At the 20th annual MozCon, Britney Muller, Founder of Data Sci 101, delivered an eye-opening presentation on AI and its impact on digital marketing.
Her session, “The Hidden Side of AI: What Marketers Need to Know,” provided a comprehensive overview of AI’s current and future potential.
Muller discussed the ethical considerations, practical applications, and limitations of AI, offering valuable guidance for marketers.
The Emergence Of Generative AI
Muller began by discussing the rise of generative AI, which lies at the intersection of AI, machine learning, deep learning, and natural language processing (NLP).
She explained:
“Generative AI, in particular, emerged from this interesting overlap of fields.
We have AI that houses machine learning. Within machine learning, there’s deep learning. And then human language comes into play with NLP or natural language processing.”
A significant portion of Muller’s presentation focused on the crucial role of training data in AI models.
She emphasized:
“I used to say AI reflects its training data, and I’m going to double down. It magnifies its training data.”
Muller highlighted the lack of diversity in datasets like Wikipedia, where contributors are predominantly male, and how this can perpetuate biases in AI outputs.
Photo taken by author at MozCon, June 2024.Practical Applications & Limitations Of AI in Marketing
What Gen AI Is Good At
Muller presented a wide range of practical applications for AI in marketing, as shown in one of her slides.
Photo taken by author at MozCon, June 2024.She explained:
“LLMs, in general, are good at all of these things, and I’m of the unpopular opinion that generating content is one of their worst capabilities. They’re way better at sentiment analysis, labeling things as categories, providing code support.”
Additionally, she shared a slide highlighting specific GenAI SEO/marketing applications, including:
- Automatic titles & meta descriptions
- Aata cleaning
- Code assistance
- Accelerating creativity and ideation
- Personalized outreach
- Sentiment analysis
- Refurbishing content
- Chatbots
- Meeting note transcription
Photo taken by author at MozCon, June 2024.What GenAI Is Bad At
Photo taken by author at MozCon, June 2024.Muller discussed the limitations of LLMs, which struggle with tasks requiring:
- Factual accuracy
- Common sense reasoning
- Understanding context
- Handling uncommon scenarios
- Emotional intelligence
- Math/counting
Marketers should recognize these strengths and weaknesses when incorporating AI into their strategies.
Prompt Engineering Tips
To help marketers utilize generative AI, Muller provided actionable tips for prompt engineering.
Photo taken by author at MozCon, June 2024.Her three suggestions were:
- Explain the task as you would to a person
- Use examples to illustrate what you want
- Give the model a “role” and tell it about the intended audience
She advised:
“Explain the task or the problem like you would to a person. There’s been so much research on prompt engineering, and oh, these things work, but these things don’t. The biggest takeaway from all of this research is examples. It’s just showing the model, hey, this is good or bad, and we want the output to look like this.”
Muller shared a slide of generative AI tools and resources such as Colab, Kaggle, GPT for Sheets, Ollama, WordCrafter.ai, and her own DataSci101.com.
Photo taken by author at MozCon, June 2024.Key Takeaways & The Future Of AI In Marketing
Muller concluded her presentation with several key takeaways captured in her closing slide.
Photo taken by author at MozCon, June 2024.She emphasized the need for a people-centered approach to AI, recognizing its potential as an assistive technology rather than a total replacement for human expertise.
Key takeaways included:
- GenAI is a predictive technology
- A model is only as good as its training data
- Marketers have the power to think up the next brilliant GenAI application
- Show up online where conversations about your product/service are occurring
She stated:
“We need to be talking more about people-centered AI, right? What will be the best model to support the people we work with? And that this is predictive technology. A model is only as good as its training data and is an assistive technology. This isn’t a full replacement of you, and it won’t be.”
In Summary
Muller’s insights serve as a valuable guide for navigating the complex world of AI.
Throughout her presentation, Muller reiterated that AI should be viewed as an assistive technology rather than a complete replacement for human expertise.
She encouraged marketers to identify tasks AI can help accelerate or automate while maintaining a human touch.
Muller’s key message to marketers is to uphold ethical practices, prioritize human needs, and capitalize on AI’s strengths while recognizing weaknesses.
Featured Image: abnalladin/Shutterstock
