Could an advanced generative AI need help from another, predictive AI?
In recent years, advancements in generative artificial intelligence (generative AI) models have revolutionized the way we approach design. With the ability to generate new concepts, designs, and more for websites, packaging, app designs, ads, and more, the possibilities seem endless.
However, while the ability to generate more options may seem like a surefire way to improve performance, it’s important to remember that simply generating more does not equate to better results. In order to truly determine the effectiveness of generative AI designs, they must be tested and evaluated.
To demonstrate this, I tested 4 different cereal packaging designs created around the same color theme and used the Midjourney generative AI. I used Neurons Predict AI to run each design and evaluate the results. Neurons Predict is an online AI tool that accurately predicts consumer responses to consumer touchpoints such as ads, packaging, websites, app designs, and much more.
Testing generative AI results…with another AI tool
First, let’s have a look at the images that the AI-generated, where the prompt I suggested was as follows:
beautiful packaging for cereal, frost color, cantaloupe color, heather color, sage color,--v 4
If you’re using Midjourney, you can copy-paste this in and try some designs and tweaks yourself.
The results came in about 1 minute and were pretty much spot on for what I was looking for…

Basically, all four designs looked pretty nice, except for obvious signs that Midjourney is not a language AI model. After all, who would buy a cereal product called “TFTOEE CATTCL”?.
But let’s forget about that for now. The brand name is not critical. If anything, we could use chatGPT to help us with that 😀
How did the package designs perform on Predict AI?
Perhaps not surprisingly, we find that the packages performed differently in terms of attention, appeal, and other aspects. Let’s take an overview shot first to see the performance:

As the results show, all packages seem to generate a lot of attention to the main logo text, and less to other elements of each package. The packages also score very high on sentiment AI scores, which means that all packages are likely to be experienced as both clear and engaging.
Thus, if brand attention is your main interest, you’re all good. But if you really need more attention to other elements, such as the product view or key information, the packages differ a lot.
Looking at the Areas of Interest (AOI) scores of how attention was distributed to specific items, some interesting differences pop out, as shown below:

The results clearly show the following:
- If you want attention to your brand, design 3 is the best
- if you want more attention to the product display, design 2 and 3 are the best
- if you want more attention to the package info, design 2 is the best
This highlights the importance of testing generative AI designs to determine which ones are likely to drive the best results. At the very core, these insights do not come as automatically as the generative AI results themselves!
Some benefits of using Neurons Predict in vetting AI
These results suggest that generative AI is indeed interesting and holds great promise to further and quicken the design inspiration process. But if anything, the generative AI approach needs to be matched by an equally strong use of vetting tools such as Neurons Predict. As such, Neurons Predict AI is a powerful tool that allows businesses and organizations to test and evaluate the results of generative AI designs in order to determine their effectiveness.
One of the key benefits of using Neurons Predict AI to test generative AI designs is the speed and efficiency with which it can process and analyze large amounts of data. Instead of manually evaluating each design option, Neurons Predict AI can quickly and accurately analyze and compare multiple designs in order to determine which ones are likely to perform best. This not only saves time, but also allows businesses and organizations to make more informed decisions about which designs to pursue.
Another advantage of using Neurons Predict AI is its ability to consider a wide range of variables and factors that may impact the performance of generative AI designs. This includes everything from the target audience and industry to the overall branding and messaging of the design. By taking all of these factors into consideration, Neurons Predict AI can provide a more comprehensive and accurate assessment of the potential performance of generative AI designs.
In conclusion, while generative AI models have opened up a world of possibilities for design, it’s important to remember that generating more does not necessarily equate to better performance. By using Neurons Predict AI to test and evaluate the results of generative AI designs, businesses and organizations can make more informed decisions about which designs are likely to be effective and drive the best results.
Written in collaboration with chatGPT 😎