Submind YouTube summaries
Thumbnail for When Does AI Inference Actually Need to Run at the Edge? | Ari Weil, Akamai

When Does AI Inference Actually Need to Run at the Edge? | Ari Weil, Akamai

Watch on YouTube

Video summary

The discussion clarifies that not every artificial intelligence workload benefits from running inference at the edge; specifically, tasks involving model training, pre-training, or post-training do not require real-time responsiveness or federation. These processes are primarily concerned with updating weights, managing embeddings for future use, and optimizing storage efficiency rather than delivering immediate answers to users. Consequently, heavy computational tasks like fine-tuning models can remain centralized without negatively impacting user experience, as the latency associated with these background operations is irrelevant to the end-user's immediate needs. In contrast, specific high-value use cases demand that inference occurs closer to the user to ensure speed, security, and reliability within strict timeframes, typically under thirty seconds. A prime example provided is a mortgage application powered by a large language model, where a user expects instant feedback on loan eligibility after providing personal data and undergoing real-time checks like credit or background verification. Similarly, in the real estate sector, companies must handle zero-click searches to maintain visibility against competitors and support complex workflows such as scheduling open houses or processing multi-step loan approvals. These scenarios involve significant financial risk and high stakes, requiring distributed AI architectures that can deliver secure, instantaneous decisions before a user loses interest or fears data compromise. Beyond financial services, the necessity for edge inference extends to physical AI devices and environments where real-time interaction is critical. This includes anomaly detection in autonomous vehicles and robots, augmented reality applications on assembly lines, and medical or military facilities utilizing VR goggles. Even simpler digital interactions like real-time multiplayer online gaming rely on low-latency inference to allow avatars and game worlds to react instantly to player actions. In these contexts, the hardware and data must be positioned near the point of interaction to maintain the seamless performance required for physical engagement or immersive experiences, distinguishing them from batch processing tasks that can tolerate higher latency. Ultimately, the industry is increasingly recognizing that the decision to distribute AI depends on whether an application requires instantaneous responses and low latency for high-risk or interactive scenarios. As companies invest in new methods for shuttling data across systems, they are deeply considering where hardware should reside, where data flows, and how the final application experience will be consumed. This strategic shift is driven by the need to support emerging sectors like online sports broadcasting, news delivery, and e-gambling, which all reward immediate feedback loops. The consensus is that while centralized training remains efficient for model development, edge deployment becomes essential whenever the user expects a real-time answer to a high-value query or interacts with a dynamic physical environment.
Read the full video transcript
Are there specific workloads or use cases that make the strongest case for running inference closer to users, devices, or data sources, or do you feel that every use case, every workload is ideal if it is running closer? >> Oh, no, definitely not. I mean, if I were going to be training a model, pre-training or post-training of my model to do anything from, you know, adding weights to the model or embeddings that I might need later, if I'm if I'm pairing down a model that I might use in a future workload, and I want to make it a smaller, you know, overall model for storage reasons and what I can hold resident in memory, we don't need those things to be federated. Those don't have to be real-time and responsive in this way. If I've already got a model, and the use case is, "Hey, I have a an LLM-backed um mortgage application uh or mor- mortgage um approval application, then I can think about when the user logs in, they're going to provide me a certain amount of data. I might want to access my data about that user, and maybe I'll perform some rag activity to do a credit check or do a background check on that person, but all of it needs to come together in under, you know, 5 seconds typically, certainly under 30 seconds, and provide me with an answer that says, "You are or not approved for this loan. You are approved for this much, and here are the next steps to actually capitalize on this loan." I think about that in a workload, like an end-to-end workload sort of perspective, imagine that I am a real estate company, and I might have anything from my online listings. I need to make sure that all of my listings render completely correctly, consistently, but certainly that quickly piece needs to be top of mind for me. I need to ensure that in LLMs, any zero-click searches are actually having my page be cited more frequently than my competitors' page, so that people find me in the first place. Just finding that real estate listing already has a significant amount of right now architectural rework and application rework that companies are going through because of this phenomenon of zero click searches and LLM scrapers. Now, that's just again finding my listing. Next thing is maybe I want to understand can I set up a open house time? Can I schedule a walk through with the realtor? Maybe I want to find out what it might take for me to actually get approved for a loan on this property. Those are different work streams that I'm building into my application and each of those as well needs to be real time. You would argue that scheduling something on the realtor's calendar is a pretty lightweight use case and you would be right. There's not much of a appeal for AI in that sort of a use case. But confirming somebody's eligibility and what level or what size of loan they could get is very much something that AI can help with and is very common in the financial services industry specifically around insurance and and warranties and mortgages that we need to run quickly, securely, and get an answer back to that user before they lose interest and walk away or before they worry that maybe their information was compromised. So, now I've got a listing that I've received, I've applied for a loan, maybe I've set up time for an open house. The next thing that I might consider is the actual set of transactions where I want to procure the home or I want to finalize that mortgage, I want to purchase the house, I want to go through other signatories. Those types of things typically as well will require some form of multi-step approval process that is currently being automated by the people that are managing it. And so those types of things where you're sitting there interacting in real time, where you're expecting a response for something that is very high value and typically has a pretty good amount of risk associated with it, that's a great example and a great use case for distributed AI because you are running inference and you want those answers to be fast, reliable, and secure. If I think about some other ones that are related, you know, anytime that I might do anomaly detection, that could be in an autonomous vehicle, it could be in a robot or other physical AI device, it could be for augmented reality if I'm doing, you know, work on a on a conveyor belt or an assembly line. I might be in the the medical field, I may be in the public sector or even in a a military sort of a facility, but there's real-time inference happening through augmented reality and virtual reality goggles today that lend themselves to this. And then we can think about, you know, even simpler things like real-time multiplayer online gaming. You're creating worlds, you're you're having avatars and other characters in the game that are reacting and responding to people in real time. These types of things you keep hearing real time or where you have a physical engagement with the application, those are typically the things that would warrant, you know, distribution to maintain that low latency and that real-time performance. But I think we're just starting to understand this as an industry, and even things like how we'll create online sports broadcasting, news broadcasting, you know, gambling {slash} you know, e-gaming, things that really do reward latency and instantaneous responses are where we are seeing companies invest in this area and start to experiment with new ways of shuttling data across systems and starting to think really deeply about where does the hardware go, where does the data go, including the models, and then where is that application experience going to be interacted with, and that's what's driving the considerations.