AI is now in its Gemini era, according to Google
You thought streaming wars were scary? Well, it seems like we’re gearing up for a whole new level.
Google teased Gemini, their multimodal AI model, a while ago, and now they’re getting ready for launch. As Google describes it, Gemini is made for multimodality — reasoning seamlessly across text, images, video, audio, and code. It has 3 models – Ultra, Pro and Nano.
Bard, Google’s chat based AI tool, is now also powered by Gemini Pro. Here is our take:
+ it’s FREE! 🙂
+ Bard does feel better after Gemini, generally.
+ It can interact with some other services of Google! (like Drive or Maps)
+ It can interact with YouTube! This means, it can watch videos and create summaries, find videos for you etc.
+ It can finally generate images based on text prompts, like Dall.e or Midjourney, but free of cost unlike the other models! This feature is still rolling out to users, so watch this space as Google will also release other upgrades soon.
+ You can ask Bard to double-check responses with the press of a button, which comes handy when you have some doubts over some of its replies.
– Overall, GPT-4 can understand your needs a bit better, research further and give you more in-depth answers.
– Midjourney version 6 is still the best image generation tool from what we’ve experimented with. And soon it will be even more easy to use through their website instead of Discord!
One Prompt, Three Models – How do image generation tools compare?
To test their capabilities, we’ve prompted the three models with the same ask – Adam and Eve meeting a tiny robot in the garden of Eden.
This is what they came up with…
😲 Imagen didn’t quite understand what we meant by Adam&Eve, so we modified the prompt a little bit.
Midjourney v6
Very artistic, different visual styles, but not exactly what we prompted. But happy to see Midjourney is an ally! 🏳️🌈
DALL.E 3 by Open AI
More accurate for our prompt, though not as visually tasty as Midjourney.
BARD/IMAGEN 2 by Google
Imagen 2 is still slowly rolling out on Bard, and we had to modify our prompt. The visual style is different and interesting, but you can notice some weird limbs and a very odd robot. Overall, it seems great for a free model.
Diversity – Worth a mention…
When we prompted it with “a man and a woman from the beginning of human civilisation”, it responded with black people as well as people who are (seemingly) from central asian cultures. These results are great and promising as we know diversity is a massive topic of conversation in AI generated content.
We hope you enjoyed our takes on these tools. Do let us know what you think and/or if you’d like any help with generating images or content with AI!