Latest / EDGE AI POD / Garbage In, Garbage Out - High-Quality Datasets for Edge ML Research

Transcript

Garbage In, Garbage Out - High-Quality Datasets for Edge ML Research · EDGE AI POD

• 0:00
  Speaker 1: Welcome to the Deep Dive.
• 0:00
  Speaker 1: Today we're tackling a really interesting problem.
• 0:04
  Speaker 1: How do we get a clear picture of how well artificial intelligence is actually performing when it's running on those tiny power-sipping devices, you know, edge devices?
• 0:14
  Speaker 2: Right, the microcontrollers, the battery-powered gadgets, Exactly.
• 0:18
  Speaker 1: Things where AI meets the well the physical world.
• 0:20
  Speaker 1: Getting that right seems like kind of a big deal.
• 0:23
  Speaker 2: Oh, it is a big deal and as you'll see, the path to reliably evaluating and, frankly, pushing the boundaries of edge AI.
• 0:33
  Speaker 2: It isn't quite as straightforward as you might think.
• 0:36
  Speaker 2: There are some unique challenges we really need to unpack here.
• 0:40
  Speaker 1: Absolutely, and that's exactly our mission today right To dive deep into how a new initiative is confronting a core issue in this field.
• 0:47
  Speaker 2: Which is.
• 0:47
  Speaker 1: Which is the current lack of standardized, high quality data for edge AI research.
• 0:52
  Speaker 1: I mean, think about it If everyone's testing their AI on totally different data sets.
• 0:57
  Speaker 2: How can you compare?
• 0:58
  Speaker 1: Exactly.
• 0:58
  Speaker 1: It's nearly impossible to get a clear sense of which approaches are truly advancing the field.
• 1:08
  Speaker 2: Okay.
• 1:09
  Speaker 2: So to help us get a handle on this, we're looking at a really insightful paper.
• 1:10
  Speaker 2: It's called Leveling the Playing Field for Edge AI Research Through High-Quality Datasets.
• 1:14
  Speaker 1: Right.
• 1:15
  Speaker 2: And this comes from the Edge AI Foundation Datasets and Benchmarks Working Group.
• 1:19
  Speaker 2: It's a real collaboration of experts.
• 1:21
  Speaker 2: You've got folks from NXP, Inatera, Harvard.
• 1:26
  Speaker 1: University, imperial College Imagimab.
• 1:28
  Speaker 2: Renice's Technical University of Denmark.
• 1:29
  Speaker 2: It's quite the list.
• 1:30
  Speaker 1: It is and you might have seen the Edge AI Foundation's logo that sort of interconnected network design.
• 1:36
  Speaker 2: Yeah, I think so.
• 1:37
  Speaker 1: Their tagline is connecting AI to the real world, which really kind of sums up what we're talking about today, doesn't it?
• 1:43
  Speaker 2: It absolutely does, it nails it.
• 1:45
  Speaker 1: Okay, so when you think about the well, the incredible progress we've seen in AI recently, Massive progress.
• 1:51
  Speaker 1: Right.
• 1:52
  Speaker 1: A lot of that has been fueled by these enormous data sets like ImageNet for image recognition.
• 1:57
  Speaker 2: That'll really change the game.
• 1:58
  Speaker 1: Or just vast amounts of text data for natural language processing.
• 2:01
  Speaker 1: These big shared data sets they've provided like a common ground for researchers train, test, compare models.
• 2:09
  Speaker 2: And the impact is undeniable.
• 2:11
  Speaker 2: The paper we're looking at even has this graph, you know, showing how AI performance and things like image recognition, reading comprehension it hasn't just improved dramatically.
• 2:23
  Speaker 1: It surpassed humans in some cases.
• 2:24
  Speaker 2: In some specific tasks.
• 2:26
  Speaker 2: Yeah, yeah.
• 2:26
  Speaker 2: Which just goes to show the power of having those large standardized data sets.
• 2:31
  Speaker 2: They really drive things forward.
• 2:33
  Speaker 1: And this is the key thing, the edge AI world.
• 2:35
  Speaker 1: It presents a very different picture, totally different, because these tiny devices are used for such a huge range of applications Monitoring crops, wearable health trackers, smart sensors.
• 2:46
  Speaker 2: You name it.
• 2:47
  Speaker 1: The landscape is just so much more fragmented.
• 2:49
  Speaker 1: We don't have that same unified data infrastructure that's been so crucial for cloud AI.
