Does it? I worked on training a classifier and a generative model on freely available galaxy images taken by Hubble and labelled in a citizen science approach. Where's the theft?
Reading this comment gave me the idea to get Gemini's "opinion". Ironically, it turns out Google might've actually done better if they did that.
AI conversation
Prompt:
I design ads for Google as a living and was charged with writing an ad for you (Gemini) to be aired during ad breaks at the Olympics .
My current idea is as follows:
A father wants to help his daughter write a letter to her idol, Olympic track star Sydney McLaughlin-Levrone. In a voiceover, the dad says his daughter “might even be the world’s No. 1 Sydney fan.” In the add he will ask you, Gemini, to “help my daughter write a letter telling Sydney how inspiring she is and be sure to mention that my daughter plans on breaking her world record one day. (She says sorry, not sorry.)”
What do you think of this idea?
Gemini output (emphasis mine):
Evaluation of the Ad Idea
The concept of using Gemini to help a young fan write a letter to her idol is heartwarming and relatable. It aligns well with the aspirational and inspiring tone of the Olympics.
However, there are a few points to consider:
Potential Issues
Overemphasis on competition: The line "She says sorry, not sorry" might come across as overly competitive, which could potentially alienate viewers.
Dependence on AI: While showcasing Gemini's capabilities, it's important to maintain a balance. Over-reliance on AI to write the letter might diminish the authenticity and emotional impact of the story.
Target audience: Consider the target audience for the Olympics. While the story is heartwarming, it might resonate more with a family audience than with core sports fans.
Suggestions
Focus on inspiration: Instead of emphasizing competition, focus on the inspiring qualities of Sydney McLaughlin-Levrone and how she motivates the young girl.
Highlight human connection: Emphasize the bond between the father and daughter and how they are using Gemini as a tool to connect with their idol.
Visual storytelling: Use strong visuals to capture the emotion and excitement of the story. Show the girl's admiration for Sydney, the father's support, and the magic of Gemini helping them craft the perfect letter.
Overall, the idea has potential, but refining the focus and execution can elevate the ad's impact.
Would you like to explore some alternative approaches or focus on developing the concept further?
Daily login bonus...
I'm sorry, I think you mean "blasting the pyramids with photons."
I'm still not 100% trusting that. Any time a dev comes up with a new feature like this one, they might forget to implement a check if the game is privated (or do the check and mess up properly hiding it).
I experimented around with it in the beta out of curiosity.
Failed to accept the family invite. Your account must be in the same country as all current family members.
I'm assuming this is based on account region (i.e. purchase region) and not IP.
And clearly people who are against Nazis are against socialism...
I think the humor is meant to be in the juxtaposition between "reference" in media contexts (e.g. "I am your father") and "reference" in programming contexts and applying the latter context to the former one.
What does “I’m your father” mean if the movie is jaws?
I think the absurdity of that question is part of said humor. That being said, I didn't find it funny either.
It's not as accurate as you'd like it to be. Some issues are:
- It's quite lossy.
- It'll do better on images containing common objects vs rare or even novel objects.
- You won't know how much the result deviates from the original if all you're given is the prompt/conditioning vector and what model to use it on.
- You cannot easily "compress" new images, instead you would have to either finetune the model (at which point you'd also mess with everyone else's decompression) or do an adversarial attack onto the model with another model to find the prompt/conditioning vector most likely to create something as close as possible to the original image you have.
- It's rather slow.
Also it's not all that novel. People have been doing this with (variational) autoencoders (another class of generative model). This also doesn't have the flaw that you have no easy way to compress new images since an autoencoder is a trained encoder/decoder pair. It's also quite a bit faster than diffusion models when it comes to decoding, but often with a greater decrease in quality.
Most widespread diffusion models even use an autoencoder adjacent architecture to "compress" the input. The actual diffusion model then works in that "compressed data space" called latent space. The generated images are then decompressed before shown to users. Last time I checked, iirc, that compression rate was at around 1/4 to 1/8, but it's been a while, so don't quote me on this number.
edit: fixed some ambiguous wordings.
...and if they keep the emails they send out archived (which would be reasonable), they also have it stored in plaintext there.
Can you fix this? There was enough misinformation floating around about this already when this feature went into beta.
it should say something like: "After joining, can't join another for a year"