How generative AI is changing the game

Panel of NFL, PGA Tour, and AWS executives recognizes 'tremendous impact' of AI implementation

During a panel discussion Monday at NAB 2024, representatives from the NFL, PGA Tour, and AWS presented detailed visions for how they are currently using or intend to use generative AI in the future.

Start a discussion AWS VP, Artificial Intelligence Products, Matt Wood We've outlined two main reasons why the latest generation of AI and machine learning is so exciting. “First, there will be a step-function change in the types of problems that can be solved. The type, complexity, and amount of data available and the context and understanding that can address those challenges will be far greater than was previously possible. Bigger, much bigger, much bigger.

“The other thing is that it increases the broad accessibility of these technologies,” he continued. It's never been easier or faster to actually use machine learning for any task. Deep data science expertise is no longer required. You don't even need to train your own model anymore. You can take what's there, use it, start tinkering with it, start tweaking it, and get it into production very quickly. ”

Wood offered specific and surprising expectations for enterprise productivity gains based on the “rich workloads” that will benefit from next-generation AI. “Today, I look at this as a technology that can increase productivity by about 10 times over the next 12 to 24 months, which is pretty amazing,” he said. Today it's probably 1.2 or 1.3 times higher. Will it be in the next five years? I think this utility will make him a 100x improvement. ”

The rise of agent-based workflows

Wood outlined three key growth areas for generative AI: media management, assistants, and agents.

The implications for media management, especially metadata tagging, are enormous. Much of what his PGA Tour and his NFL have said on this subject can be found below.

The second area is obvious and already pervasive. The industry will see the rise of AI assistants that will make life easier in many ways.

But the final use case Wood outlined seemed to have the potential for a major innovation breakthrough: agent-based gen-AI workloads.

“Today, these agents can complete tasks in about five minutes or so. Just offloading them works pretty well. Today, longer than that is not possible. But over the next year, In 2 years, 10 years, we'll be able to delegate more and more long-running tasks to AI for 15 minutes or even more than just 5 minutes. 5 hours, or even 5 days. That way, we can go from being 1.2 times more productive today to 10 times more productive in the next year or two.”

NFL Deputy CIO Aaron Amendria We explained how the application of optical tracking and AI to game operations has revolutionized more than just fan engagement.

“We launched Next Gen Stats in 2017 using sensors worn by players,” he said. “Collect data live and with low latency to unlock insights for broadcasters, storytellers, and even game operations. We strengthened our player health and safety efforts in 2019. We're bringing that data and insight directly to the fans. If you're at a game, you can hold up your phone and see an AR overlay of what's happening on the field. That's the kind of experience we're building for our fans.”

And he's thinking even further ahead. “More cameras, more sensor data, and more data points around the game itself. We're trying to build these systems that have multiple layers of value. These systems… We orchestrate multiple types of AI, computer vision, machine learning, generative AI, and an orchestration layer on top, often using natural language processing.”

“You can't have metadata tags for that.”

“If you're manually tagging metadata, you'll never discover everything,” Amendria said. Generative AI, natural language search, computer vision indexing, and media and content archiving (what's coming in, what's happened in the past, etc.) provide much more detailed information than traditional tagging. information will be available. ”

“You can't put metadata tags in all these contexts,” he said, citing the example of search and Patrick Mahomes being seen out in the rain with a Pepsi sign on his back and some fans wearing face paint. Ta.

He also expanded on the training and fine-tuning these models can perform to truly begin to understand the game. “We can train AI how to recognize certain soccer actions that cannot be tagged. [for example] This particular yard line, this player's fair catch signal, but I would love to see that in this particular state night game as well. ”

Once that model is built, these search contexts will be available across different interfaces used by people within the organization, rather than being siled into one specific role.

Amendria said: “This will reveal the value of our content across the organization at a larger scale. The amount of data coming in at scale is huge, so in the future we will be able to build in automation based on this. can.”

