Snowflake Arctic AI, Microsoft's Tiny AI Model, GPT beats Ophthalmologists, $6B for xAI
PLUS HOT AI Tools & Tutorials
Welcome to our weekly news roundup!
I'm glad to announce that our newsletter has grown to over 5,000 subscribers 🍾 Thank you for your support. We promise to keep bringing you the most valuable AI stories!
To celebrate 5000 subscribers, we invite you to participate Subscription Giveaway and make our newsletter even better. Take a short survey and tell us about your experience + participate in the free Giveaway of a Premium Subscription for 6 months.
And by the way, this week we had posts about LLMs and how they really help to save time. The topic was so broad that we split it into two posts. Here they are:
This Creators’ AI Edition:
Featured Materials 🎟️
News of the week 🌍
Useful tools ⚒️
Weekly Guides 📕
AI Meme of the Week 🤡
AI Tweet of the Week 🐦
(Bonus) Materials 🎁
Featured Material 🎟️
Snowflake Releases Open-Source Arctic AI Model
An unexpected turn of events (or is it just me being inattentive?). We have a really serious player in the AI market here. Snowflake Inc. has unveiled its own big open-source language model that will compete with solutions by OpenAI, Google, Meta, Mistral, and many others. It's called Arctic.
The company says Snowflake Arctic is today's most open-source enterprise-level LLM program. Moreover, it's also one of the most powerful, aided by a “mixed experts” architecture that optimizes the model to run a wide range of use cases. Here's what Snowflake CEO Sridhar Ramaswamy said about the new model:
I think this is going to be the foundation that’s going to let us — Snowflake — and our customers build enterprise-grade products and actually begin to realize the promise and value of AI. You should think of this very much as our first, but big, step in the world of generative AI, with lots more to come.
This model utilizes a unique Dense-MoE hybrid transformer architecture to handle SQL generation, encoding, and instruction execution efficiently. The company says that with its ability to operate on low budgets, Arctic offers a high-performance solution for enterprises looking to integrate advanced AI without significant cost.
The model now includes two versions: Arctic Base and Arctic Instruct.
The first, Arctic Base, offers a framework for general AI tasks and will suit most Snowflake enterprise customers. Arctic Instruct, on the other hand, was designed for narrower tasks. This platform offers customized performance improvements such as accurate command response and advanced query processing. Snowflake says both versions are designed to be highly flexible and scalable to meet various business needs.
Keep your mailbox updated with key knowledge & news from the AI industry
Snowflake Arctic is open-sourced under the Apache 2.0 license to increase transparency by providing unauthorized access to both the model weights and the underlying code. The open-source code includes detailed documentation and data recipes, which will also help users customize the model to their needs.
And one more important detail: Snowflake claims that this model was built from scratch in less than three months and at about one-eighth the cost of training similar LLM programs. By establishing a new framework for LLM training and training costs, the company says, enterprises will be able to build significantly cheaper generative AI models on a large scale.
Sharing is caring! Refer someone who started a learning Journey in AI!
News Of The Week 🌍
Microsoft Launches Tiny AI Model Phi-3
Microsoft announced its smallest AI model yet, Phi-3. This model, measuring just 3.8 billion parameters, was learned from ‘bedtime stories’ created by other LLMs. Thanks to innovations in learning, the company says this family outperforms the same and next-size models on a range of tests assessing language, programming, and math abilities.
The new model is available in the Microsoft Azure AI Model Catalog and on Hugging Face, as well as Ollama, a lightweight framework for running models on a local machine. Microsoft says it will also be available as an NVIDIA NIM microservice with a standard API interface that can be deployed anywhere.
Nvidia Buys Two AI Startups for Chip Production
What could be better than buying an AI startup? The rule of thumb is acquiring two startups at once. That's what Nvidia has done. According to The Information, the company has acquired two Israeli developers, Run:ai and Deci. The first deal has already been confirmed, while only insiders discuss the second.
