2 months ago
15 ARTICLE AD
Whether you like it or not, the world has decided to embrace artificial intelligence fully. That means there’s AI infused into everything, even your favorite messaging app. One of those apps is likely one of Meta’s properties: WhatsApp, Facebook, or Instagram. Soon, you can interface with one of them to generate an image.
I got to try this out at Meta Connect, hosting demonstrations on Samsung Galaxy S22s. The demos also showed off Llama 3.2’s image generation capabilities. These new models are Meta’s first major vision models to understand images and text within a dialogue. So far, it’s nothing I haven’t already done with Google’s AI. That is to say, I ran into most of the same blockages I do with Gemini, wherein I realized that I still don’t understand this whole querying thing. While I didn’t mind the final product (the header image above), it wasn’t what I envisioned generating.
My original idea was to have a photo of a family dressed as characters from Sonic the Hedgehog flying through space. I started by asking Meta AI, inside the Messenger app, to “imagine a family of three as varying characters from Sonic the Hedgehog.” (I’m prepping for Halloween.) It depicted a stereotypically nuclear family: a mom, a dad, and a young child. Everyone in the photos had a Sonic costume on. The “mom” in the photo was Knuckles, the dad was Sonic, and the kid was Tails. The kid was also a boy, so I asked Meta AI to change “the gender of the child in the middle.” It produced a photo with another boy, changing the scope of the entire image completely.
I got annoyed at this point in the demonstration, so I decided to try a different approach. I asked Meta to change the background of the photo it had produced to something that looked like a “space theme.” It worked as directed, so I asked Meta to “change the post to flying.” That’s how I ended up with the header image above. It somehow evolved from a family to a man dressed as Sonic and three other children.
Ultimately, I didn’t get what I envisioned. Perhaps I should have been as detailed as possible to have that image generated a certain way. But can you blame me? This AI stuff is as new to me as it is to the rest of the world. I expect to see more of this kind of imagery sprouting up in my Instagram feeds as this feature rolls out.
