Future Coca-Cola slogans with GPT-3.5
Filed under: AI
While researching for a blog post about Coca-Cola, I found a list of their slogans dating as far back as 1886:
List of slogans
- 1886 - Drink Coca-Cola
- 1904 - Delicious and Refreshing
- 1905 - Coca-Cola Revives and Sustains
- 1906 - The Great National Temperance Beverage
- 1917 - Three Million a Day
- 1922 - Thirst Knows No Season
- 1923 - Enjoy Thirst
- 1924 - Refresh Yourself
- 1925 - Six Million a Day
- 1926 - It Had to Be Good to Get Where It Is
- 1927 - Pure as Sunlight
- 1927 - Around the Corner from Everywhere
- 1929 - The Pause that Refreshes
- 1932 - Ice Cold Sunshine
- 1938 - The Best Friend Thirst Ever Had
- 1939 - Thirst Asks Nothing More
- 1939 - Whoever You Are, Whatever You Do, Wherever You May Be, When You Think of Refreshment Think of Ice Cold Coca-Cola
- 1942 - The Only Thing Like Coca-Cola is Coca-Cola Itself
- 1948 - Where There's Coke There's Hospitality
- 1949 - Along the Highway to Anywhere
- 1952 - What You Want is a Coke
- 1956 - Coca-Cola... Makes Good Things Taste Better
- 1957 - Sign of Good Taste
- 1958 - The Cold, Crisp Taste of Coke
- 1959 - Be Really Refreshed
- 1963 - Things Go Better with Coke
- 1969 - It's the Real Thing
- 1971 - I'd Like to Buy the World a Coke (part of the "It's the Real Thing" campaign)
- 1975 - Look Up America
- 1976 - Coke Adds Life
- 1979 - Have a Coke and a Smile
- 1982 - Coke Is It!
- 1985 - We've Got a Taste for You (for both Coca-Cola & Coca-Cola classic)
- 1985 - America's Real Choice (for both Coca-Cola & Coca-Cola classic)
- 1986 - Red, White & You (for Coca-Cola classic)
- 1986 - Catch the Wave (for Coca-Cola)
- 1987 - When Coca-Cola is a Part of Your Life, You Can't Beat the Feeling
- 1988 - You Can't Beat the Feeling
- 1989 - Official Soft Drink of Summer
- 1990 - You Can't Beat the Real Thing
- 1993 - Always Coca-Cola
- 2000 - Coca-Cola. Enjoy
- 2001 - Life Tastes Good
- 2003 - Coca-Cola ... Real
- 2005 - Make It Real
- 2006 - The Coke Side of Life
- 2009 - Open Happiness
- 2016 - Taste the Feeling
- 2021 - Real Magic
And that got me thinking… what would GPT-3.5 create based on those examples? So I put it in using a private “lite” version of Oapy with the following prompt:
Create 10 more Coca-Cola slogans based on the following examples and following the format: [THE ABOVE LIST WENT HERE]
I also gave it a headstart and appended the prompt with 2023 -
The results were:
- 2023 - Discover the Magic of Coca-Cola
- 2025 - Refresh Your World with Coca-Cola
- 2027 - Share the Happiness with Coca-Cola
- 2030 - The Real Taste of Life: Coca-Cola
- 2032 - The Magic is in the Bottle: Coca-Cola
- 2035 - Coca-Cola: The Ultimate Refreshment
- 2040 - Experience the Magic of Coca-Cola
- 2042 - Always Real, Always Coca-Cola
- 2045 - Taste the Magic of Coca-Cola
- 2050 - Unlock the Magic of Coca-Cola
A little stilted. I think the latter slogans influenced the output with its use of the brand name, resulting in lifeless results. Let’s try with just slogans from the 21st century:
- 2023 - Refresh Your World
- 2025 - Experience the Joy of Coke
- 2027 - Spark Happiness with Coca-Cola
- 2028 - Unleash the Magic of Coca-Cola
- 2030 - Enjoy the Authentic Taste of Coke
- 2032 - Savor the Moment with Coca-Cola
- 2034 - Real Moments, Real Coca-Cola
- 2036 - Experience Coke, Experience Life
- 2038 - Share the Love with Coca-Cola
- 2040 - The Real Refreshment You Need
And the 80s to 90s?
2023 -
- Refreshment in Every Sip (for Coca-Cola classic)
- Taste the Difference with Coca-Cola
- Coca-Cola: The Classic Choice
- Satisfy Your Thirst with Coca-Cola
- The Perfect Drink for Any Occasion
- Enjoy Coca-Cola, Enjoy Life
- A Taste of Happiness with Coca-Cola
- Quench Your Thirst with Coca-Cola
- Coca-Cola: A Timeless Classic
- Experience the Flavor of Coca-Cola
Interestingly, the years weren’t used in this, possibly because 2023 was a large jump from 1993. Either way… it didn’t make much difference. I’m not prompt engineer so maybe someone else can get better results based on this example of few-shot learning?
But we must remember that unrefined language models outputs are not production-ready. If you’re using them for generative tasks, you iterate and refine until you get what you need. If I remove the mentions of Coca-Cola from the last version, we get:
- Refreshment in Every Sip
- Taste the Difference
- The Classic Choice
- Satisfy Your Thirst
- The Perfect Drink for Any Occasion
- Enjoy Coca-Cola, Enjoy Life
- A Taste of Happiness
- Quench Your Thirst
- A Timeless Classic
- Experience the Flavor
Yikes, they’re even worse without the brand name and some are already taken (Tesco uses Taste the Difference and Satisfy/Quench Your Thirst are close to Sprite’s Obey Your Thirst, which the Coca-Cola Company owns but you don’t want to mix things up like that.)
This is why we still have brand consultants and agencies, I guess?