The purpose of the acquisitions is simple and obvious—Nvidia wants to reduce the cost of building AI models.
And here's an interesting fact. Right now, Nvidia is the most active venture capitalist in AI. Last year alone, the company invested in more than 30 startups! Apparently, Nvidia is aiming for the title of the most active buyer of AI startups.
Adobe Firefly Tools are Now Available in Photoshop Desktop Beta
Adobe has released an updated version of the Firefly model and added it to the beta version of Photoshop. The company called (well, certainly) this tool “the best model for generating images.” The AI has learned to create realistic images and understand more complex cues and prompts.
Adobe says it should render things like typography, iconography, raster images, and line art more accurately and is “significantly” more adept at depicting dense crowds and people with “detailed features” and “a variety of moods and expressions.”
xAI Closes In on $6 Billion Fundraising
No week is complete until we get a big news story about Elon Musk. This time, our attention was caught by an insider from Bloomberg, who said that the startup xAI is close to raising $6 billion in investments and getting a valuation of $18 billion. The specific list of funds is unknown, but it is believed that one of the main investors will be Sequoia Capital.
Well, we can't yet say if Grok is capable of beating ChatGPT, but Musk clearly has enough finances to call xAI a competitor to OpenAI.
GPT-4 Outperforms Doctors in Diagnosing Eye Problems
OpenAI's GPT-4 model beat doctors at assessing eye problems in a University of Cambridge study. The researchers showed how the GPT-4 model performed in 87 different scenarios. According to the results, the AI failed to beat expert-level ophthalmologists and made several errors in making diagnoses. However, GPT-4 fared better than trainees and junior specialists. The scientists noted that AI may catch up with top doctors in the field of ophthalmology soon.
Well, it sounds like we may soon be able to save quite a bit on insurance!
Apple Releases Open Source AI models Designed to Run On-Device
Apple has released a series of OpenELM language models that run directly on user devices such as smartphones and laptops. These small models are different from most existing solutions, which require a connection to the cloud and are designed to generate text. With them, the device's power consumption is almost zero, and they don't need an internet connection.
OpenELM consists of 8 models, each customized and pre-trained for optimal performance in different environments. These models range in size from 270 million to 3 billion parameters, allowing them to handle text-generation tasks efficiently.
Somebody, but Apple knows how to raise expectations. This is not the first (or even the tenth) news about how the company's engineers are developing AI, but we still haven't heard an official announcement. Should we waiting for WWDC 2024?
Creators’ AI Subscription is the best gift for Creators diving into AI.
Useful Tools ⚒️
Assista AI – A central nervous system for all your productivity apps
Insights by Ayraa – A new dimension in enterprise search AI
Dart – The ultimate AI project management tool
PaddleBoat – Perfect your sales pitch with realistic AI roleplays
CopyCopter – Create viral videos for TikTok and Shorts from text with AI. It’s a tool designed to generate short social media videos from text content automatically. No video editing skills are required. CopyCopter uses LLMs to automatically edit your video, making it visually appealing and potentially increasing its virality.
Weekly Guides 📕
How to Use Adobe Firefly 3 (& Why It's the Only AI Image Generator You Should Use)
How to Use & Earn from Meta Ai Whatsapp | Meta Ai
Llama 3 8B: BIG Step for Local AI Agents! - Full Tutorial (Build Your Own Tools)
Your guide to using artificial intelligence for business growth
AI Meme Of The Week 🤡
AI Tweet Of The Week
Jensen Huang personally brought the new AI training CPU to the OpenAI office!
(Bonus) Material 🏆
Perplexity AI is raising a lot more money for a $2.5-3 billion valuation
Meta’s A.I. Assistant Is Fun to Use, but It Can’t Be Trusted
Here Come the Anti-Woke AIs - by WSJ
2024 AI Index Report - Stanford University
Ads for Explicit ‘AI Girlfriends’ Are Swarming Facebook and Instagram
Take our survey to participate in the Premium Subscription Giveaway!