Meta is going all out to contest ChatGPT for mindshare in Generative AI.
In the past week, Meta has…
– Added AI chatbot in WhatsApp, Instagram and Facebook Messenger mobile apps, all powered by its new AI engine Llama 3.
– Released Llama 3 as an open-source model (which you can try out in the high-speed AI platform Groq)
– Launched the Meta.ai site which works pretty much like ChatGPT and Google Gemini. It can generate text and images at a good speed.
– Collaborated with Google to provide search engine results in the chatbots above.
I did some quick tests and here are my initial thoughts:
Meta.ai is easy to use, but I couldn’t paste more than 2200 words into it for processing. The image generation is convenient but the quality of images is so-so. I would say it’s a step behind DALL-E 3 (in ChatGPT 4) and it cannot yet understand common AI Art prompts like “pixel art”.
I am surprised that Meta.ai is integrated with Google search. I guess both companies really want to take down Microsoft and OpenAI. I managed to find an accurate description of “Ian YH Tan” using Meta.ai, so thumbs up, hah!
Llama 3 initially impresses when used on the Groq platform: it crunches text quickly and gives me good results in summarizing a speech and doing translation. However, it wasn’t good in copyediting this LinkedIn post. It showed me edited text that were the same as my original text. ChatGPT did a better job.
The tide of the Generative AI war has turned.
Rolling out Gen AI via the three of world’s most popular apps (WhatsApp, Instagram, Messenger) and integrating Google search is going to help Meta catch up and possibly overtake OpenAI in popularity.
However, for my use cases, Llama 3 is still not going to replace ChatGPT or Midjourney. But I’m sure Meta is working on this.
Fun times!
(Cat in a hat image generated by me in Meta.ai)