🤩OpenAI’s Exciting New Mini Model

OpenAI launches GPT-4o Mini—a smaller, cheaper, and smarter AI model

Key Points 

• Compared to GPT-3.5 Turbo, GPT-4o Mini is more than 60% cheaper to run.

• For now, the new lightweight model will support text and vision through the API.

• The company plans to expand its capabilities to handle all multimodal inputs and outputs, such as video and audio, soon.

🧑‍💻News - OpenAI just rolled out GPT-4o Mini, their latest small AI model, which is cheaper and more powerful than GPT-3.5. This model is mainly for developers who want to build apps and tools without breaking the bank on a larger model like GPT-4. But it's also going to be available for regular users on the ChatGPT web and mobile apps.

Starting today, if you’re using ChatGPT on a Free, Plus, or Team plan, you'll be able to use GPT-4o Mini instead of GPT-3.5 Turbo. Enterprise users will get access next week. While GPT-3.5 won’t be an option anymore for ChatGPT users, developers can still use it through the API if they prefer.

For now, the new lightweight model will support text and vision through the API. The company plans to expand its capabilities to handle all multimodal inputs and outputs, such as video and audio, soon.

🕵️How does it compare? 

• Compared to GPT-3.5 Turbo, GPT-4o Mini is more than 60% cheaper to run. It will cost developers 15 cents per million input tokens and 60 cents per million output tokens.

• The model has a context window of 128,000 tokens, roughly the length of a book, and its knowledge is current up to October 2023.

• While OpenAI hasn't shared its exact size, they say it's similar to other small models like Llama 3 8b, Claude Haiku, and Gemini 1.5 Flash.

• Additionally, according to George Cameron from Artificial Analysis, GPT-4o Mini is super fast, with an output speed of 202 tokens per second.

‘Godmother of AI’ Fei-Fei Li’s new AI startup reaches $1 billion valuation in four months

Key Points 

• World Labs is developing AI with advanced 3D spatial intelligence for various tech fields.

• The startup has raised $100 million and is valued at over $1 billion in just four months, reflecting a broader trend of investing in AI following ChatGPT’s success.

👩‍🔬News - Fei-Fei Li, the renowned computer scientist also known as the "godmother of AI," has launched a startup named World Labs, which has already hit a valuation of over $1 billion in just four months.

For those who might not be familiar, Li is best known for her contributions to computer vision, including developing ImageNet, a key visual database for object recognition. She also led AI at Google Cloud and now advises the White House task force on AI.

🥸What’s the plan for World Labs? According to an anonymous venture capitalist, World Labs is working on a model that can understand the three-dimensional physical world—how objects fit together, where they are, and what they do. This matches up with a Reuters report from May, which said World Labs aims to make AI capable of advanced reasoning by processing visual data like humans do.

In an April TED Talk, Fei-Fei Li explained that her startup is focusing on algorithms that can turn images and text into 3D environments and act based on those insights, a concept called "spatial intelligence." This could boost advancements in robotics, augmented reality, virtual reality, and computer vision, potentially transforming fields like healthcare and manufacturing if Li’s ambitious vision comes to fruition.

🌐Capitalizing on the AI investment boom? World Labs has raised funds twice, with the latest round bringing in about $100 million. It's backed by Andreessen Horowitz and Radical Ventures (where Li joined as a partner last year). However, these investments actually highlight a broader trend of venture capitalists eagerly supporting bold AI startups, especially following the unexpected success of OpenAI’s ChatGPT.