• 2:55
  Speaker 2: It's a good analogy you used before comparing a scooter to a semi truck.
• 2:58
  Speaker 2: Different purposes, wildly different requirements.
• 3:02
  Speaker 1: Yeah.
• 3:03
  Speaker 2: And this inherent diversity.
• 3:04
  Speaker 2: It's made even trickier by the fundamental nature of edge devices themselves.
• 3:09
  Speaker 2: Constraints, you mean Exactly they have to operate under incredibly tight constraints, Very limited power, often running on tiny batteries, right Strict thermal limits they can't overheat in these tiny packages.
• 3:19
  Speaker 2: And tiny memory footprints, both for storing the actual AI model and for the.
• 3:25
  Speaker 2: You know the working RAM they use during operation right.
• 3:28
  Speaker 1: So, while the trend in cloud AI has been towards bigger, more complex models, often meaning better accuracy- yeah, bigger is often better in the cloud that sheer size becomes a massive roadblock when you're trying to cram AI onto a microcontroller with like kilobytes of RAM the resources just aren't there.
• 3:46
  Speaker 2: You can't run those huge networks.
• 3:47
  Speaker 1: Okay, so big data fueled cloud progress, but the edge is diverse resource constrained.
• 3:53
  Speaker 1: This leads us to the core issue what specific problem is this edge AI foundation initiative really trying to solve here?
• 4:03
  Speaker 2: So the paper really highlights a key challenge because we lack these standardized ways to evaluate performance using realistic data.
• 4:10
  Speaker 2: Many research publications in the tiny ML the edge ML fields.
• 4:14
  Speaker 2: They tend to present an overly optimistic picture.
• 4:17
  Speaker 1: Optimistic how.
• 4:18
  Speaker 2: About how much models can actually be compressed and optimized without significantly losing performance.
• 4:23
  Speaker 1: Okay, hang on.
• 4:23
  Speaker 1: So you're saying the really impressive efficiency gains, these tiny model sizes we sometimes read about in papers.
• 4:30
  Speaker 2: Yeah, the ones designed for these tiny devices.
• 4:32
  Speaker 1: They might look great on paper, but they don't always translate to real world success when you try to put them into actual products.
• 4:40
  Speaker 2: That's precisely it.
• 4:41
  Speaker 2: The authors emphasize that these models are often evaluated on what they call toy examples.
• 4:47
  Speaker 2: Toy examples yeah, very small, simplified data sets.
• 4:50
  Speaker 2: They just don't capture the complexity, the variability, the sheer messiness of real-world data.
• 4:56
  Speaker 2: So for our listeners, especially if you're involved in developing or deploying edge AI solutions, it's really crucial to be aware of this Research claims based just on these limited data sets might not accurately reflect how those models will perform in your specific use case.
• 5:12
  Speaker 1: Right, Because that overestimation it's not just academic, is it?
• 5:16
  Speaker 1: It could lead to real problems.
• 5:18
  Speaker 2: Oh, absolutely Significant delays, wasted resources.
• 5:21
  Speaker 2: When companies try to deploy these models in actual products and then discover whoops, they don't actually meet the necessary accuracy or efficiency targets.
• 5:29
  Speaker 1: That's a really critical point.
• 5:31
  Speaker 1: It's like testing that fuel efficient engine prototype only in perfect lab conditions.
• 5:36
  Speaker 1: You need the highway test, the bad weather test.
• 5:38
  Speaker 2: Exactly, you need the messy reality.
• 5:41
  Speaker 1: So, if these toy examples are giving us a potentially skewed view, what's the Edge AI Foundation's working group proposing?
• 5:50
  Speaker 1: What's the actual solution?
• 5:51
  Speaker 2: Okay.
• 5:51
  Speaker 2: So their core solution is to establish a publicly accessible repository, a collection of realistically sized data sets, and these are specifically curated for tiny and edge ML applications.
• 6:05
  Speaker 2: The fundamental idea is to move the whole research community away from those overly simplistic examples.
• 6:12
  Speaker 1: Towards data that actually represents the challenges you face in real world deployments.
• 6:16
  Speaker 2: Precisely Data that looks like the real thing.
• 6:18
  Speaker 1: Okay, so this repository, that's the key to creating that level playing field they talk about in the paper's title.
• 6:22
  Speaker 1: That's the idea.
• 6:23
  Speaker 1: How does having playing field they talk about in the paper's title.
• 6:24
  Speaker 2: That's the idea.
