model that's in some service that they don't understand what the training is, you know, and even like one thing that became very good early on, of course, was being able to generate human figures uh in certain ways, but there were certain things that were challenged.
00:40:19.519 -->
00:40:24.400Uh, you know, I I uh literally could have hold up my hands.
00:40:24.960 -->
00:40:31.840So uh I I've I've I wanted to get there's like a somebody sold sells this thing, it's like a little fake finger you put underneath.
00:40:31.920 -->
00:40:38.800And so you just wear this in any place you are, and if someone says, Hey, I don't like what you said, they're like, Oh, that was an AI generated video.
00:40:39.199 -->
00:40:40.800Look at the extra fingers.
00:40:41.280 -->
00:40:42.559That wasn't me.
00:40:42.719 -->
00:40:47.760Uh so I gotta there's a great way to get out of uh out of crime cameras, I suppose.
00:40:47.920 -->
00:41:00.480But the the other challenge was the the reason why those things are very good at generating human figures, not necessarily fingers and toes, is because they were generated off of a lot of training data that had a lot of human figures in.
00:41:00.559 -->
00:41:09.039I won't go deep into what that what that was, but there's a lot of lot of training data that you didn't have control over that led to that first foundation model.
00:41:09.199 -->
00:41:11.280And then now what do you do with that?
00:41:11.360 -->
00:41:18.880And this is why anybody that's thinking of getting started, it's not just grab the some latest and greatest model.
00:41:18.960 -->
00:41:22.400You have to understand what what led to the current state.
00:41:22.719 -->
00:41:26.559How do I make this you know achieve my first goal?
00:41:26.639 -->
00:41:37.199And then how do I continue to evolve myself or this model or my process, which is there's a lot of stuff that we don't necessarily know about what got us to to step one.
00:41:37.599 -->
00:41:37.920Yeah.
00:41:38.159 -->
00:41:47.519So um, if you're in this line of work for long enough, one of the key skills that doesn't often get talked about is your ability to break things.
00:41:47.840 -->
00:42:00.960Um I don't necessarily mean, you know, grabbing a hammer and smashing something, but uh when you have a new technology, one of the first things we do is we ask, how how do we think we can make this not work properly?
00:42:01.199 -->
00:42:02.800Where do we think it's weak?
00:42:02.960 -->
00:42:04.719Where do we think it'll get confused?
00:42:05.199 -->
00:42:13.360Um because you can't properly use a tool until you understand its full limitations, whatever that tool is.
00:42:14.159 -->
00:42:19.119Um if you try to hammer in a screw, that's not gonna work very well.
00:42:19.519 -->
00:42:21.519Uh it's the same thing with these tools.
00:42:21.599 -->
00:42:29.119If you try to some of them are amazing, a lot of them are amazing, but they might not be good at some specific thing.
00:42:29.440 -->
00:42:29.679Right.
00:42:29.840 -->
00:42:37.360So you talked about the the generative AI and it couldn't, it was having trouble making a a five-fingered hand consistently, right?
00:42:37.519 -->
00:42:39.519But it did a lot of other things really amazing.
00:42:39.840 -->
00:42:47.840So if you're using if we're using that tool and we understand that limitation, we might be asking to generate images that don't have people's hands in them.
00:42:48.239 -->
00:42:53.280Um it's the same for any other of these AI tools, these vision tools that we use.
00:42:53.679 -->
00:42:57.920Um you don't, you don't, you know, it goes back to Einstein, you right?
00:42:58.000 -->
00:43:01.599If you judged a fish by its ability to climb a tree, it wouldn't seem very smart.
00:43:01.920 -->
00:43:07.679Where the the key is to to break the tools and to understand their limitations.
00:43:07.760 -->
00:43:22.639Because if you've if you've broken it, if you know how it won't work and you know what it's not good at, then you can really deploy it in a way that it's going to be effective and that you're gonna get the full full value out of it, whatever its ceiling is you want to get there.
00:43:24.079 -->
00:43:32.480I I laugh at what Cal said about the breaking because I I I coach high school track and field, and I'll have kids ask me, so what do you do for your job?
00:43:32.880 -->
00:43:37.440And so so I tell him, I said, Well, well, I break AI so that we can make it work.
00:43:38.800 -->
00:43:40.079That's awesome.
00:43:40.320 -->
00:43:47.119Well, I'm glad that I got fine folks like you both uh breaking AI so that we can make it work better.
