Few days ago Adobe released a statement that they’ve created their own image generation AI, big news considering Adobe being a monopolist in the creative services software and the new AI is a perfect match for their premier products like Photoshop, Premiere, Illustrator and surely their latest acquisition Figma.
While Microsoft (disguised as an OpenAI) and Google are fighting a grueling battle over the AI supremacy and who will win bulk of the user base and create a monopoly in what they believe is the future of the internet there is a new challenger on the horizon, and it’s the one that’s been synonymous with the art of creativity for decades.
What’s more impressive is that Adobe claims that their AI, unlike the others from this space, is not trained using copyrighted materials. Adobe said that all the training data is either out-of-copyright data or belongs to the Adobe Stock library. Sharing that information is certainly a breath of fresh air considering that other companies are usually not open in discussing which data was really used for training their models.
DALL E was released in January 2021, more than 2 years ago, and it’s seen many different use cases, either stand alone or paired with other Open AI’s products like ChatGPT, for things like generating unique children bed time stories with audio, text, images and many other applications.
While DALL E’s has an obvious advantage in form of working well with ChatGPT and other OpenAI products, Firefly might be a completely different beast since their endgame could be integrating with Photoshop and other Adobe products and not necessarily being mainly oriented to non-Adobe-owned applications.
Considering all above, DALL E and Firefly have a similar story, purpose and capabilities hence it might be fun experiment comparing what will they spit out given the same inputs. Bear in mind that this comparison is never fully fair because the results are only as good as the prompt used in generating them.
- input: coffee cup floating on a distant planet, digital art
Open AI DALL E:
Adobe Firefly:
2. input: small robot in the fantasy land looking at a billboard with text Java
Open AI DALL E:
Adobe Firefly:
3. input: animated basket floating on the moon, digital art
Open AI DALL E:
Adobe Firefly:
Objectively looking, personal artistic tastes aside, both models are very impressive, fun to play with and with plethora of potential real word use cases. Even though DALL E had two year head start, Firefly doesn’t seem to be lagging behind and takes DALL E head-on by generating great results for all prompts given to it.
While generally I quite like the Peter Thiel’s book Zero to One and his infamous point that ’Competition is for losers’ in this case of AI battles strong competitions and new big time players like Adobe entering the scene is quite intriguing to me, and apart from the downside of massive spam of countless ChatGPT this … ChatGPT that … blog posts in all feeds, subreddits, hacker news and most other places on the internet the AI software really is delivering breakthroughs and it might just deserve all the attentions it’s getting right now.
