TLDR

Text-to-image generation is one of the most exciting applications in generative AI. Its visual nature creates a dynamic appeal that goes beyond text, and it currently offers a broader range of applications than video. For creators, it hits that perfect sweet spot where mature use cases meet cutting-edge technology, which inspired this issue. Today, we'll explore the evolving landscape of image generation by crafting visuals that creatively represent Big Tech's AI strategies—as animals.

Estimated read time: 5mins

The Unfold

Generation Stage

While the purpose is not to make this a full-blown research, it remains important that we define some methodology and ground rules here.

• Inference of company’s AI strategy → animal depiction. This will be done by feeding a common prompt into ChatGPT, with the format as below:
  
  Imagine [company] AI strategy as an animal. What animal would it be?
  

• The generated animal depiction will then be used as the seed to feed into selected image generation applications, with the contestants in the table below:
  
  Create an image of [company] AI strategy as a [animal], incorporating the [company] logo in a subtle and clean way.
  

• The Big Techs are Amazon, Apple, Meta, Microsoft, Google, Nvidia, and Tesla

• We are adopting a “one-shot” approach where the first image generated by the prompt will be selected for comparison

👏 And here are the results …

Amazon

Hungry Strategist: DALL-E exceeded expectations with its ability to seamlessly integrate a company's logo into a tech-inspired design. While Ideogram and Flux didn’t quite align with my initial prompts, they brought their own unexpected creativity to the table. As for Imagen, it almost felt like it had a mind of its own.

Apple

Hungry Strategist: This is utterly surprising that the Apple brand / logo has been so well recognized and highlighted across all models (Flux as the slight deviation with its blue variation). A guess would be how vastly accumulated the Apple brand image has been in terms of perception, in contrary to the following Meta’s case.

Meta

Hungry Strategist: Meta logo / brand was not recognizable by the models across. One hypothesis is that Meta rebranded from Facebook since 2021 which is not long ago, effectively decreasing its brand accumulation across the internet. Perhaps we can alter the prompt to use “Facebook” instead, or to emphasize that Meta is in fact a company.

Pointing at a blue Gmail logo and calling it Meta, ouch! 🤯

Microsoft

Hungry Strategist: While we are half way through the list, it’s interesting to observe a pattern that DALL-E perceives the notion of “AI strategy” best in the pool with it’s futuristic style rendering. Ideogram shows an impression that closely resemble Microsoft’s brand in the early-2000. Flux seems to be confused a bit as to what it should emphasize. Imagen is making me to suspect if it has been performing mere Google searches all this time.

Google

Hungry Strategist: DALL-E and Ideogram surely scored big on this one. With Google’s full suite of products across various sectors, my solemn wish was for the model to somehow capture this essence and label each tentacle with a different Google product. While this may be too much of a stretch especially without more specific prompting instructions, both truly went over-and-beyond to answer my expectations. Imagen at least got the logo colors right. Flux is taking things the easy and lazy way.

Nvidia

Hungry Strategist: While the generations mostly covered the speed, power and streamlined structure of a cheetah, the logos in general seemed to be embedded in the most obtrusive manner, which makes one ponder if all training data related to Nvidia has been of this sort.

Tesla

Hungry Strategist: Ideogram and Flux have an edge in this round, cleverly combing the markings that would have been present if the panther was a tabby cat, together with the “T”-shaped logo of Tesla. The black and white contrast strengthens the elegant and minimalistic vibe.

It’s a Wrap!

Some thoughts I have extracted from this fun exercise:

• ImageGen comprehension is deepening, achieving desired outcomes with less efforts - While much of the focus has been on how realistic and visually appealing the outputs can be, I've found the input side to be even more intriguing—specifically, how well does the AI understand me, my needs, and abstract concepts? For instance, I didn’t just ask the models to create conceptual images of each company; I also requested that their logos be integrated subtly and artistically. In this respect, DALL-E 3 truly stood out, understanding my vision and passing my own version of a ‘Turing test’—one where I feel like a customer whose creative requests are being expertly met.

• Lots of untapped potentials in the field, it’s still anyone’s game - The global AI image generator market was valued at USD 349.6 million in 2023 and is projected to grow at a CAGR of 17.7% from 2024 to 2030. While the market is already competitive, this exercise shows that success often hinges on how closely a model’s training data aligns with the desired image or artwork niche. Some models excel in photorealistic scenes, while others specialize in graphic design. This mirrors the trends we're observing across all generative AI forms—text, video, audio, and beyond—where the real competition at the application layer is just beginning.

• There’s more to explore even with existing options - Recognizing that there is a whole world of applications I have not touched on or may not be the best tool for this activity:

  • Midjourney - Simply because I no longer have a subscription. (Calling for sponsors 😂)

  • Stable Diffusion - Tested, but it generally performs with poor understanding of company brands (with the exception of Tesla and Apple)

  • Starryai, Nightcafe - Will try out next time!

  • Microsoft Copilot Image Generator - Tested, this operates on the DALL-E model as well though I noticed that the results are more flatly designed

  • Adobe Generative Fill - Resides in Photoshop and like Stable Diffusion, it does not seem to understand company brands well

Food of Thought

Why so serious?