Can AI Write a Better Mother’s Day Card? How my mom reacted, and… – GeekWire

The AI-generated card I got my mom for Mother’s Day, with a font similar to my handwriting.

There’s nothing like a heartfelt Mother’s Day message to let your mom know how much you appreciate her and everything she does. As I sat down to write mine this weekend, some natural questions popped into my mind: Would this be a good job for generative AI? If so, how would my mother feel about it? And how would I like it?

Several factors worked in my favor: My mom has put up with my “experiments” all my life, and she just finished listening to our GeekWire Podcast episode where our voice clones read an AI-generated script.

So she had a lot of context and I knew she’d be a good sport once I told her what I’d done, even if an AI-generated message might not have the same impact as a personal note.

This is the prompt I gave OpenAI’s ChatGPT, Microsoft Bing and Google Bard.

Write a heartfelt Mother’s Day message to my mom and tell her how much I love her and am thankful for everything she does for me, from me, her son, Todd. Include examples such as her going to my tennis tournaments when I was a kid, and mention how proud I am of how she runs and enjoys the family business, and how proud Dad would be of her doing that too. Name her own dedication to her “Morning Messages” ministry, and how many lives she touches in the process.

Here’s a PDF if you want to read each of the responses. I immediately ignored Bing because of the opening line, which read like a student mimicking the question in a misguided attempt to pass the test: “I just wanted to write you a heartfelt message to tell you how much I love you and am thankful for everything you do it for me.”

Overall, ChatGPT (with GPT-4) was much more natural in its language. I loved the simple, short sentences. “I remember when I was a kid you were always there at my tennis tournaments. Rain or shine, win or lose, you had my back. I didn’t say it at the time, but it meant a lot to me.”

However, something about the first Bard version caught my attention and called my mom my “biggest fan.” This is what my dad would quietly tell me when my mom and I had a fight. (“You know, Todd, your mom is your biggest fan.”)

Yes, I could have mixed and matched and edited, but that would have complicated things and diluted the pseudoscientific nature of the experiment. Based on that detail, I literally went with Bard:

Dear Mom,

Happy mothersday! I hope you have a great day full of love, laughter and joy.

I am so thankful to be your son. You’ve always been there for me no matter what. You have supported me through thick and thin and I know I can always count on you.

I remember when I was a kid you used to go to my tennis tournaments. Even when it rained or frosted, you were there to cheer me on. You were my biggest fan and I knew I could always count on you to be there for me.

I am so proud of the woman you have become. You are a successful business woman and you are also a wonderful mother and grandmother. You are always so positive and cheerful, and you always know how to make me laugh.

I know Dad would be so proud of you running the family business. You did a fantastic job and you make him proud every day.

I am also so proud of your dedication to your Morning Messages. You touch so many lives with your words of encouragement and hope. You are an inspiration to me, and I know you are an inspiration to so many others.

I love you more than words can express. Thank you for being the best mother a son could wish for.


I found a simple Mother’s Day card template in the online design program Canva and used a font similar to my own handwriting to further develop the illusion (or trickery, depending on your perspective).

After printing it on cardstock on our home printer, I put it in an envelope to give to my mom on the drive to my sister’s house for Mother’s Day dinner. I also folded up a printout of my AI question and all the answers and put it in the envelope so we could talk about the process and her thoughts after I was clean.

As she was reading I watched her to see how she reacted, and parts of the note clearly resonated with her, but I could see that other parts (i.e. always knowing how to make me laugh) were not her really seemed. After she thanked me for the thoughtful card, I let her look at the printout and let her know what I had done.

Essentially, she said she was fine with it, especially after seeing the instructions I gave the AI ​​models and seeing that I chose from several posts to get one that I thought was the best.

An important piece of background is that my mother is someone who goes to the drugstore and very carefully selects cards with the right message for a friend or family member, then adds her own personal note.

She saw the process I went through with the AI ​​models as a step further.

As she could see on the printout, I not only gave the AI ​​directions on what to write, but also selected the message that resonated most from the various AI models.

“You chose what was on it,” she said. “That actually makes more sense than a store-bought card.”

She astutely pointed out that the approach reflects the natural evolution of the proposed auto-replies and auto-completion features common in text and email programs today.

Her biggest complaint: I printed the card on a full sheet of paper instead of making it foldable, which means she can’t easily put it upright with her other cards on the kitchen table. Noted for next time.

How did that make me feel? I actually felt pretty good until I shared the process and result with some of my colleagues during the writing process. One of them said the language seemed insincere, which I can understand and appreciate after reading it again.

There were definitely phrases and sentences in the post that I wouldn’t naturally use (starting with the opening line where I wished her “love, laughter, and joy”).

In the end, the words weren’t mine, even though the ideas and direction were.

After conducting my experiment, I learned that Greg Gottesman, co-founder and general manager of Pioneer Square Labs in Seattle, tried a similar experiment last year with his wife, children, and parents, sending them messages of love and appreciation with an earlier version of ChatGPT, as a test.

His wife, Shannon, knew right away what was going on.

“I love you too, ChatGPT,” she replied.

If I were to go this route for a personal note or card in the future, I’d probably use the AI ​​models more for inspiration and boost. I’d at least edit the results instead of taking any of the AI-generated posts verbatim.

Obviously a handwritten note, from the heart and in my own words, would have been better. But not everyone is capable of that. And sometimes it’s nice to have a little help, even if the helper isn’t human.

Have you tried using AI to write personal notes? How were they received? What did you learn? Email for possible inclusion in a follow-up article.

Adblock test (Why?)

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *