Artificial Intelligence (AI) has revolutionized many industries, and political marketing is no exception. AI algorithms can help political campaigns in a variety of ways, such as micro-targeting voters, predicting election outcomes, and analyzing large amounts of data. With the help of machine learning algorithms, political campaigns can now analyze vast amounts of data on voters’ demographics, behavior patterns, and political preferences to understand their target audience better. They can then tailor their messaging and advertisements to resonate with specific groups of voters. The use of AI in political marketing is still in its early stages, but it has already shown promise as a powerful tool for campaigns to reach voters effectively.
The Paragraph Above Was Written with AI
Since there’s nothing worse than starting a writing project with a blank page, I opened ChatGPT and typed in the prompt:
Prompt: “Write a 100-word introduction paragraph about how AI can be used for political marketing”
Wow. That’s almost too easy. Maybe I should have ChatGPT write the rest of the article and take the afternoon off.
Prompt: “Should I take the afternoon off and let ChatGPT do all my work?”
Result: As an AI language model, ChatGPT can assist with various language-related tasks, but it is not intended to replace human effort entirely. While it can provide helpful information, generate text, and answer questions, it may not always understand the context and nuances of a particular task or be able to handle all the responsibilities required. It’s important to use AI tools like ChatGPT as an aid, rather than relying on them completely.
Fine, I guess I’ll just use it to help me out.
Copywriting with AI is nothing new. Programs like Jasper have been available for years and are used by marketers and copywriters with varying levels of success. What is new is the quality of the output, and the commitment from big players like Microsoft and Google to develop AI further.
AI writing, particularly of non-creative, research-based content, has already taken the academic world by storm and threatened to be the end of school writing assignments. There have been countless stories around AI recently, with some saying it’s the end of creative jobs and others saying it will change the world and the way we use the web. There’s little doubt AI will significantly alter the way we work, but it’s difficult to know how big a shift is coming when dealing with programs that are still in their infancy.
ChatGPT currently runs on GPT-3 (Generative Pre-trained Transformer 3), and improvements are on the horizon, with GPT-4 expected to roll out later this year. This post won’t get into the details of how these programs work, but rest assured that better results and more capabilities are coming soon. This is only the beginning.
A Search Engine Replacement?
A potentially huge use for AI programs is to return answers to questions you may have used a search engine for in the past. Here’s what I get when I type the prompt:
Prompt: “Write 50 words on the history of ChatGPT”
ChatGPT is a state-of-the-art language model developed by OpenAI, which was first released in June 2020. It is based on the Transformer architecture and is fine-tuned on a diverse range of internet text to produce human-like responses to text-based inputs. ChatGPT has become widely popular in the language generation and conversational AI communities due to its impressive language generation capabilities.
That’s pretty good, and helps you understand why Google is concerned about the impact it could have on their search-based business model, so much so that they’ve called in their founders Larry Page and Sergey Brin to help work on their response. Last week, Google unveiled their own AI language model, Bard. Is it the same thing as ChatGPT? Let’s see what ChatGPT has to say about it.
Prompt: “Write 75 words about the difference between Bard and ChatGPT”
Bard and ChatGPT are both AI models developed by OpenAI, but they differ in their specific use cases and capabilities. While ChatGPT is a conversational AI model trained to generate human-like responses to text inputs, Bard is a creative AI model designed to generate original text in a specific style or genre. ChatGPT can be used to build chatbots, answer questions, and generate text, while Bard can be used to generate poetry, fiction, song lyrics, and more. Additionally, ChatGPT is trained on a large corpus of text data, while Bard is fine-tuned on a smaller, more specific dataset to enable it to generate text in a particular style.
That’s interesting, but is it accurate? They don’t cite their sources, and more research would be necessary to confirm it. Accuracy has been a bit of a problem with OpenAI based text results. Take this prompt from Google Bard’s unveiling last week.
Prompt: “What new discoveries from the James Webb Space Telescope (JWST) can I tell my 9-year-old about?”
Bard returned a list of several discoveries, with the last one being:
Took the very first pictures of a planet outside of our own solar system.
