Meta everywhere: Llama 3, Meta AI & Horizon OS
Meta everywhere: Llama 3, Meta AI & Horizon OS
10 places you can use Llama 3 for free
Total reading time around 5 minutes.
Welcome to Visually AI!
🔮AI News This Week
Meta News Everywhere
I feel like I’ve been seeing Meta news everywhere in my generative AI corner of the world, over the past week.
I tried to simplify it and provide some helpful links for you to check out the things you care about.
Llama 3
Meta open-sourced Llama 3 a few days ago, and it’s pretty impressive, and everywhere.
I won’t go into the performance against other LLMs, but I’ve used it several times on Meta AI in my browser, Groq, and Poe. It gave me quality answers without the typically obvious AI words, such as “delve”, “unleashed,” etc.
You might have noticed that there are two versions of Llama 3: Llama 3 8B and Llama 3 70B. So, what's the difference between them?
Think of it like this: the number (8B or 70B) refers to the model's "brain size".
A larger brain size means the model has been trained on more data and can understand more complex tasks.
Llama 3 8B:
Smaller brain size
Faster and more efficient for quick tasks and conversations
Good for simple questions, generating text, and basic tasks
Llama 3 70B:
Larger brain size
Slower but more accurate for complex tasks, in-depth conversations, and nuanced understanding
Good for advanced tasks, long-form writing, and tasks that require more thought and analysis
According to
, you can run it locally with no internet access on GPT4All, LM Studio, and Ollama.Online, you can use Llama 3 on these sites:
Meta AI Web: Access Llama 3 directly on Meta's website or through Meta apps like Facebook and Instagram, available in supported locations. (link)
Perplexity Labs: Experiment with Llama 3 models, including llama-3-70b-instruct and llama-3-8b-instruct, with generous token limits for thorough testing. (link)
HuggingChat: Log in or create a free account, select "Meta-Llama-3-70B-Instruct" as your current model, and start chatting with Llama 3. (link)
Vercel Chat: Test both Llama 3 models for free, and compare their responses and token usage side-by-side to see their strengths and cost efficiency. (link)
Replicate: Try out Llama 3 models in the Replicate Playground, accessing the 8B model in meta-llama-3–70b and the 70B model in meta-llama-3–8b. (link)
Groq: Experiment with Llama 3 models, including Llama 3 8b-8192 and Llama 3-70b-8192, on the Groq platform. (link)
Poe: Access Llama 3 through the Poe platform, a web-based interface that allows you to experiment with various AI models, including Meta-Llama-3-70B-Instruct. (link)
Meta AI
Meta AI, powered by Meta Llama 3, is a new AI assistant that promises to help you get things done, learn, create, and connect with what matters to you.
It's available in English in these countries: Australia, Canada, Ghana, Jamaica, Malawi, New Zealand, Nigeria, Pakistan, Singapore, South Africa, Uganda, Zambia and Zimbabwe
Meta says it can help you plan events, answer questions, and even generate images based on text input.
Key features include:
Seamless search integration across Facebook, Instagram, WhatsApp, and Messenger
Text-to-image generation using the Imagine feature, now faster and higher quality
Integration with Ray-Ban Meta smart glasses and Meta Quest
I have not tried the image generator yet, because you have to log in with your Facebook account and I just don’t feel like doing that.
But I have seen several images other people generated and the quality is impressive.
You don’t have to log in with a Facebook account to use Meta AI on your browser, but you won’t be able to generate images or save your chats. You can try it here.
Meta Horizon OS
And there’s more Meta news that dropped today, as I wrote this newsletter:
Meta is opening up its Meta Horizon OS operating system to third-party hardware makers, allowing for a more open and diverse ecosystem of mixed reality devices.
This means more choices for consumers and businesses, and more opportunities for developers to build mixed reality apps.
Here are the key takeaways:
The Meta Horizon OS combines advanced technologies like inside-out-tracking, hand tracking, and spatial anchors to enable immersive mixed reality experiences.
Developers can use the Meta Horizon OS to create social, interactive experiences that integrate with users' identities, avatars, and social graphs.
The Meta Horizon Store will feature apps from multiple developers, making it easier for users to discover new experiences.
📰 Other AI News this Week
I’ll make this much shorter and digestible, after the lengthy Meta news!
VASA-1: VASA-1 is a new AI model developed by Microsoft Research that generates lifelike talking faces from audio in real-time. It is not available to the public, and there are no plans to release it, at this time. (link)
VideoGigaGAN: Adobe Research published a report about VideoGigaGAN, a new generative video super-resolution (VSR) model that can upsample videos up to 8× while preserving rich details and maintaining temporal consistency. (link)
AI Music: There has been a lot of news about the realistic music and voices generated by Udio and Suno, and now we have popular musicians like Drake releasing diss tracks using AI-generated voices of Tupac and Snoop Dog. (link)
Poe: Poe now has multi-bot chat so you can chat with other bots by tagging with an
”@-mention” or choose “Compare” for one of the suggested bots. I tested it with an uploaded image and compared the result to Gemini Pro, Claude 3 Haiku, and Mixtral8x7B Chat, below. (link)
You could have your AI service, tool, or event seen by Visually AI’s community of over 9,400 subscribers:
Sponsored by Blaze AI
Let us be blunt, we've never done a sale this high, and it's only for 5 days, so don't miss out on this chance to light up your business with Blaze:
Blaze 4/20 Sale from April 20-25, with 42.0% off our annual plans.
Produce blog posts, social media content, and ad copy - all in your brand voice, at Blaze.ai.
🚀 This Week’s AI Tools
Digest: Build your personalized daily newsletter with content from your favorite apps. (link)
Collato: Transforms images, transcripts, and meeting notes into customizable documents. (link)
Lazy AI: Build and deploy full-stack web apps with text prompts. (link)
CV & Resume ATS Optimize + Match-JOB GPT: ChatGPT Pro users can create resumes and CV’s optimized for ATS and tailored to job descriptions with interview tips. Note: I used this GPT to create a fantastic resume for my 18-year old son, with limited work experience. (link)
Stylar Image Editor: AI-powered image editor with generative fill for object removal and replacement, background generation or replacement, image-to-image style transfer, and more. (link)
🖼️ Image Prompts
Prompt: Desert nomad facing a sandstorm maelstrom, sun-bleached bones, sands of time, oasis mirage, ancient scars, stark survival essence
Prompt: Luminous canyon, vivid sunset, river of stars, surreal, cascading waterfalls, bioluminescent flora, hypnotic depth, ethereal, floating rocks, cosmic backdrop
🎁 Get it free: The AI Visual Creator’s Toolkit
Boost your content with my all-in-one, free visual AI toolkit!
Access AI-powered tools for AI-generated images, image editing, and more: Get your toolkit
Thanks for reading, and have a creative week!