• 6:25
  Speaker 1: How does having these common high-quality data sets help exactly?
• 6:29
  Speaker 2: Well, by providing this shared foundation of data, researchers, developers, everyone can compare their various techniques much more effectively.
• 6:37
  Speaker 2: Techniques late, well.
• 6:38
  Speaker 2: Whether they're focused on shrinking the model size, minimizing power consumption, boosting accuracy, improving precision, whatever their goal is, they can compare them in a much more meaningful, objective way, because everyone will be evaluating their innovations against the same benchmarks, using data that reflects real-world complexities.
• 6:57
  Speaker 1: Okay, that sounds like a well, a really crucial step Moving towards more reliable progress in the field.
• 7:03
  Speaker 1: What are some of the specific goals the working group has set out for this initiative?
• 7:08
  Speaker 2: They have several key objectives laid out.
• 7:10
  Speaker 2: Firstly, obviously, provide these realistically scaled data sets and, importantly, these won't be static.
• 7:17
  Speaker 2: The idea is they'll be actively curated and expanded by the broader community.
• 7:22
  Speaker 1: Dynamic.
• 7:23
  Speaker 2: Secondly, they want to foster open research, research into the most effective ways to actually do the AI inference on these data sets, taking into account critical factors like power, memory, accuracy, precision all the edge constraints.
• 7:36
  Speaker 1: Makes sense.
• 7:37
  Speaker 2: And finally, they intend to facilitate the development and, crucially, the sharing of diverse optimization techniques within the community.
• 7:44
  Speaker 1: So it's not just about giving people the data.
• 7:46
  Speaker 1: It's about building a whole collaborative ecosystem around it to advance the tech.
• 7:51
  Speaker 2: That's exactly it, and it's important to understand the work group's role here.
• 7:55
  Speaker 2: They're quite clear about this.
• 7:56
  Speaker 2: So, they're not positioning themselves as, like, the judges of submission quality or the managers of some big centralized benchmarking system.
• 8:06
  Speaker 1: OK.
• 8:06
  Speaker 2: Their role is more about being an enabler.
• 8:08
  Speaker 2: Centralized benchmarking system.
• 8:10
  Speaker 2: Their role is more about being an enabler.
• 8:11
  Speaker 2: They're providing the essential resources, the data sets that will empower the community itself to conduct more informed, more transparent comparisons between different approaches.
• 8:19
  Speaker 1: Got it Empowering the community's collective knowledge, not imposing from the top down.
• 8:24
  Speaker 1: Now the paper talks about the importance of carefully considering use cases and selecting data sets.
• 8:29
  Speaker 1: Given how diverse edge AI is, how are they actually approaching that?
• 8:33
  Speaker 1: It seems tricky.
• 8:34
  Speaker 2: It is tricky that heterogeneity is a huge challenge, but they propose a pretty smart approach actually, which is Focusing on the fundamental technical requirements that different use cases impose.
• 8:44
  Speaker 1: Yeah.
• 8:44
  Speaker 2: Rather than trying to define specific benchmarks for every single possible application out there.
• 8:49
  Speaker 1: Okay, explain that a bit more.
• 8:55
  Speaker 2: So, for instance, instead of creating a specific benchmark just for, say, smart doorbells, they would analyze the underlying technical demands that a smart doorbell places on an AI system, things like real-time object detection.
• 9:05
  Speaker 1: Under different lighting maybe.
• 9:07
  Speaker 2: Exactly Varying lighting conditions.
• 9:09
  Speaker 2: With a limited power budget, they break it down to those core requirements.
• 9:13
  Speaker 1: Okay, so it's more granular, focusing on the technical build-up blocks, not the specific end product.
• 9:18
  Speaker 1: What kinds of technical requirements are they looking at then?
• 9:21
  Speaker 2: They've identified several key categories.
• 9:23
  Speaker 2: One really crucial distinction is between real-time processing and batched processing.
• 9:28
  Speaker 1: Okay.
• 9:29
  Speaker 2: For many edge applications, like, say, anomaly detection in industrial equipment, you need that immediate analysis.
• 9:35
  Speaker 2: So real-time single sample operation is a top priority.
• 9:39
  Speaker 1: Can't wait for a batch to process if a machine is about to fail.
• 9:42
  Speaker 2: Precisely.
• 9:42
  Speaker 2: Another vital factor is the energy budget.
• 9:45
  Speaker 2: A tiny sensor running on a coin cell battery.
