AI marketing videos: so bad they’re good? 🤨

Plus, saving 4+ hours using AI to create a 20-minute event speech

Author

Charlie Braithwaite
May 05, 2023

A newsletter about AI and Marketing by Tom & Charlie

Welcome to AI Marketing School, where we’re helping marketers harness AI to its full potential. Share the love. Forward this email over to your favourite marketing pals, or go ahead and sign them up here.

 

AI Marketing School Newsletter #19

AI-generated marketing videos: so bad they’re good?

A couple of AI-generated ads have recently hit the headlines… for all the wrong reasons.

Synthetic Summer, posted by London-based production company Private Island on Instagram, features all of the usual trappings of a stereotypical beer commercial.

Sunny weather? Check.

Barbecue? Check.

Smiling beer drinkers? Check.

Flaming balls of fire and giant beer cans for hands? Wait—what?

It’s so bad it’s good, and it gets even weirder the more you rewatch it.

Similarly, Twitter user Pizza Later posted an AI-generated ad for a fictional pizza restaurant called Pepperoni Hug Spot.

The tagline, ‘Peperoni Hug Spot: like family, but with more cheese’ is quite brilliant. As are the ingredients, “cheese, pepperoni, vegetable, secret things”.

Why should marketers care?

  • Relying solely on AI-generated ad content (ChatGPT for the script, Midjourney for the images, Runway for the video clips) is clearly the future, but currently, the tech hasn't quite caught up with the hype.

  • This is especially true if your ad involves eating or drinking—AI clearly still struggles to reproduce these actions.

  • However, adopting a ‘so bad it’s good’ strategy might be a great way to drive attention.

  • So if you’re bold enough to throw your company’s creative credibility down the drain, you could well go viral. It’s like shooting the moon when playing hearts. Be so bad that you’re good. 

How I saved 4+ hours using AI to create a 20-minute event speech

A client asked me to write a 20-minute speech for an upcoming event. After deciding to use the same topic as a recent webinar they held, I set to work.

I used AI to help me write it in a fraction of the time these projects used to take.

Here was my exact process:

  1. Create a webinar transcript: I downloaded the webinar recording onto my computer before uploading it into Otter.ai. Then, it set to work automatically creating a transcript.

  1. Find the useful nuggets: I copied the transcript (400 or so words at a time) into ChatGPT, before asking:

    ‘I want you to analyse the text below and present the most useful insights back to me in bullet point format. Follow the Pareto Principle, giving me the 20% that provides 80% of the value’.


    I then copy and pasted these bullets into a Google Doc until I had a list of the hour-long webinar’s most useful insights.

  2. Divide up the insights into sections: I then wanted ChatGPT to help me work out how to present these insights in powerpoint format. I didn’t want to follow the original webinar slides as many were redundant or featured branded slides from the company’s partners.

    So, I told ChatGPT:

    ‘I’m creating a 20-minute speech entitled ‘TITLE’, aimed at helping marketers understand ‘TOPIC’. I’m going to provide you with the key points. I want you to divide these up thematically into slides so that I can create a powerpoint presentation to accompany the speech’.

    It divided the insights up into section titles, and for each section, explained which point should go in which slide.
     

  3. Flesh out the insights: The next step was to turn these bullet points into a genuine speech. I knew a 20-minute speech should be roughly 2,500 words long, so I divided this process into about 10 steps. 

    This is because I’ve found that ChatGPT can get a little lost when asked to provide more than 250 words of output at a time. It can handle it, but there’s often unnecessary repetition and it seems to run out of steam.

  4. Sense check: I pasted the script back into the main doc and reviewed it manually, clarifying anything that seemed uncertain or warranted further explanation.

  1. Double check spelling, grammar, etc.: Lastly, I pasted the entire transcript into Grammarly Premium to check for spelling and grammar mistakes.

Creative inspiration

Ethan Mollick used Midjourney to create commercial level images from text alone. Below is his "modern outfit inspired by Van Gogh”.

Food for thought: AI will soon renegotiate contracts for you

Hate negotiations? If so, we have good news: AI will soon be able to drive a hard bargain on your behalf.

Walmart recently announced that it’s been using Pactum AI to renegotiate with human sellers since 2021.

Not only has the AI come to an agreement with 68% of sellers, but it’s also driven cost-savings of around 3%.

This is a huge sum of money for a massive retailer like Walmart—and incredibly impressive given recent rises in inflation. 

The takeaway is clear: it’s only a matter of time until marketers are also using AI for contract renegotiations.

Tool to try (and tool to watch)

Bing Chat is now open to everyone who has a Microsoft account, with plug-ins coming soon. For example, it’s working with OpenTable to allow users to complete restaurant bookings just by asking the Bing chatbot.

Cloud-based content management solution Box has now integrated AI into its Box Content Cloud, teaming up with OpenAI. Companies will soon be able to feed vast quantities of their own data into the platform before automatically generating on-brand content. Watch this space.

Recommended resources📚 🎥

ChatGPT has sparked speculation about artificial general intelligence. But the next real phase of AI will be in specific domains and contexts. While it’s easy to fall into the trap of seeing OpenAI as the sole gatekeeper of this technology, this fortunately is far from the case.

The US Supreme Court’s upcoming decision on Andy Warhol Foundation for the Visual Arts, Inc. v. Goldsmith could shift the interpretation of fair use law—and all the people, and tools, that turn to it for protection.

The move comes after a sensitive data leak last month, when Samsung engineers uploaded internal source code and meeting notes to ChatGPT and accidentally leaked it. Oops.

Thanks so much for reading. With the abundance of content on the internet, your attention means a lot. If you have any suggestions or feedback on how we can improve the newsletter, please shoot us a reply. We'd love to hear from you!

Until next time. Happy marketing.

Tom and Charlie