- The Summary AI
- Posts
- 🔥 GPT-4o "Mini" Released
🔥 GPT-4o "Mini" Released
PLUS: Microsoft Designer AI App
Welcome back!
OpenAI has launched GPT-4o mini, a budget-friendly, multimodal AI model that will replace GPT-3.5 in ChatGPT, providing users with a smarter option. Additionally, API users will benefit from significantly reduced costs. Let’s unpack this...
Today’s Summary:
🆕 OpenAI introduces GPT-4o mini
📱 Microsoft Designer AI app
🤔 YouTube creators’ content used in AI training
đź’» Codestral Mamba code generator
🏛️ J.D. Vance AI regulatory agenda
đź“š Andrej Karpathy launches AI-education startup
2 new tools
TOP STORY
OpenAI launches budget-friendly GPT-4o mini for ChatGPT and API
The Summary: OpenAI has unveiled GPT-4o mini, a smaller version of its GPT-4o model that is 60% cheaper than GPT-3.5 Turbo. The new model supports text and vision inputs, with plans to add audio and video capabilities. GPT-4o mini will replace GPT-3.5 Turbo in ChatGPT for free and paid users, aiming to make the lower tier model smarter.
Key details:
Priced at $0.15/1M input tokens and $0.60/1M output tokens
Scores 82% on MMLU benchmark, surpassing Gemini 1.5 Flash (79%) and Claude 3 Haiku (75%)
128K token context window and knowledge up to October 2023
Multimodal capabilities: currently supports text and vision inputs
First model to apply "instruction hierarchy" for improved safety
Why it matters: This release is a step towards more affordable and accessible lower-tier AI for users and developers. By replacing GPT-3.5-Turbo for free users, it will increase capabilities and multimodality of the base model. The API version competes directly with Gemini-1.5-Flash, which currently had the best price-to-performance ratio.
TOOLS
Microsoft Designer AI tool for iOS and Android
The Summary: Microsoft Designer, an AI-powered creative tool, is now available on iOS and Android. The app offers AI-driven image generation, editing, and design features. Capabilities include image restyling, custom stickers, and greeting card creation. Microsoft aims to make creativity more accessible with AI assistance across its ecosystem.
Key details:
Free app available on iOS, Android, and Windows devices
15 free daily "boosts" for AI-powered creations
New features: Restyle image, Frame image, and Replace background
Integration with Microsoft 365 apps for Copilot Pro subscribers
Why it matters: This move brings new advanced AI design capabilities to mobile users. As tech giants compete in the AI space, consumers gain access to a wide range of creative options on their everyday devices.
DATA PRIVACY
YouTube creators’ content used in AI training
The Summary: Major tech companies, including Apple and Nvidia, have been using YouTube video transcripts to train their AI systems without creators’ permission. The data comes from an open-source collection. While this practice surprised creators, the use of publicly available data for AI training often falls under fair use.
Key details:
170K YouTube videos text transcripts were used for AI training
Includes content from popular creators like MrBeast and MKBHD
Companies involved include Apple, Nvidia, Anthropic, and Salesforce
Data source compiled by nonprofit EleutherAI
Why it matters: This controversy shows the complex challenges of data sourcing for AI development. There is a gap between legal rights and creator expectations in the AI era. This highlights the need for clearer guidelines on using publicly available content for AI training.
QUICK NEWS
Quick news
Codestral Mamba, a Mamba2 open-source model specialised in code generation
J.D. Vance AI Agenda: Reduce Regulation
OpenAI co-founder Andrej Karpathy launches AI-education startup
TOOLS
🥇New tools
Microsoft Designer - Create, design, & edit with AI
Otio - The fastest way to do research
That’s all for today!
If you liked the newsletter, share it with your friends and colleagues by sending them this link: https://thesummary.ai/