If you've attended our Intro to AI for Marketers course or have searched for articles on piloting your first project with AI, you'd see that natural language processing (NLP) and natural language generation (NLG) are two of the most tangible and understandable ways for marketers to get started with AI.
The potential for more intelligent marketing with NLP and NLG is seemingly limitless in advertising, communications, content marketing, email marketing, and social media marketing, but many critical questions remain unanswered.
At MAICON 2022, our Chief Content Officer Mike Kaput sat down with Christopher Penn (Trust Insights), Lisa Spira (Persado), and Jeff Coyle (MarketMuse) for a discussion on the capabilities of NLP and NLG in storytelling.
They discussed:
Watch this video plus all MAICON main stage keynotes, sessions, and panels with the MAICON 2022 On-Demand Bundle.
To kick off the session, Chris Penn helped us level set with definitions and explanations of NLP and NLG. Chris explained, "Natural language processing is when you're taking language and trying to teach machines to recognize it, deconstruct it, [and] understand how language works. [Natural language] generation then takes that learning and turns it into 'let's make new language.'"
Chris continued, "The underlying technology for all of this is something relatively new called a transformer. It's the ability for machines to look at blobs of tokens, text, or any kind of input, and then say, 'What have I learned? What is likely to come next?'"
"And those transformers are what have revolutionized and made possible everything we're seeing. Everybody who is doing up-to-date language generation and language processing is using transformers."
When it comes to large language models, their capabilities are very robust and striking. Jeff Coyle explained how newer language models are growing in size and capabilities at an alarming rate. This doesn't mean much to the average marketer, but the important part is that these large language models are being trained on data sets. They'll be quicker and more likely to be able to be tuned and specifically trained to support a particular business use case.
LLMs can now be trained using any data set and are constantly improving and iterating. The first ones were trained on Wikipedia and/or larger data repositories, so as the models learn more and work with better information, these models improve.
Two years ago, Google announced its Multitask Unified Model, MUM. Jeff explained, "MUM not only understands language but also generates it. It's trained across 75 different languages and can do many different tasks at once, allowing it to develop a more comprehensive understanding of information and world knowledge than previous models. And MUM is multimodal, so it understands information across text and images and, in the future, can expand to more modalities like video and audio."
AI technology is changing at a record pace. The panel talked about how writing technologies were not great four years ago, and last year were going from okay to good—and are now continuing to improve.
As Chris Penn said, "Right now, the fine-tuning is still not quite there. It's much better than it used to be, but it's still not as good as you writing it yourself." That said, you don't need to know the ins and outs of transformer models to understand that hundreds of technologies are building upon these AI systems, and some can help you.
You may not get all the content you want to create for your business with AI, but AI can help you achieve specific goals. An example of this would be if you want to analyze sales calls to try and understand:
Lisa Spira explained how her love of language, stemming from childhood, led to a degree in linguistics and continued studies in onomastics, which in turn led Lisa to Persado, where she heads up content intelligence. Her team brings human intuition to machine learning and works to iterate their product so humans can work more closely and more effectively with the machines.
Mike has a journalism background, and Chris and Jeff have computer science backgrounds. This was also discussed in the Next-Gen Agency workshop; three panelists have journalism degrees, and one has a data science degree. An excellent blend for us all to learn from each other!
AI technology can analyze large amounts of data at scale in ways that are impossible for a human to achieve. Technology can help a content creator overcome a mental block, an email marketer with testing, analysts with data analysis, and more.
Industries will evolve, technology will evolve, and human roles will need to be adapted thanks to AI. It's up to the company and the employee to stay ahead of the industry to know what tasks machines can do, how the human will be in the loop, and in what areas humans need to keep or gain skills.
Human progressive and human regressive companies both exist. A regressive company will look at AI and say, "This is a cost savings tool." If they have ten people on staff who perform the same job and can automate 40% of the roles, then they'll remove four people and preserve the remaining six. Regressive companies, those that want to ONLY maximize profits, will cost people their jobs.
Progressive companies will look at those same people and say: "We've got 10 people. We can automate away 40% of the role. What else can we train those people to do? What else can we have them upskill?"
Our panelists offer these tips on piloting AI:
Do you want to be a next-gen marketer? Use the resources at the Marketing AI Institute, read our blog posts, take our Intro to AI for Marketers class, attend webinars, join our community, download reports, guides, and templates (all free), read Marketing Artificial Intelligence, look into AI Academy for Marketers and Piloting AI Bundle, and our annual MAICON—Marketing AI Conference.
Get access to all MAICON main stage keynotes, sessions, and panels with the MAICON 2022 On-Demand Bundle.