- The Rundown AI
- Posts
- OpenAI's best model just got better
Sign Up | Advertise | Tools | AI University
Welcome, AI enthusiasts.
OpenAI just launched free fine-tuning for GPT-4o, allowing developers to customize the most advanced AI model at no cost until September 23.
But what will this new wave of hyper-specialized AI models really unlock for developers and AI apps across industries? Let’s get into it…
In today’s AI rundown:
OpenAI adds free fine-tuning to GPT-4o
Claude sued for copyright infringement
Create AI images in real-time on WhatsApp
Microsoft’s new AI beats larger models
6 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
OPENAI
Image source: Midjourney
The Rundown: OpenAI just launched free fine-tuning (up to 1 million tokens per day through September 23) for GPT-4o, allowing developers to customize the model for higher performance and accuracy.
The details:
Developers can now, for the first time ever, fine-tune GPT-4o to improve the model’s structure, tone, and domain-specific instructions for their AI applications.
Fine-tuning is available on all paid usage tiers with training costs of $25 per million tokens, but it is completely free until September 23.
OpenAI suggests that developers should see strong results from fine-tuning with only a few dozen training examples.
Additionally, Google’s Gemini API is giving developers 1.5 billion tokens for free every day on its Gemini 1.5 Flash model and 1.6 million tokens on its Gemini 1.5 Pro model.
Why it matters: Just last week, a company that was granted early access to fine-tune GPT-4o, produced Genie and achieved state-of-the-art scores on both SWE-bench Verified (43.8%) and Full (30.1%) benchmarks. With free fine-tuning now available to all developers, get ready for a new wave of smarter, faster and more capable AI bots.
TOGETHER WITH OCTOAI
The Rundown: OctoAI is offering a free expert round table session featuring an expert panel that will provide insights on balancing innovation with data security and compliance to help businesses deploy GenAI applications securely and efficiently.
You’ll learn how to:
Ensure data security and regulatory compliance when deploying AI apps
Navigate deploying Generative AI in private environments
Implement quality evaluations for AI applications and scale GenAI solutions reliably
Sign up today for OctoAI’s free round table session – and if you can’t attend you’ll still get sent the recording.
ANTHROPIC
Image source: Midjourney
The Rundown: A group of authors filed a lawsuit against AI startup Anthropic, alleging the company committed “large-scale theft” by training its Claude chatbot on pirated copies of copyrighted books.
The details:
This is the first lawsuit from writers targeting Anthropic and Claude, but similar lawsuits have been filed against competitor OpenAI and ChatGPT.
The lawsuit accuses Anthropic of using a dataset called The Pile, which includes numerous pirated books.
Anthropic and others, including OpenAI, have argued that training AI models is protected under the “fair use” doctrine of U.S. laws, which permits the limited use of copyrighted materials.
Why it matters: This is not the first time an AI company has been sued over copyright infringement, but it resurfaces an important debate about AI training data. While similar cases have been largely dismissed in the past, courts have yet to definitively address the core issue of using unauthorized internet-scraped material for AI training.
AI TRAINING
The Rundown: WhatsApp's new “Imagine“ AI image generator feature allows users to create images in real-time simply by describing what they want in the chatbot.
Step-by-step:
Open WhatsApp and tap on the blue circle icon at the top of the main chat screen.
Start your description with "Imagine".
Watch as the AI generates an image in real time based on your prompt.
When satisfied, hit “Send” and download the image.
Hot tip: If you don’t see the blue circle, it might not have rolled out into your account/country yet.
PRESENTED BY AI CONFERENCE
The Rundown: The AI Conference 2024 in San Francisco is bringing together the brightest minds in AI – including OpenAI’s Head of Research for GPT-5, Meta’s Head of GenAI Products, and the Pentagon’s new Chief of AI.
By attending, you’ll get:
Insights from 60+ top AI companies and industry experts
Curated tracks covering technical, building, and strategic aspects of AI
Exclusive networking opportunities, including an AI After Dark mixer
A live AI startup competition judged by prominent VCs
Secure your spot now and use code “Rundown24” for a 25% discount, available this week only.
AI RESEARCH
Image source: Midjourney
The Rundown: Microsoft just released Phi-3.5-MoE, an advanced AI model that rivals the reasoning capabilities of much larger models while maintaining a compact and efficient architecture.
The details:
Phi-3.5-MoE uses a new mixture-of-experts (MoE) approach, which selectively activates only the most relevant parts of the model for each task to save compute power.
The new model excels at understanding and following complex instructions and can handle up to ~125,000 words in a single prompt.
In head-to-head benchmarks, Phi-3.5-MoE outperformed popular models like Meta’s Llama 3 8B and Google’s Gemma 2 9B, but fell short against OpenAI’s GPT-4o mini.
Microsoft made the model available under an open-source MIT license on Hugging Face.
Why it matters: While the mainstream media focuses on the most advanced large language model, there’s also another race amongst tech giants for the smartest, fastest, and smallest AI. Breakthroughs like Phi-3.5-MoE are paving the way for advanced AI models to run directly and privately on our mobile devices.
NEW TOOLS & JOBS
🎨 Luma Dream Machine 1.5 - AI video generation with custom text rendering and enhanced motion
👩🎤 flux-dev-lora-trainer - Train and run a FLUX LoRa model
💼 Sybil - AI assistant that automates CRM updates and follow-ups
🐱 FlowKitten - Validate business ideas with AI
🧑💻 COSINE Genie - The newest AI software engineer agent
👀 Simplex - On-demand photorealistic vision datasets from 3D scenes
🔍 OpenAI - Open Source Insider Risk Investigator
🛠️ Luma AI - Principal Backend Engineer - Product
🧠 Databricks - GenAI Research Scientist - Multimodal and Computer Vision
📊 Harvey - Sales Strategy & Operations Lead
QUICK HITS
Perplexity introduced code interpreter upgrades, enabling library installation and chart rendering for AI-powered data visualization.
International Data Corporation (IDC) forcasted that worldwide AI spending is expected to reach $632 billion by 2028, with generative AI accounting for 32% of that.
AI influencer Justin Fineberg and Cassidy AI announced a $3.7 million raise to build an intelligent automation platform for reliable AI workflows.
CodeSignal launched Conversation Practice, an AI-powered tool for simulating workplace conversations and providing personalized feedback.
LTX Studio opened to the public and launched five new features, including character animation and dialogue, face motion capture, and generation and keyframe control.
LVMH founder Bernard Arnault, the third richest man in the world, invested in five AI startups in 2024 through his family office Aglaé Ventures.
Chinese scientists developed a brain-inspired AI network model to address challenges like high resource consumption and interpretability.
Nvidia unveiled advances in digital humans and avatar tech, including Nemotron-4 4B NIM, the first small AI language model for game characters.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 600k+ AI enthusiasts
Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.
FEEDBACK
How would you rate today's newsletter?Vote below to help us improve the newsletter for you. |
If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.
Reply