TL;DR
- Industry experts discuss the transformative role of generative AI on media technologies at Streaming Media Connect 2023 in a panel discussion moderated by Streaming Media contributing editor Nadine Krefetz and featuring George Bokuchava of Tulix, Globo’s Jonas Ribeiro, independent AdTech consultant C.J. Leonard, and Darcy Lorincz of the Barrett-Jackson Auction Company.
- The panelists stressed the irreplaceable role of human expertise, advocating for a multi-disciplinary approach in AdTech development. They also advocated for a balanced approach to training large language models on open versus enterprise data.
- Automated digital product placement emerges as a new frontier, with Globo using AI to integrate products seamlessly into scenes.
- Generative AI could also power what Bokuchava calls “dynamic ad generation,” creating real-time, hyper-relevant ads based on current market conditions, global events and other data points.
Imagine a world where AI not only assists in content creation but also plays a pivotal role in monetization strategies for streaming media. That future isn’t far off, as shown by a group of industry experts gathered for a panel discussion, “How Generative AI Will Impact Media Technologies,” at Streaming Media Connect 2023.
The conversation spanned the gamut from the use of public versus enterprise data to the transformative potential of generative AI on the advertising technology space.
Moderated by Streaming Media contributing editor Nadine Krefetz, the panel featured diverse viewpoints and approaches to harnessing the burgeoning technology. George Bokuchava, CEO of digital distribution platform Tulix, delved into the complexities of using AI to implement encryption and digital rights management (DRM) systems. Jonas Ribeiro, digital products, platform and AdTech manager at Globo, discussed the cautious approach needed when utilizing open data. Advertising technology and operations veteran C.J. Leonard, owner of MAD Leo Consulting, highlighted the time and cost-saving aspects of generative AI in advertising and content creation. Darcy Lorincz, chief technology officer at Barrett-Jackson Auction Company, offered his insights into how the renowned auto auction house is employing generative AI to create high-quality videos and other data-rich assets for promoting their vast inventory of collectors cars.
While the panelists were unanimous in acknowledging the transformative power of generative AI, they also stressed the irreplaceable role of human expertise. As Bokuchava put it, “Without programmers, we cannot implement it; it’s just not enough.”
This sentiment was echoed across the panel, highlighting the necessity for a multi-disciplinary approach. In a landscape where streaming platforms and consumer behavior are in constant flux, the panel agreed, the collaboration between developers, data scientists and AI experts is crucial for building robust, scalable and secure AdTech platforms.
Open Data Vs. Enterprise Data
In an industry that thrives on data, the panelists were quick to address the role of using public data versus enterprise data to train the large language models that power generative AI.
Ribeiro, with his extensive experience in data analytics for media companies, emphasized the importance of a balanced approach. “We are using both,” he said. Open data provides better results, generally speaking, but also demands more caution, he explained, because LLMs can be influenced by virtually anything on the internet. “So for this, we have a lot of people to check the outputs.”
Cost is another factor. “Not everyone can afford to have private data,” Ribeiro said, but for certain specific workloads it makes more sense. “We are trying to use the private data and work on it to get a more global perspective of what we are doing and what we are delivering to our customers and clients.”
Lorincz chimed in about the benefits of using proprietary data. “The majority of the 50 years of automotive information we have is ours, we own it,” he said. “Having that data is part of our competitive advantage.”
Training LLMs on proprietary data is a must for public-facing applications such as customer service, Lorincz insisted. “You have to train your own model if you really want the results you’re looking for,” he said. “When you train your own model… then you’re all-in on only your stuff. It’s only going to be talked about in your tone.”
Use Cases for the AdTech Space
Generative AI was able to significantly boost the number of auto listings and auctions at Barrett-Jackson, Lorincz said. “We have to write tens of thousands of car descriptions every month for our listing service or auction, wherever those vehicles may appear, and that was a lot of heavy lifting. A lot of editorial, a lot of knowledge you had to have, or just a lot of research,” he explained.
“We put the research tools, the information on every car sold for 50 years, everything, into our own language model. And now we can generate that editorial in seconds. It still needs people, because you still need to do some moderation, but as the machines learn more and more it’s less effort for us, so we can scale a business. Ultimately, we can do a million listings now and I don’t think that would ever been possible for any number of people before.”
Education and training is one area generative AI will definitely have an impact, Leonard predicted, pointing to traditionally time-intensive tasks that could be streamlined such as employee onboarding and organizational documentation. “Advertising is a high turnover space. If you’re in a job more than two years you’re the oldest tenured person there,” she noted, describing the typical six-month learning curve a new hire requires before they start delivering ROI. “Gen-AI will help in the future with education and ‘How do you get that person up to speed quickly,’ and maybe take out some of the roadblocks that we’ve had previously,” she added.
Automated digital product placement is another frontier that’s ripe for transformation in the AdTech sector, according to Ribeiro. He detailed how Globo is using AI to identify opportunities for seamlessly integrating products into various scenes. “So we identify some opportunities [for] putting a bottle up on the table, that maybe can be water or a beer or soda, and have more type of formats for the publishers so they can impact a lot of people more in a directed way,” he noted. While this technology is still in the research phase, its potential to revolutionize the way advertisers engage with audiences is significant, offering a more dynamic and personalized experience.
Among the panelists, Bokuchava arguably had one of the most groundbreaking ideas for leveraging generative AI in the AdTech space. While the industry is already familiar with the concept of “dynamic ad insertion” — the real-time placement of pre-made ads into streaming content — Bokuchava introduced a more advanced notion: “dynamic ad generation.”
“Imagine you have a company profile and allow for AI to generate the ad dynamically based on market conditions, based on the latest info, based on whatever is going on in the world,” he proposed. This concept takes dynamic ad insertion to the next level by not just placing the ad, but actually creating it in real-time based on various data points.
The implications are profound. Instead of merely inserting pre-made ads into content streams, advertisers could use AI to generate ads that are hyper-relevant to the current moment, making them more effective and engaging. While this idea is still in the conceptual stage, its potential to revolutionize the AdTech industry is immense, offering a more dynamic, targeted, and timely advertising experience.
