"No one AI model class will do the best advertising for marketers without the skills of others. The future will be a carefully orchestrated ecosystem - a team of specialists not a single genius. Epsilon's proprietary stack of roughly 15 machine learning models, embedded deep inside its ad server, was built from scratch precisely so it could make real-time buying decisions at the level of individual people, not just sites or audiences."
"Steve Nowlan illustrates the mismatch through an analogy: Imagine a college senior who has read everything on the internet - every research paper, Wikipedia entry, and forum thread. They seem extraordinarily smart, but put them to work inside one of the most complex, fastest-moving financial markets and they will get crushed by something that has been doing exactly this job, at exactly this speed, for over a decade."
"Every day, Epsilon's models process 800 billion bid requests. For each one, they make approximately 15 AI-driven decisions. All of it happens in under 10 milliseconds. It is a performance no LLM could replicate. Not because the tech isn't impressive but because it is the wrong kind of impressive. The 'P' in GPT stands for pre-trained. These models are frozen in time, shaped by human learning."
Epsilon, owned by Publicis Group, advocates for a specialized AI ecosystem in advertising rather than relying on a single dominant model. Steve Nowlan, SVP of decision sciences at Epsilon, illustrates this through an analogy: a highly knowledgeable person lacks practical expertise in complex, fast-moving environments. Epsilon built a proprietary stack of approximately 15 machine learning models embedded in its ad server to make real-time buying decisions at the individual level across 270 million unique U.S. individuals. The system processes 800 billion bid requests daily, making 15 AI-driven decisions per request in under 10 milliseconds. This identity-centric approach differs from most platforms that buy against broad contextual signals. Large language models cannot replicate this performance because they are pre-trained and frozen in time, making them unsuitable for real-time, individual-level decision-making in advertising.
#ai-ecosystem-in-advertising #machine-learning-models-specialization #real-time-bidding-decisions #identity-centric-targeting #llm-limitations-in-advertising
Read at Digiday
Unable to calculate read time
Collection
[
|
...
]