00:43:47.360 -->
00:43:53.679And I I could I only wish that we could we could go I could go eight hours and keep on going with with all y'all.
00:43:53.760 -->
00:43:56.079So thank you very much for sharing what you did.
00:43:56.320 -->
00:44:01.440I don't want to take away because people you talked about a great great stuff in the webinar, so we'll have links.
00:44:01.519 -->
00:44:05.840So people go check out the webinar, stay in touch with the series.
00:44:06.079 -->
00:44:19.360We'll also make sure that we have links out to folks that are here on on the podcast because this is stuff that this is a really exciting topic, and I love that in the SNEA community, there's opportunities to keep participating and having these conversations.
00:44:19.519 -->
00:44:20.639So I look forward to that.
00:44:20.800 -->
00:44:27.760I would tell people the classic thing if you don't think that we do certain assumptive things with vision, it's a very simple test.
00:44:27.920 -->
00:44:32.559Everybody knows what an elevator is, we know how it works, and we trust that it generally works as expected.
00:44:32.800 -->
00:44:43.840But if you think that your trust in that output is so good, then what I want you to do is I want you to go up to an elevator, I want you to turn around backwards, I want you to press the button, and I want you to close your eyes.
00:44:44.000 -->
00:44:47.519When you hear the ding and hear the door open, I want you to jump in backwards.
00:44:48.159 -->
00:44:52.239You are never gonna feel a feeling like that moment in your life.
00:44:52.480 -->
00:44:56.079Because what's the first thing you do when you get an elevator?
00:44:56.239 -->
00:44:57.679You check to see the elevator's there.
00:44:57.840 -->
00:45:00.480You don't know you do it, but you do it.
00:45:01.199 -->
00:45:13.840And if you don't do it, you know why you do it because your heart will make motions that you've never experienced on any roller coaster as you jump backwards into what you know is an elevator.
00:45:14.159 -->
00:45:17.119I I don't suggest people do that just in case there's no elevator there.
00:45:17.199 -->
00:45:19.199Uh uh, you know, so we're on that.
00:45:19.280 -->
00:45:22.480Uh I I recommend a friend is there just in case.
00:45:22.559 -->
00:45:31.760Uh, but uh but that tells you this is a very challenging technology problem that has a very human element to it.
00:45:32.000 -->
00:45:40.320And we tech will get there with human help, and it will continue to bounce off back and forth between human and tech.
00:45:40.480 -->
00:45:44.559And this is the value of of kind of what what these folks are doing.
00:45:44.719 -->
00:45:52.559So, with that, I'll I'll ask you both or all to kind of sort of do a quick wrap and remind folks where we can find you uh and any quick uh exiting words as well.
00:45:52.719 -->
00:45:55.199Eric, uh I'll say uh sorry, Eric Gamble.
00:45:55.280 -->
00:45:56.960I'll I'll go around Robin here.
00:46:00.320 -->
00:46:15.760Um with any analytics, any any any of the modeling, whether it's AI, I mean whether it's vision or just basic AI, um comes down to patterns and changes and patterns and and and go back to that core concept every time.
00:46:17.360 -->
00:46:19.519Absolutely, absolutely awesome.
00:46:19.679 -->
00:46:24.079And Eric, if when folks do want to get connected with you, what's the best way to do so?
00:46:24.239 -->
00:46:33.440Beyond, of course, we'll have some some uh some links to your SNEA persona, of course, and and people can get to the SNEA events and and we'll we'll be able to meet you IRL, hopefully.
00:46:33.519 -->
00:46:35.440But what's the best way to catch you online?
00:46:35.920 -->
00:46:39.199Uh you can get catch me on LinkedIn for sure.
00:46:39.440 -->
00:46:42.239Um you'll find me.
00:46:42.480 -->
00:46:46.639I love how LinkedIn became like unironically my favorite social media network.
00:46:46.719 -->
00:46:51.199Because all every we all just like they all disbanded, everybody disappeared off the other one.
00:46:51.360 -->
00:46:54.000So LinkedIn is the last uh the last bastion.
00:46:54.320 -->
00:46:54.800Yeah, yeah.
00:46:55.199 -->
00:46:56.480There's multiple of me out there.
00:46:56.559 -->
00:46:59.599I I think you'll see my background right now has a track on it.
00:46:59.679 -->
00:47:00.719So nice.
