Alibaba's Model Outperforms LLama 3.1, YouTube AI Tools, and Amazon AI for Sellers
PLUS HOT AI Tools & Tutorials
Hi!
Today, we had an unusual guest in the weekly roundup, which led me to believe we don't follow enough non-US companies in the AI industry. We often discuss OpenAI, Google, Meta, and Anthropic. But how frequently do they release families of 100 LLMs at once? That's what I'm saying.
So, let's not hesitate to review this week's most exciting AI news, guides, and tools.
Don't miss our review of the new OpenAI o1 model:
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 🎟️
Alibaba Unveils Family of 100 LLMs Called Qwen 2.5
Chinese company Alibaba has introduced the Qwen-2.5 family. This collection of advanced LLMs includes 100 models designed to solve a wide range of problems, including math, coding, and general-purpose tasks. While the sheer number of AIs released alone makes this a significant event, the main reason we're talking about Qwen 2.5 is performance.
Qwen 2.5 models range in size from 0.5 to 72B parameters and support more than 29 languages. The Qwen-2.5 series has been trained on a dataset of 18 trillion tokens, allowing the models to support up to 228 thousand tokens. The company says this will enable them to be used in a wide range of AI applications on the edge and in the cloud in sectors ranging from automotive to gaming to scientific research.
In particular, Qwen 2.5 outperforms even LLama 3.1 405B in some tests. And the integration of reasoning techniques such as Chain of Thought, Program of Thought, and Tool Integrated Reasoning further enhances the performance of these models. As a result, the Qwen-2.5 series not only outperforms Meta's Llama 3.1 405B and 70B models, but also competes with GPT-4.
According to Geeky Gadgets, Qwen-2.5 offers three categories of models, each designed for specific tasks:
Basic models: ranging from 5.5B to 72B parameters, provide general-purpose capabilities, making them versatile tools for a wide range of applications.
Coder models: available in 1.5B, 7B, and 32B parameter options, these models are carefully optimized for coding tasks, allowing developers to streamline their workflows and increase productivity.
Math models: Reflecting the dimensions of coder models, these specialized models are fine-tuned for mathematical problems, offering powerful tools for researchers, analysts, and educators in quantitative fields.
Keep your mailbox updated with key knowledge & news from the AI industry
The developer focuses on accessibility, so most Qwen-2.5 models are released under the Apache 2.0 license. This means enthusiasts can flexibly integrate these tools into their applications without significant legal restrictions. However, there are some limitations: 3B and 72B models are unavailable under this license. Qwen-2.5 models are available on Hugging Face and for local deployment using LM Studio.
While OpenAI, Google, Meta, and Anthropic are on the radar, Qwen has also become popular among developers worldwide. The number of Alibaba model downloads has surpassed 40 million Hugging Face and ModelScope downloads, and Qwen has been used to create more than 50,000 products.
So, the release of Qwen 2.5 family might be a good reminder that the battle for leadership in the AI industry is only gaining momentum.
News Of The Week 🌍
Amazon Debuts Its AI Assistant for Sellers Called Project Amelia
Amazon sellers now have access to an artificial intelligence-based assistant to help them grow their business. The new platform answers questions about metrics, advises sellers, and supports problem-solving. The assistant, codenamed Project Amelia and built on AWS' Amazon Bedrock, is available in beta in the U.S. starting Thursday.
Amazon also said that some sellers in the U.S. will have access to the feature immediately, and then a broader rollout will begin nationwide. Later this year, the feature will appear in more countries and languages other than English.
Amazon continues to develop AI products aimed at entrepreneurs. The company also recently launched its video generator for advertising.
OpenAI Gives Access to o1 for Enterprise and Education Customers
OpenAI o1 is now available to all ChatGPT Enterprise and ChatGPT Edu users. However, the company clarified that this is still an “in preview” launch, so speed limits will apply. Both the o1-preview and o1-mini models have a speed limit of 50 requests per week. These models represent OpenAI's most advanced attempt to create an AI capable of deep multi-step reasoning.