• 9:48
  Speaker 2: It operates under vastly different power constraints compared to, say, a wall powered smart camera.
• 9:55
  Speaker 1: Absolutely.
• 9:55
  Speaker 1: You can't drain a watch battery in an hour with a power hungry AI model.
• 9:59
  Speaker 2: No way.
• 10:00
  Speaker 2: And that links to another key consideration, always on operation.
• 10:04
  Speaker 1: Right Things that are constantly listening or watching.
• 10:06
  Speaker 2: Yeah, health monitors, security sensors.
• 10:08
  Speaker 2: They need to continuously monitor their environment and that has huge implications for power consumption, sustained performance, even the long-term reliability, the durability of the hardware itself.
• 10:20
  Speaker 2: They also consider the basic nature of the task.
• 10:22
  Speaker 2: Is it classification, like identifying a specific sound or prediction, or is it more like regression or data transformation?
• 10:30
  Speaker 1: Different types of output Right.
• 10:32
  Speaker 2: And finally they look at the data modality.
• 10:34
  Speaker 2: Is the device processing sequential time series data like sensor readings from an accelerometer, or individual independent frames or images, like from a camera?
• 10:44
  Speaker 1: Okay.
• 10:44
  Speaker 1: So by dissecting all these diverse applications into these core technical requirements, they can develop data sets and benchmarks that have broader relevance.
• 10:52
  Speaker 1: They apply across a wider range of edge AI scenarios.
• 10:56
  Speaker 2: That's exactly the goal.
• 10:57
  Speaker 2: It allows researchers to focus on optimizing for specific technical challenges, challenges that are relevant to loads of potential applications, rather than getting stuck on very narrow application-specific data sets.
• 11:10
  Speaker 1: That makes a lot of sense.
• 11:11
  Speaker 1: Now, the paper also outlines how the Edge AI Foundation actually intends to improve the quality and usefulness of these data sets.
• 11:18
  Speaker 1: What's their strategy there?
• 11:20
  Speaker 2: Yeah, they've laid out a pretty comprehensive multifaceted strategy for that.
• 11:24
  Speaker 2: They plan to start by leveraging existing data sets as foundational base sets.
• 11:30
  Speaker 1: So building on what's already out there.
• 11:32
  Speaker 2: Exactly, not reinventing the wheel every time.
• 11:34
  Speaker 2: They also emphasize the critical importance of variety, actively seeking out and incorporating a wide spectrum of scenarios variabilities within the data.
• 11:43
  Speaker 1: Why is that so important?
• 11:45
  Speaker 2: To ensure that the AI models trained and tested on these data sets are robust, that they can actually handle the complexities of the real world.
• 11:52
  Speaker 2: You don't want a data set that only shows perfect, ideal conditions.
• 11:55
  Speaker 2: That's not reality.
• 11:57
  Speaker 1: No real world.
• 11:57
  Speaker 1: Data is messy noise, different environments, unexpected stuff.
• 12:01
  Speaker 1: The AI needs to handle all that.
• 12:04
  Speaker 2: Absolutely.
• 12:04
  Speaker 2: Another crucial element is metadata.
• 12:08
  Speaker 1: The data, about the data.
• 12:10
  Speaker 2: Right.
• 12:10
  Speaker 2: They're committed to ensuring all data sets come with complete and accurate labeling, Plus information about how the data was collected, how it was processed.
• 12:19
  Speaker 1: Why does that matter?
• 12:20
  Speaker 2: It builds trust.
• 12:21
  Speaker 2: It allows researchers to really understand the data thoroughly and then interpret their results with confidence.
• 12:27
  Speaker 2: And, finally, they recognize that AI is well constantly changing, so they're planning for continuous adaptation.
• 12:36
  Speaker 2: The data sets need to evolve to keep pace with the latest advancements in AI models and techniques.
• 12:42
  Speaker 1: So this isn't just a static library they're building.
• 12:44
  Speaker 1: It's meant to be a living, dynamic resource that grows and adapts as the field moves forward.
• 12:49
  Speaker 2: That's the vision.
• 12:50
  Speaker 1: The paper mentions an initial focus on visual wake words.
• 12:53
  Speaker 1: Can you elaborate a bit on why they're starting there?
• 12:56
  Speaker 2: Yeah, they're initially focusing on visual wake words using images.
• 12:59
  Speaker 2: That's a pretty common task, right, and many battery operated devices that need to respond to some visual cue.
