Recent developments in AI have been nothing short of amazing. With the rapid pace of advancements,...
Beyond ChatGPT. A Guide for Marketers.
Recently, especially since the launch of ChatGPT, generative AI has been making headlines and its presence is increasingly felt everywhere. Microsoft has been the biggest newsmaker in the last several weeks, with their strategy to disrupt the search landscape as well as integrate natural language processing tools to their MS Office and other products. Google responded with the soft launch of Bard, Meta continues integrating AI across all of their products, opening access to developers for further experimentation, and Amazon’s AWS claims “the broadest and deepest set of artificial intelligence (AI) and machine learning (ML) services and supporting cloud infrastructure.”
One of the best summaries of advances in AI in the last 6 months is by Martin Harbech:
“In the past six months, AI has helped…
Win a fine arts competition.
Immortalise Darth Vader.
Pass a MBA exam at an Ivy League university.
Assist in a legal court case.
Win a Cannes Short Film award.
Interview the ghost of Steve Jobs on a podcast.
Create video content from just text input.
Start an 'infinite conversation'.
Explore the ‘dark matter’ of the protein universe.
Enable a developer to talk to her inner child.
Unlock speech-to-speech translation.
A few (of many) recent AI headlines. 👆”
Today, we’ll talk about how generative AI relates to marketing.
Exploring Generative AI: Key Facts for Marketers
Generative AI is a type of artificial intelligence that uses algorithms to generate content. It is capable of producing completely unique content, ranging from images to text and audio. Its applications are incredibly varied, from creating interactive gaming experiences to aiding in digital marketing content production.
This type of artificial intelligence works by analyzing large amounts of data and using the patterns it sees to produce something new. For example, it could take a collection of images and generate an image based on the patterns it finds. This image will be completely unique and unlike anything previously seen.
AI also uses deep learning techniques to understand how different elements of information interact with one another. See below on “How AI Works”.
Value of Generative AI
Generative AI is:
- Capable of creating unique marketing content. It can generate thousands of variations of images, videos, and texts, allowing marketers to develop campaigns efficiently.
- By bringing together consumer insights and AI, it is possible to create content that is engaging and personalized. This can help create texts, images, and videos that stand out and resonate with the target audience.
- More cost-effective than other methods of content creation, enabling marketers to save money while achieving high quality.
- Incredibly fast, accelerating content creation multiple fold.
Despite 3 rounds of quality assurance by humans, we are seeing incredible efficiencies:
Understanding How AI Works
We believe it is very important to understand these core principles to know what’s possible and what the limitations are.
The basis of generative AI is an LLM, a large language model with billions of parameters in it. GPT-3, a model used by most companies, has been trained on 175 billion parameters. GPT-4 has just been released, but its size isn't known yet. But what we know is that the GPT-4 model is multimodal (can work with text, images and video at the same time).
While it is incredibly impressive, there are several things to keep in mind:
- In “GPT”, “P” stands for pre-trained
- It is not real-time and has a finite dataset
- At the moment of this writing, the GPT-3.5 through September 2021.
In the current version of GPT models, the text is created by predicting the next word based on the data it was trained on.
Let’s take a look.
Here, we ask AI to write about 10 top things to do in New York City.
And here is how AI decides what words to choose:
Issues Related to Generative AI
One of the most significant issues with generative AI is that it can produce content that is not always accurate or trustworthy. It happens when the dataset used to train the artificial intelligence lacks the information you are asking to write about. The AI will keep predicting the next word when it generates content, and it will do that based on the data it has been initially trained on. Remember, for now, it’s not real-time – it’s a finite dataset in the foundational model, such as GPT-3.5 or even GPT-4.
Another issue is the possibility of bias. It has the exact same cause. If the training dataset contained biased texts, AI's output is also likely to be biased. With 175 billion of parameters and some training data going as far as centuries, there will no doubt be biased content.
Hence a crucial need for human Quality Assurance when you use AI to create marketing or any other content. Intentful has 3 QA tiers before delivering the work to a client, and artificial intelligence can additionally perform self-checks (however, a subject matter expert is required for any AI-aided work). For Enterprise clients who would like to have their own AI platform, we highly recommend implementing a similar review and approval process before the content is published.
Train AI to know your brand
A GPT model can be trained. The best example is ChatGPT – it is a GPT model trained to have a conversation.
If AI could be trained on those 175 billion parameters and knows history, physics, languages, and more, or like ChatGPT, it could be trained to have a conversation, AI can be trained to know your brand – brand's voice and facts, what ads and content have worked best, and become a "super employee" that helps your human team deliver more, faster. It can't replace your core staff, who determine content strategy and creative; rather, it serves to augment your content team's productivity. Intentful trains AI to understand our customers' businesses. While we can’t share examples of clients’ trained models, see what is possible by taking a look at how we collaborated with Oxford University's Saïd Business School to produce AI-generated content for a debate at Oxford Union. We taught AI to impersonate William Shakespeare, Jane Austen, Oscar Wilde, and Winston Churchill pretty well.
AI Use Cases
In content creation, the use cases are diverse. While AI can generate images, videos, and audio, for now (March 2023), the text is the most advanced part of generative AI and can be applied to commercial use immediately. Images and videos had a major breakthrough, but they are not yet as ready for commercial use by businesses.
Some ideas on the types of content you can produce with the help of AI.
- Ad copy (Google Ads, display, etc.)
- Content matched to customer intent
- SEO-friendly website copy
- Blog posts and articles
- Email marketing
- Press releases
- Content plan, article, and blog post ideas
- Writing or updating product descriptions
- Unique content and variations for SKUs
- Product FAQs
- Upselling and cross-selling content
- Category descriptions, product tags, meta
- Dynamic variations of e-commerce ad copy
- Help Center articles
- Knowledge base content
- Answers to common customer questions
- Product and how-to guides
- Troubleshooting content
- Tutorials, and more
- Employee handbooks and onboarding materials
- Company policies
- Product manuals
- Sales scripts
- Compliance training programs
This list can be continued, and if you’d like to learn more about generative AI and how it can be useful for your business, join one of Intentful’s events and meet us online or in real life. Intentful is a frequent speaker on the practical use of AI, including at the CMO Summit in New York, The Future of Marketing conference in London, the eTourism Summit, Google and MMA Possible event in Miami, University of Oxford, University of Exeter, to name a few.
We also run custom educational sessions for enterprises around the world. Contact us to schedule a workshop for your company.
AI in Marketing Use Cases Webinar
Intentful’s Marina Petrova (CEO) and Bruce Amick (COO) will host an online session Monday, March 20, at 10 am EST, dedicated to how marketers can use artificial intelligence.
Here is what the webinar will include:
- Practical use cases of AI and NLP in marketing
- A real-time demo: See how AI can create content for you
- What is NLP and AI, in simple terms
- The benefits of using AI in your marketing efforts (AI does NOT equal people losing jobs)
- How does AI know what content to create?
- Why are some facts made up, and how to manage that?
Register for AI Use Cases in Marketing webinar.
Check our Events page for other upcoming webinars.