4A's Noteworthy News - 26th Edition
THE BIG STORY: Is AI hitting a wall?
Recent headlines have speculated about an "AI slowdown," with both OpenAI and Google confronting diminishing returns from traditional scaling laws—the notion that more data and compute automatically yield better AI models. OpenAI's upcoming "Orion" model reportedly brings only moderate gains over GPT-4, and Google's Gemini has similarly struggled. Yet, OpenAI CEO Sam Altman’s bold declaration on X that "there is no wall" pushes back against the narrative. So, what’s really happening?
The truth is nuanced. Traditional scaling techniques, like feeding models ever-larger datasets or deploying more GPUs, may well be reaching their limits. Challenges include the scarcity of high-quality training data, spiraling compute costs (e.g., OpenAI’s $100B data center proposal), and physical infrastructure constraints. However, AI developers are pivoting, focusing on innovation beyond brute-force scaling:
Hyperparameter tuning (optimizing external model settings (e.g., learning rate, batch size) to improve performance) and synthetic data generation (new data sets generated by AI); results have been mixed.
Post-training optimization, which fine-tunes models for specific tasks.
Test-time compute, where models "think" longer during problem-solving, promising better reasoning capabilities.
Additionally, the rise of AI "agents" like OpenAI’s planned "Operator," Google’s rumored “Jarvis,” and Anthropic’s “computer use” - all of which can automate complex workflows - signal a shift in focus.
For agencies hoping for a break from the relentless pace of AI change, this may not be the slowdown you are looking for. AI's trajectory may not be slowing as much as it is recalibrating. Expect tools that may be less about "more power" and more about "smarter power," demanding updated strategies for leveraging these advancements.
OTHER INTERESTING AI NEWS FOR AGENCIES
Mastering Prompt Engineering: Shelly Palmer offers an updated guide to crafting effective prompts for generative AI tools. Get the insights here.
Generative AI in Media Buying: From small agencies to Fortune 500s, learn how gen AI is reshaping this crucial advertising discipline. Full story.
AI Overviews Reshape SEO: Key Insights from 200,000 Keyword Study: Semrush’s analysis of 200,000 AI Overviews shows that Google’s AI-generated summaries favor informational queries, often skipping top-ranking pages, and rarely include ads—revealing new SEO opportunities and challenges for businesses aiming to stay visible in AI-dominated search results Explore the data.
AI Transforming Real-World Business: Microsoft shares case studies of how companies are thriving with AI integration. Learn more.
Generative AI’s Labor Impact: Harvard Business Review examines the labor shifts spurred by gen AI. Read the research.
Parents vs. Gen AI: Mediapost explores why parents are skeptical of AI, even as its influence grows. Get the full perspective.
Perplexity AI’s Ad Ambitions: Perplexity introduces ads to its platform after months of rumors and vague announcements. Dive in.
RECENT CAMPAIGN EXAMPLES, USE CASES AND CASE STUDIES:
How Taco Bell's AI Boosts Sales One Byte at a Time - Yum Brands, the parent company of KFC, Taco Bell, and Pizza Hut, shared that its investment in AI-driven marketing is paying off by boosting consumer engagement, reducing churn, and driving more purchases.
ThredUp's recent dive into AI-driven tools is paying off, with tangible results indicating a notable shift in shopper behavior and engagement. Since launching its visual search and Style Chat chatbot in August, ThredUp reports that customers using these tools are four times more likely to complete a purchase than those who don’t.
NOTE: This post originally appeared on LinkedIn here.