If you want to know more about o1-preview, read our review:
YouTube AI Tools Will Let You Generate Ideas, Titles, and Videos
On Wednesday, at its Made on YouTube event, Google announced a series of AI features on the platform, including a couple that might change how creators make videos. The first feature is a new “Inspiration” tab in the YouTube Studio app. This offers to do much of the work for the creator. The AI can suggest a video idea, develop a title, and even write a plan to realize it. The developers call this solution a helpful brainstorming tool and recommend trying it to create entire projects.
The second feature is Veo, a background and video generation platform created by Deepmind engineers. This tool is now integrated into YouTube Shorts and will be part of the Dream Screen feature. However, Veo is now limited to generating videos up to six seconds long.
YouTube's plans to integrate AI features for its creators look pretty interesting. But it once again makes me wonder how precarious a position developers of other AI tools are in.
By the way, you can create content for TikTok & Shorts using AI right now. Follow this newsletter to find out how:
Sharing is caring! Refer someone who recently started a learning journey in AI. Make them more productive and earn rewards!
BlackRock, Microsoft, and MGX Launching $100B AI Investment Plan
If you thought the AI industry was approaching the limit regarding investment size, you’re wrong. BlackRock Inc., Global Infrastructure Partners, Microsoft, and MGX are launching the most significant fund to invest in AI infrastructure.
The partnership initially aims to raise $30B in private equity capital, which could later turn into $100B in total investment if debt financing is included. The money will be used to manufacture chips, build data centers and power plants, and support project participants.
The critical point is that we're not just discussing Microsoft projects. The company claims to support “open architecture and a broad ecosystem” that will give access to many developers. In this way, the entire industry will benefit from the initiative.
Runway Brings AI to Hollywood by Partnering with Lionsgate
A famous video generation model developer has partnered with one of Hollywood's biggest studios. In a mutually beneficial collaboration, Lionsgate, the company behind the John Wick and Twilight franchises, has gained access to a specially developed AI. At the same time, Runway will have access to an extensive library of movies and shows. Lionsgate will use the model to edit and produce future content.
I don't want to excite ahead of time, but a partnership like this looks like the start of something bigger. According to Michael Burns, vice chairman of Lionsgate Studio, the company will save “millions and millions of dollars” by using the new model. Right now, Lionsgate plans to use the AI tool for internal purposes like storyboarding and later for background and special effects creation.
Useful Tools ⚒️
Zivy – Automatically prioritize and organize messages in Slack
SocialAI – Your AI-powered social network
Thunderbit – Personalized web AI copilot
Atlantix – A platform to discover, develop, & commercialize innovations
Surfsite AI – AI assistants for SaaS professionals
Surfsite is a productivity tool that uses AI to help creators and entrepreneurs manage their work more effectively by integrating various tools and providing intelligent assistance. It's not a bad solution that optimizes many processes without costing time. Here are the key features of this platform include:
Centralized management of tools and data
AI assistants to help with tasks and decision-making
Real-time insights and analytics
Integration with familiar tools like Google Docs, Jira, Trello, etc.
Share this post with friends, especially those interested in AI stories!
Weekly Guides 📕
How to Make AI-Powered Interactive Dashboard | Bricks Tutorial
Create Cinematic Ai Videos for FREE with Minimax! - Prompts & Tutorial
Cursor AI tutorial for beginners
How to Make AI Images/Videos of Yourself! (For Free)
OpenAI Releases Smartest AI Ever & How To Use It
Ultimate Guide to Writing Effective AI Prompts
AI Meme Of The Week 🤡
AI Tweet Of The Week
Our tweet of the week again goes too well with the meme above.…
(Bonus) Materials 🏆
How CEOs Are Using Gen AI for Strategic Planning
One in five GPs use AI such as ChatGPT for daily tasks, survey finds
I Stared Into the AI Void With the SocialAI App
A bottle of water per email: the hidden environmental costs of using AI chatbots
Can AI improve one of the most expensive (and boring) parts of work?
Share this edition with your friends!