Unfortunately, users on Twitter were quick to point out that the first picture of a planet outside our solar system was taken in 2004 at the European Southern Observatory, and Alphabet lost $100 billion in market value after the presentation. It turns out AI can be very wrong and very confident about it, just like that one relative at Thanksgiving.
If you’ve visited COMPETE’s site before and have been targeted with our display ads, you may have seen some with an animated shark mascot. One of the first things I did with Midjourney was create a shark character wearing sunglasses.
Impressive results. But for all its amazing qualities, AI art typically has some odd flaws if you look closely. In my shark example above, the anatomy of the shark is very hit or miss. Midjourney famously has difficulty rendering human hands with five fingers, and Stable Diffusion occasionally renders two heads on a body.
For my shark creation, I refined the prompt, asking for the shark to walk on its legs, stand in front of the Capitol, and be drawn in a cartoon style. This is what I got:
The results are interesting, but not what I had in mind, and even with several rounds of refined prompts, I was never able to get something I would actually use. Certainly, over time the results will improve, but for now, these are good starting points for a brainstorming session or as quick concepts before creating sketches.
Selecting stock photography is always challenging, particularly with its noted lack of diversity and models that appear in many places. Setting aside the issue of creating images of “fake news”, it is an appealing idea to generate a unique image that is exactly what you need. Below is a common type of stock photo we might want, with a prompt of:
Prompt: “Realistic photograph of a diverse group of people smiling outside of a voting booth”:
For editorial use of public figures who have many images online, AI is able to create interesting results that may be good enough to use as an illustration for a social media post or meme.
The Problem with AI Art
AI Art is using human-made artwork as the foundation of its data set in order to create new images based on user prompts. However, the artists who created the original work are not credited or compensated. Artists are mad. Presently, there is little to no recourse an artist can take, and it is difficult to remove the artwork from the data sets. There have also been instances of photos from private medical records being found in the data sets, and Getty Images is currently suing Stable Diffusion due to the program scraping all of its watermarked sample images. The end result of this is still to be determined, but it’s worth keeping in mind before using for paid client work.
A Potentially Dangerous Slope in a World Full of Fake News
The possibilities for AI art to be used to create disinformation are huge. Back to the issue of AI art being used to create “fake news”: it was used earlier this month to create pro-police images that supposedly took place at a French protest.
Many are hoping governments step in to regulate this, but I don’t think that is likely, and it’s hard to imagine most people being able to discern what is real and what isn’t. Since the dawning of photo manipulation, from spirit photographs to photoshopped images, people have not proven to be great judges of reality. Don’t expect that to change.
The Legal Question
How is COMPETE using AI?
Prompt: “Write an Instagram caption for a photograph of a politician eating a burger at a campaign stop.”
“Fueling up for the campaign trail, one burger at a time 🍔🇺🇸 #BurgersForChange #CampaignFuel #ElectionSeason”
Prompt: “Write a funny Instagram caption for a photograph of a politician eating a burger at a campaign stop”
“Taking a break from kissing babies to kiss this burger 🍔😘 #ElectionFuel #BurgerCravings #PoliticiansHaveCravingsToo”
Prompt: “Write a sad Instagram caption for a photograph of a politician eating a burger at a campaign stop”
“Even politicians need a moment to escape from the stresses of the campaign trail 🍔💔 #ComfortFood #ElectionBlues #InNeedOfABreak”
We are not currently using AI image generators for any client work, but may do so for the creation of social media memes or as placeholders in designs to show clients what type of image we are looking to receive.
The implications of AI upending search engines could be huge to all Digital marketing industries and is something that we will be watching closely to help our clients tailor their online presence to accommodate these tools.
AI will continue to advance and generate content to the preferences of the end consumer but will give its best output with a human guiding it. Watch this space to learn more about how we use this technology in the future.
Prompt: “Write a witty closing sentence to a post on using AI for political marketing”
“With AI on your side, your political campaign will be running like a well-oiled machine, leaving your competitors in the dust and your voters saying ‘what just happened?'”
Want to chat with us about AI, Political Digital Marketing, and/or the intersection of the two? Get in touch with me!
Or, if you are looking for a gift for your kids, I am also a published children’s book author and illustrator! Check out my work here (I promise I didn’t make these with AI): https://adamperrybooks.com