Wood emphasized the importance of this level of depth in metadata tagging. “I think the level of detail available in tags is much greater and more extensive than ever before. I was trying out the system the other day and an athlete was running and I asked the AI ​​to tag him. What came back was incredibly detailed, not only describing the type of concrete and what the athlete was wearing, but also the brand, manufacturer, and color of the sneaker. and identified it as a brand-specific color.

Amendolia has added the dimension of multimodal functionality. “Also, [with] Multimodal search capabilities (audio, visual, and data) allow you to uncover the spirit and emotion of your play. I want to find strong, aggressive plays. I want to see the concentration and determination on the players' faces. You can give it back. You can't put metadata tags on it. ”

Live production in the cloud: time, economy, and environment savings

Julie Souza, Head of Global Sports, AWS Here's what he said about innovation in all leagues, including the NHL: “Behind the camera, so to speak, we are seeing the growth of live cloud production technology. The NHL also did this on March 22nd, broadcasting live remotely produced NHL games.

“It means a lot,” she continued. “The savings add up on many levels. First, I've heard from colleagues at the league that preparation to produce a live show can take several days, maybe three or four. But with live cloud technology, that can be reduced to just a few hours.

“But I love this fact,” Souza continued. “Traditional game production methods would have emitted more than 2.05 tons of carbon dioxide. That would mean he would have to plant 34 trees and grow them for 10 years.”

Perfect for live, live commentary, and AI dubbing for all shots

Scott Gutterman, PGA Tour SVP, Digital Operations We provided other examples of how generative AI is opening up new fan engagement opportunities on a global scale. “We have an initiative called Every Shot Live that we started at The Players Championship four years ago. We are the first golf entity to publish every shot from the first drive to the last putt. ”

This means that during the tournament, the PGA Tour will have 48 streams on Thursday and Friday, approximately 24 of which will occur simultaneously. These streams will run alongside TV and ESPN Plus broadcasts, but will not have commentary.

“It's just a natural sound,” Gutterman said. “You can hear the players, too. These feeds are sent all over the world. It would be nice to have commentary for every stream, but it's not practical to find commentators for all 24 different streams. .”

TOUR is evaluating various generative AI technologies to determine what more can be done to drive individual streams. “Can we use generative AI narration?” he observed. “And if a Japanese network wants to see Hideki Matsuyama, can we get it to them? And can we do it in Japanese?”

Part of this was moderating the panel. thursday night football Sideline Reporter Kaylee Hartung, I'm a little worried. “What about the incredible background and history that some commentators have? [legendary sportscaster] Jim Nantz and his 37 masters [tournaments], for example? ” She asked, “Won't that be lost when these new tools come out?”

Scott had a thoughtful response. “For us on tour, it's more about empowering the broadcasters. Imagine giving Jim Nantz the opportunity to know, 'Today we're going to follow these six of his players.' After these players complete a hole, they have five talking points.If Jim Nantz [commentator/retired pro golfer] Trevor Immelman If necessary, they can see the opportunity come and incorporate it into the broadcast if they wish. ”

Multimodalization: biology, emergent properties, etc.

Asked to tell a broader story about the impact of generational AI, Wood offered a composite story that included biology and emerging properties. “Certainly these machine learning models have some very interesting properties. The bigger they get, the more modalities you combine, the more some of these properties start to emerge.”

In biology, predicting how proteins fold into a stable state is an intractable problem, he explained. But today's AI and machine learning models do this in about 10 seconds. “And the new property of these multimodal models appears to be detecting the protein's most stable state energy function.”

His advice for sports rights owners and distributors: “I want to be able to create larger models than ever before, capture multiple modalities, drive the manifestation of these new properties, and channel them into entirely new products and experiences for fans. I intend on doing it. “

Mr. Amendria ended his talk with an insight that resonated not only with the panelists but with the entire audience. “The only difference at this particular moment is that this technology everyone In your organization. It's not just for engineers. It's not just for data scientists. In the future, everyone in an organization will use AI to get their work done, increase efficiency, and create new and innovative things. You will be involved in every aspect of the business. ”