00:47:00.960 -->
00:47:01.599All right.
00:47:01.760 -->
00:47:06.719And Cal, where uh what what what what would you leave with the folks uh before we go?
00:47:06.800 -->
00:47:09.840And and where do we find you to keep this conversation going?
00:47:11.199 -->
00:47:17.280So I'm gonna steal a quote that Eric used uh in in our presentation, right?
00:47:17.440 -->
00:47:20.239All models are wrong, but some are useful.
00:47:20.480 -->
00:47:29.920And that's if you can remember that going into uh any any models you use, but in my case, vision models, you'll you'll be successful.
00:47:30.480 -->
00:47:36.960The models are never gonna be perfect, but if you understand how they can be used, you can do a lot of cool things with them.
00:47:37.199 -->
00:47:43.199And uh, if you need to or have any interest in getting in contact with me, LinkedIn is also the way to do it.
00:47:43.360 -->
00:47:49.280Uh, there are not too many cow foshies on there, so um it shouldn't be too hard to find me.
00:47:49.599 -->
00:47:50.400Nice, nice.
00:47:50.559 -->
00:47:55.840Well, it's like they like the old saying, uh, you know, democracy is the worst kind of government except for all the others.
00:47:56.719 -->
00:47:57.360Precisely.
00:47:57.599 -->
00:48:03.360And it that's we have to really, it's it's fun to contextualize this because that is exactly it.
00:48:03.440 -->
00:48:06.000Like we just make this like everything just works.
00:48:06.320 -->
00:48:10.159Like, well, assume it doesn't, and then work upwards from that assumption.
00:48:10.320 -->
00:48:24.000So uh and Eric Smith, of course, uh last and certainly not least, uh, where do we find you and what do you what do you want to share with the folks about you know this session and and what we've got ahead in 2026 for the SNEA community?
00:48:24.639 -->
00:48:24.960Sure.
00:48:25.119 -->
00:48:30.559Um I'm not sure if you can hear me uh the heads such as that thing.
00:48:30.800 -->
00:48:35.440Um I would say uh you know go out and break things.
00:48:36.079 -->
00:48:46.960You know, seriously, go out and break things, uh, try try doing things, um, you know, fail fast and see what works and what doesn't work, and uh and um you're gonna learn a lot.
00:48:47.039 -->
00:48:51.199And that's that's I think the best way to get introduced to AI and the topics in general.
00:48:51.280 -->
00:48:53.840Uh, we have a great set of what you know webinars coming up.
00:48:54.079 -->
00:49:00.079You know, consider going to them, looking at the demos, and uh and and the best way to get a hold of me is also on LinkedIn.
00:49:00.480 -->
00:49:01.440So thanks everyone.
00:49:02.159 -->
00:49:03.519LinkedIn's where the party's at.
00:49:03.599 -->
00:49:06.800And if you want to connect with me, there's way too many air crates in the world.
00:49:06.880 -->
00:49:10.159Uh, but there's only one disco posse, that's the easiest way to find me.
00:49:10.320 -->
00:49:15.199Uh so if you search for uh disco posse, there's definitely only one of those on LinkedIn.
00:49:15.360 -->
00:49:19.920Uh, and please, I recommend folks connect, watch other great podcasts.
00:49:20.079 -->
00:49:25.920So thank you all for joining and sharing your knowledge and and stuff today, and check out the webinar.
00:49:26.000 -->
00:49:26.800It's fantastic.
00:49:26.960 -->
00:49:28.639These are practical things.
00:49:28.800 -->
00:49:35.039So this isn't just like, let's talk for 45 minutes on the ethics of AI and have no outcomes at the end of it.
00:49:35.280 -->
00:49:42.079This is like you walk in, we walk through hard problems, you show us how to solve those hard problems, and you give us ways to experiment.
00:49:42.320 -->
00:49:46.159So this is practical for the best of us and for the rest of us.
00:49:46.320 -->
00:49:48.480And uh uh thank you all for sharing this stuff.
00:49:48.559 -->
00:49:54.880So, with that, make sure you, of course, go get to your latest SNEA online and in-person events.
00:49:54.960 -->
00:50:00.079So make sure you go to SNEA.org, check out all of the other links to upcoming in person events.
00:50:00.239 -->
00:50:12.6392026 is going to be a great year in technology in the world, and I look forward to seeing all of you on many, many more of these kind of podcasts and IRL somewhere at a SNEA event.