• 13:04
  Speaker 1: But while using minimal power no-transcript.
• 13:10
  Speaker 2: Plus, there's a decent amount of existing image data out there that can serve as a foundation to build upon.
• 13:15
  Speaker 2: Okay, so they'll be working hard to assess how well models trained on these initial sets actually generalize to real-life scenarios and making sure the data sets cover a really broad range of conditions, variations, all that messy reality we talked about.
• 13:30
  Speaker 1: And they plan to expand beyond images later.
• 13:32
  Speaker 2: Yes, absolutely yeah, plans are in place to extend to images later.
• 13:34
  Speaker 2: Yes, absolutely yeah.
• 13:34
  Speaker 2: Plans are in place to extend to other data modalities soon after.
• 13:37
  Speaker 1: Okay, it sounds like a really well thought out and, frankly, incredibly important undertaking.
• 13:43
  Speaker 1: Now the paper keeps stressing the vital role of community participation.
• 13:47
  Speaker 1: How can people, researchers, companies, individuals actually get involved?
• 13:52
  Speaker 1: How can they help ensure these data sets are high quality and relevant?
• 13:55
  Speaker 2: Community involvement is absolutely paramount.
• 13:57
  Speaker 2: They make that very clear For this initiative to really succeed and have a lasting impact, they need the community.
• 14:03
  Speaker 1: So how do they get involved?
• 14:04
  Speaker 2: The Edge AI Foundation is actively encouraging participation in several key ways.
• 14:09
  Speaker 2: First, they're specifically asking for recommendations and contributions of existing data sets.
• 14:14
  Speaker 1: So if someone knows a good recommendations and contributions of existing data sets.
• 14:17
  Speaker 2: So if someone knows a good data set yeah, data sets that could be valuable starting points, or maybe data sets that community thinks would really benefit from being improved or expanded they want those suggestions.
• 14:26
  Speaker 2: Ok, they also really want feedback feedback on how well the current data sets meet the needs of various use cases, and they're actively soliciting suggestions for expanding the variety, the scope of the data included.
• 14:39
  Speaker 1: So if a company, for example, has a valuable data set they've been using internally, or if they see gaps in what's available, they have a channel to contribute that expertise, that data, absolutely.
• 14:48
  Speaker 1: What else?
• 14:49
  Speaker 2: They're also inviting the community to contribute improved or expanded test sets.
• 14:53
  Speaker 2: Help raise the bar for performance.
• 14:55
  Speaker 2: Drive more innovation.
• 14:57
  Speaker 1: Makes sense.
• 14:57
  Speaker 1: Better tests mean better models.
• 14:59
  Speaker 2: And perhaps most critically, they're seeking assistance with the huge task of data labeling and annotation.
• 15:05
  Speaker 1: Ah, the often unglamorous but essential work.
• 15:09
  Speaker 2: Totally essential To ensure the accuracy, the completeness of the data sets.
• 15:14
  Speaker 2: They're even looking for contributions of labels for data sets that might currently lack comprehensive annotations.
• 15:21
  Speaker 1: It really sounds like a truly collaborative effort.
• 15:24
  Speaker 1: Everyone in the Edge AI ecosystem seems to have a potential role to play.
• 15:28
  Speaker 2: That's the goal.
• 15:30
  Speaker 1: Now, towards the end of the paper, there's a very clear call to action.
• 15:33
  Speaker 1: It's directed right at the Edge AI Foundation community.
• 15:35
  Speaker 1: What are the main things they're urging people to do?
• 15:39
  Speaker 2: Yeah, the central message there is a strong, direct invitation to get actively engaged.
• 15:44
  Speaker 2: They are urging members and potential members to provide their valuable insights.
• 15:49
  Speaker 1: Insights on what.
• 15:50
  Speaker 2: On the specific data-related challenges they are actually facing in their own work right now and to clearly articulate the use cases they believe the working group should prioritize.
• 15:59
  Speaker 1: So real-world feedback.
• 16:01
  Speaker 2: Exactly that direct feedback loop is essential to make sure the Foundation's resources are focused on the most pressing real-world needs of the community.
• 16:09
  Speaker 1: And beyond feedback.
• 16:11
  Speaker 2: Well, they're explicitly requesting contributions, Contributions of realistic data sets and relevant evaluation methods for those use cases that are identified as most critical.
• 16:21
  Speaker 1: Okay.
• 16:22
  Speaker 2: And they're actively encouraging participation in the collaborative development of best practices for optimizing AI models for deployment on these data sets.
