How Meta AI compares to ChatGPT, DALL-E, and Google Gemini


Last week, Meta released the latest update to Meta AI , which further integrates chatbots into the company’s applications and adds new features such as GIF generation. Meta AI is debuting Llama 3 , the company's latest large-scale language model, and a new Meta.ai website for accessing chatbots outside of Meta's social media apps.

Llama 3 currently has up to 70 billion training parameters, and a version based on 400 billion parameters is coming soon, so it is undoubtedly a powerful tool. But for the average person, what can it do that other tools like ChatGPT , DALL-E , and Google Gemini can't? This is where you might want to use Meta AI against your competition.

Meta AI can generate GIFs

While the most popular image-generating AI to date has been limited to static frames, Meta AI can now automatically animate images for you. Technically, this is already possible in services like DALL-E, but requires the user to prompt for a series of frames and then manually stitch them together. Meta will now do the middle work for you, although relying on artificial intelligence isn't perfect, as I saw in my testing.

The process is very simple. First, ask the chatbot to generate your image.

Then, tell it to animate the image, converting it into a GIF. In my tests, telling it to just make a GIF from the start didn't work.

That's it. However, the results were not foolproof for me. The above example of the cat chasing a ball of yarn works well, but this hummingbird seems to just be hovering in space, jabbing its beak into the feeder.

Meanwhile, the world's youngest boy, 17 (my cue just says "kid"), was surprised by the magical confetti that emerged from his birthday cake. (I asked the AI ​​to create a GIF of him blowing out candles.)

Meta AI can generate images as you type

Another unique feature of Meta AI is the ability to generate images as you type. It's very simple to do so, and while the images suffer from the same dream logic as much AI art, the speed with which they pop up is impressive.

To generate an image as you type, just go to Meta.ai (the feature doesn't appear to be working on the company's social media apps yet, at least in my own testing), start with the "imagine" command, and start describing what you image of. The chatbot will preview new results for nearly every character you type, allowing you to change the prompt as you type to tailor it to your liking.

For example, when I asked the chatbot to "imagine a dog chasing a ball," it showed me a different breed of dog that nearly every new character had. When I tell it to specify "imagine a dog chasing a ball down the street," it intuits that I want the dog to chase a ball down the street. When I specify that the dog itself should be chased by a clown, simply adding the word "and" after "chased by a clown" adds a third clown to the frame.

On its own, it's a dazzling stunt. I'm used to AI art taking a few seconds to generate each individual frame, whereas this seemed to push the art out in real time instead. But what takes it beyond gimmick territory is that pressing "Enter" gives you four different results and allows you to make a video of all the other options so you can review your work and go back to the drawing board if needed.

Meta AI can browse the internet

Not every AI chatbot has real-time access to the internet, at least for free users. This is most evident on ChatGPT, where data on the free version is limited to information before January 2022. Meta AI, on the other hand, has access to Google and Bing.

This means you can use Meta AI to answer questions that require it to access current information. You can ask it, "Tell me about the weather in New York this weekend," or "Did the White House announce anything important today?" You can even ask it to compile information, saying things like, "How was the Dragon's Dogma 2 review going?" The command. or “What’s the best movie to watch this weekend?”

For each response, Meta AI tells you which search engine it used and which sources it cited.

Meta isn't the only company with this capability, as Google Gemini also provides real-time access to the Internet. Meta's chatbot is different in that it's not limited to Google searches, and it handles sources differently. Sometimes, the sources cited by Google Gemini "are not necessarily the sources that Gemini used to generate the response" - they may just be sources that Google thinks you will find useful. Meta AI, meanwhile, always seems to disclose with 100% certainty the sources it uses, but only those sources.

Meta AI can integrate with Facebook, Instagram, Messenger and WhatsApp

Whether that's a good thing or a bad thing depends on who you ask, but Meta AI is now included in Facebook, Instagram, Messenger and WhatsApp apps. Integration varies with each app, in Messenger and WhatsApp Meta AI will appear as a contact. In Instagram, it will appear in the search bar. In Facebook, it will appear in the search bar and in your feed.

In most cases, this means integration outside of the latter two applications is negligible. For search bar integration, Meta AI will now suggest searches and questions as you type them, although you can ignore these suggestions by simply not clicking on any suggestions with the Meta AI logo next to them (if you click on them, you'll be taken off your search and enter a conversation with the chatbot). Regular searches will still appear, with a magnifying glass next to them.

It's even more powerful for feed integration in Facebook. In some posts, Meta AI will now appear below the post to give you more information about the post or possibly suggest questions about the post. Your feed may also occasionally be interrupted by a generic card prompting you to use Meta AI, which functions like an ad.

So when should you use Meta AI?

With its latest update, Meta AI is several steps ahead of its most mainstream competitors, which means it may now be your best go-to for AI-generated gifs. Image generation is also powerful, now with a unique "imagine" feature, and Meta AI has one more search engine than Google Gemini.

Still, you might want to stick with ChatGPT, DALL-E, and Google Gemini so you can browse the different results, especially when it comes to the still often dreamlike nature of AI art.