The perfect guacamole recipe should include ripe, mashed avocados. Spices such as fresh lime juice, coriander, red onion, salt and paprika are often added to the dish.
But guacamole’s newest ingredient, courtesy of nonprofit marketing organization Avocados From Mexico (AFM), is a dash of artificial intelligence.
With 250 million pounds of avocados set to be consumed, AFM introduced a multi-modal AI tool called GuacAImole that uses image recognition, text, and rendering. Create personalized guacamole recipes.
GuacAImole can be used in three different ways. Users can upload a picture of the ingredients they have on hand or a completed meal. They can also choose to respond to text prompts that range from serving size to flavor profile to spice level. GuacAImole then creates a recipe that includes an ingredient list, step-by-step instructions, and AI-generated visuals to show what guacamole should look like.
“We want to make products as fun and dynamic as other food categories,” says AFM President and CEO Alvaro Luque.
While GuacAImole isn’t afraid of sweets, most of the advice leans toward savory sensations. After uploading an image of Jell-O, the AI forces the user to add a quarter cup of strawberries to a fairly standard guacamole recipe, creating a slightly sweet and tangy twist on the snack.
“We wanted to simplify it for consumers,” says Luque. “We understand that artificial intelligence is the big new thing. “So how do we, as a brand, be creative enough to use technology to offer consumers a solution that works for you at the right moment, and that is the Super Bowl?”
DIP SERVICE
But even GuacAImole comes with some creative limitations. After digesting an image of a couch, some sort of error message flashes on the screen. “Hmm, looks like it. . . interesting. I’m not sure we can make it any better, but here’s a guac recipe that’s always good!” Some ingredients that do not match well with avocado will also direct users to more traditional recipes.
GuacAImole was built with a script for OpenAI’s GPT-4 with Vision, which analyzes images uploaded by users. Once the image is decoded, GPT-4 steps in to retrieve the ingredients or food and create a guacamole recipe. This output is also fed into DALL-E 3 to create concept images.
“We knew we wanted images to emerge, and so two of the three multi-modal AI tools we used are image recognition and image generation,” says Luque.
Product marketer is launching AI tool to capitalize on guacamole’s popularity during Super Bowl festivities. In the four weeks leading up to Super Bowl weekend, the amount of avocados Americans consume will cover a football field with more than 25 yards of guacamole. AFM has been a consistent Super Bowl advertiser, skipping the big game only twice since 2015, in 2021 and this year.
“Our goal is to be back 100 percent,” Luque says. “We see this as a break in our Super Bowl history.” This past football season, the AFM placed more emphasis on college football.
GuacAImole arrives a year after AFM botched its AI activation last year. The group had planned to use ChatGPT as part of AFM’s Super Bowl campaign but later canceled those plans. Luque says it was a “miscommunication” and explains that using ChatGPT is one thing, saying you’re partnering together is quite another.
This year, the playbook focuses more on the tool AFM has created for consumers. “They want to be creative and entertain their friends and family,” Luque says. “So this is a tool that can give you a lot to make game day different.”