• 16:31
  Speaker 2: How do we best tune models for this data for deployment?
• 16:35
  Speaker 1: on these datasets?
• 16:35
  Speaker 1: How do we best tune models for this data?
• 16:37
  Speaker 1: So it's a real chance for companies, for researchers, to directly influence the direction and the focus of this whole important initiative.
• 16:44
  Speaker 2: Absolutely.
• 16:44
  Speaker 2: And there's one final, very direct and significant call to action in there.
• 16:50
  Speaker 1: Which is.
• 16:50
  Speaker 2: They are strongly urging companies to join the Edge AI Foundation community.
• 16:55
  Speaker 1: Ah, okay, not just participate, but formally join.
• 16:57
  Speaker 2: Right.
• 16:58
  Speaker 2: This isn't just about contributing a data set here or some expertise there.
• 17:01
  Speaker 2: It's about becoming an integral part of this larger collaborative ecosystem, an ecosystem that's collectively working to tackle these fundamental challenges in Edge AI.
• 17:10
  Speaker 1: And what's the benefit for the companies who join?
• 17:12
  Speaker 2: Well, by becoming members, companies get a unique opportunity.
• 17:15
  Speaker 2: They get to directly shape the development of the data sets, the evaluation methods, the best practices, all the things that will ultimately define the future path of this rapidly evolving field.
• 17:25
  Speaker 1: So direct influence.
• 17:26
  Speaker 2: Direct influence, yeah, and that can provide significant benefits back to their own research and development efforts.
• 17:32
  Speaker 2: It could potentially lead to a real competitive edge in this dynamic market.
• 17:36
  Speaker 1: So for any companies listening right now, companies involved in developing or deploying AI on edge devices, this sounds like an incredibly valuable opportunity not just to contribute, but also to gain a strategic advantage by being right at the forefront of these crucial standardization efforts.
• 17:54
  Speaker 2: That's precisely the key takeaway here it's a chance to actively participate in shaping the fundamental tools, the resources that the entire industry will increasingly rely on to effectively evaluate and, frankly, drive advancements in edge AI.
• 18:10
  Speaker 1: Okay, so let's just summarize the essential points for our listeners.
• 18:12
  Speaker 1: What's the absolute core message from this deep dive into the Edge AI Foundation's initiative?
• 18:18
  Speaker 2: I think the most crucial thing to grasp is that the Edge AI Foundation is proactively tackling a major bottleneck, a bottleneck that's really hindering the progress of Edge AI right now.
• 18:30
  Speaker 1: Which is that scarcity of standard, high-quality data.
• 18:33
  Speaker 2: Exactly, and this isn't just some abstract academic problem.
• 18:36
  Speaker 2: It directly impacts our ability to develop and accurately assess real-world AI applications, the AI running on those low-power, resource-constrained devices that are becoming more and more part of our daily lives.
• 18:49
  Speaker 2: So, by addressing this data gap, the foundation aims to provide a much clearer understanding of what's truly possible, what's actually achievable at the edge.
• 18:58
  Speaker 1: And by establishing this common ground, these realistic data sets, the hope is to catalyze even faster innovation and get more practical AI deployed.
• 19:06
  Speaker 2: Yes, leading to the practical deployment of AI across a wide spectrum of everyday devices, really connecting AI to the real world.
• 19:12
  Speaker 2: Just like their tagline said, and as we've really emphasized, active participation from the community is just vital for this to work.
• 19:19
  Speaker 2: Therefore, if your company is involved in any aspect of Edge AI designing algorithms, deploying solutions, anything in between we'd strongly encourage you to consider the significant value of joining the Edge AI Foundation community.
• 19:34
  Speaker 1: It's that opportunity to contribute, gain access to resources.
• 19:37
  Speaker 2: And ultimately play a key role in shaping the future of this incredibly promising field.
• 19:42
  Speaker 1: Absolutely, and you know, as Edge AI becomes more and more deeply integrated into the fabric of our lives, it really makes you think, doesn't it?
• 19:50
  Speaker 1: How will these kinds of collaborative initiatives, the ones focused on standardized, high quality data, how will they fundamentally transform not just the capabilities but also the reliability of the AI we interact with every single day?
• 20:02
  Speaker 2: It's a huge question.
• 20:03
  Speaker 1: It is Definitely something to ponder.
• 20:05
  Speaker 1: Thanks for joining us for this deep dive.