AI Tool Comparison - ChatGPT, Copilot and Gemini

Making an old-fashioned, "low tech" sweet treat - New York Style Crumb Cake

I haven't quite figured out what kind of a relationship I want to have with AI (Artificial Intelligence). Like a freight train coming down the tracks I know that it will be upon us quickly (or already is), and much like email; the World Wide Web; or social media it's poised to disrupt life as we know it.

I expect that many of you have already had the opportunity to dabble with AI - that's if you're not already using it on a regular basis as a part of your job or your job search. From a job seeker perspective people are tripping over themselves to become experts with AI and job search. Whether it's using AI's writing skills to tailor resumes and cover letters to match job descriptions or using it to prepare marketing plans as a part of requested projects by prospective companies, there are no shortage of opportunities to delve into the AI world.

My history with AI

Last year I used (or tried to use) ChatGPT to map out a bicycle route of craft breweries in Monmouth County, NJ. I appreciated the effort that ChatGPT made, but it wasn't particularly successful (the data OpenAI pulls from is not current). I guess in the craft brewery world things change a lot so some of the information was off target (brewery location addresses, directions).

That was the summer of 2023 when it felt like AI was ready to overwhelm us, and the only game in town was ChatGPT from OpenAI. In the fast-paced world of AI technology there are additional heavyweights I wanted to focus on - Google's Gemini (formerly known as Bard) and Microsoft's Copilot. I thought this week I would compare the three in a simple way and report on the results and what I observed.

Here were my objectives

• Use AI in a practical application.
• Initiate a comparison to see what results are returned.
• Report findings and share observations.

Use AI in a practical application

Why New York Style crumb cake?

When I'm out riding my bike on the weekend there are a couple of places that are on my short list if I'm looking for a square of crumb cake (it's just a personal favorite). The Colts Neck Deli and the Allenwood General Store are my primary stops based on where I'm riding. I started wondering though if I could replicate the crumb cake they sell and produce it at home.

Why AI?

Like many of you I'm fascinated by AI and as a consumer curious about its many uses. The fact that it is readily available almost makes it a crime for us not to be figuring out ways to put it to use.

Why not a Google search if you wanted a recipe?

I didn't want just any recipe. I wanted to be specific and I wanted to see how AI would handle the request. I know Google (or any other search engine) will do the job. I wanted to narrow my choices based on the prompt and I didn't want to travel down the rabbit hole of looking at recipe after recipe. I also didn't want the beautiful images of these delicious-looking crumb cakes to sway my choice.

Words matter - It's all about the prompt

My prompt was the same for all three AI tools:

"Can you find a recipe for New York Style crumb cake that can be made in an 8" x 8" pan?"

All three provided credible recipes, yet all three differed with one main ingredient. In the Gemini recipe it was the inclusion of buttermilk. Copilot used sour cream. ChatGPT used all ingredients I had in the pantry/refrigerator.

I think we're all learning that refining our prompt (request) should help in providing what we're looking for in a more efficient way. As I spend more time using AI I'll devote more time to crafting better prompts, but for purposes of this exercise I just went with the prompt above and did not refine it.

Results and Observations - part 1 | Differences in the AI tools

All three AI tools provided credible recipes, and I believe any of them would have been good. Here are a couple of categories where I noted differences:

References

Gemini and Copilot cited their references on where the recipes came from. ChatGPT was more elusive. When I asked specifically for a reference this is what it responded with:

"I can't provide a specific citation for this recipe as it's a combination of common ingredients and techniques used in New York Style Crumb Cake recipes. It's a classic recipe that has been adapted and shared by many home cooks and baking enthusiasts over the years. However, you can find similar recipes in various cookbooks, baking websites, and food blogs that feature New York Style Crumb Cake."

Advertisements & other oddities

ChatGPT was the purest. It just delivered what I asked for.

Gemini provided two recipes (I only asked for "a recipe"). Both were almost carbon copies of the ingredients, but the quantities were slightly different.

Copilot threw in a group of images for advertisers, and that was something I would expect from a search engine, but not from my AI tool.

Results and Observations - part 2 | the crumb cake itself

After all this research about crumb cake I at least had to try out one of the recipes. Below is an image of a comparison of one of the crumb cake squares I bought in the general store with the one I made at the house. I can tell you that I got the crumb part right. Maybe not the best crumb to cake ratio, but the taste of the crumbs from the recipe was pretty good. The cake itself was a little dense. I had to use a round pan because my 8" square pan was being used for something else. That might have accounted for the crumbs not being piled as high as on the general store square.

Conclusions

This exercise that gave me a chance to use AI in a fun way, and I would encourage anyone that hasn't played around with it yet to start exploring so you can be ready for the wave of AI that will be hitting us (if it isn't already here!).

The other thing I learned is that as a crumb cake baker I'm not ready to give up my day job just yet (of course, I'd actually have to get a day job first - lol).

Happy Valentine's day and have a great rest of the week and weekend!

Dave

About David Shultis

I am a senior level B2B marketing and product management professional with experience in the life science/med device/manufacturing space. Returning to a leadership role (formal or informal) where coaching and mentoring a team or working with a cross-functional team would be an important component of the position. This could be in the form of a director of commercial excellence, marketing director, a group product director or as an individual contributor in a senior product management capacity. I'm especially interested in under-performing departments, startups or rebuilds where I can contribute by making a positive change through leadership and the implementation of procedures and metrics.

When not in pursuit of my next opportunity you can find me biking in and around New Jersey (just passed 1,000 miles this year) or trying to get my walking in with a 15k step goal each day.

Photo(s) of the week. For those of you music fans out there you may have heard of The Stone Pony, a popular concert venue in Asbury Park, NJ. Last week marked its 50th Anniversary. I was riding through the area early in the morning and took a few pictures a couple of days after it celebrated its 50th.