?? The Creativ Brief: Retrieval Augmented Generation

I learn something new every time we have a product call at Creativ Strategies.?

This week my partner Vibhu Bhan explained a bleeding-edge AI concept, RAG, retrieval augmented generation.

RAG is an AI framework that overlays a large language model (LLM) onto a body of knowledge to make it queryable.?

Chat GPT stands for chat generative predictive transformer. Their LLM is general, fed by millions of disparate pieces of text, to create a model that can predict the next word or series of words in response to a prompt. The model is general, the body of knowledge general, the output is general. That’s why one can ask for haikus about cookies, code a restaurant website, or generate 5 interesting carrot cake recipes… I’m hungry.

Generally, a RAG model is trained on a specific body of an authoritative, specific body of work to create a queryable system that will go and retrieve information when prompted to do so. You can think of this as a supercharged robotic librarian or golden retriever. You say fetch and the system retrieves it.?

RAGs also differ from search engines in one key aspect, conversation ability. A search engine is keyword prompted while a LLM can process and respond with full sentences or paragraphs. Traditional internet search engines like Google use similar AI-data science to index websites and generate a predictive list of links when queried. However, a RAG could give you specific samples, quotes, images, and information in a conversational format.?

Critically, RAGs are also deployed to specific bodies of knowledge. The body of knowledge is constantly updated, authoritative for the use case, and bespoke to the industry while the LLM is trained to sort and filter that specific body of information in question.?

A lawyer could ask a RAG model trained on a body of IP law to retrieve “fair-use cases from the last decade for copyrighted images where the claimant won”.?

A history professor could query a RAG model trained on Medieval texts to find “instances of Catholic doctrine pertaining to Germanic tribes from 1200-1500 AD”.?

A consumer could query an online retailer for “wide-brimmed fedora hats organized by highest reviews.”?

Specific bodies of information. Large language models trained on that information. A queryable search. Conversational prompting. That’s Retrieval augmented generation. Now about that hat…?

3 Stories Dominating Media and Tech Headlines

At the D23 Expo, Disney unveiled a broad range of upcoming movies, TV shows, and projects across its major brands, including Marvel, Star Wars, Pixar, and Disney Animation. Key highlights included the first look at "Moana 2," announcements of "Avatar 3" and "Toy Story 5," and new projects like a live-action Monster Jam movie, with performances and teasers adding excitement to the event.

Why it matters: By unveiling new content and expanding beloved franchises, Disney reinforces its ability to capture diverse demographics, set industry standards, and influence the competitive landscape across media, streaming, and film.

"Deadpool & Wolverine" has become the latest Marvel movie to surpass $1 billion in global box office earnings, marking a significant milestone for Disney’s MCU. The film, starring Ryan Reynolds and Hugh Jackman, is on track to become the highest-grossing R-rated film ever, highlighting its strong appeal despite recent struggles within the franchise.

Why it matters: This success underscores the enduring popularity of superhero films and the critical role they play in Disney's entertainment strategy, especially as the MCU seeks to maintain its cultural and financial dominance amidst evolving audience preferences.?

NBCUniversal’s streaming service Peacock is making a significant play for NBA media rights, investing $2.45 billion annually to secure exclusive NBA games starting in the 2025-26 season. This decision follows the success Peacock saw after streaming an exclusive NFL playoff game, which added millions of subscribers who stayed for both sports and non-sports content.

Why it matters: The move is crucial as it positions NBCUniversal to compete with streaming giants by leveraging live sports to drive subscriber growth, boost advertising revenue, and ensure long-term viability in the rapidly evolving media world.

Creativ Spotlight - Greg Harrison, MOCEAN, Up Next Podcast

Greg Harrison, Chief Creative Officer at MOCEAN, joined Gabriella Mariella on the Up Next Podcast to talk all things AI and creative work. Greg shares valuable insights into the practical applications, benefits, and implications of AI for creative services, agencies, and marketing professionals.

Checkout the Podcast HERE!!!!

Stat of the Week - Trends in Social Shopping?

Facebook leads the pack, with 44% of respondents in the Q2 survey indicating they would likely purchase a product directly from the platform, an increase from 34% in Q1. YouTube followed closely, with 39% of consumers expressing willingness to buy from it, up from 31% in the previous quarter.

Instagram (29%) and TikTok (26%) trailed closely behind in terms of social shopping likelihood, with both showing increases from the previous quarter's numbers.

Despite the recent rise in TikTok Shop’s popularity, older social platforms like Facebook and Youtube are still leading the way in social shopping.? For businesses and advertisers, social commerce is not just an emerging trend but a critical channel for reaching consumers.

One Fun Thing - Generative AI Hype

Generative AI has built massive hype from its tangible nature. It can create efficiencies, generate essays, and code a website. But it might not be amazing as people think, according to Eric Siegel. AI has more pertinent, important, and useful applications for predicting outcomes than generating internet memes.?

Checkout this video in which he explains applications for predictive AI in today’s world.