. . AI: Double WOW: Stunning Videos and 1Mil Token Lengths (2.15.24)

Friends, today's edition is nothing short of containing jaw dropping updates. The top two stories require no introduction, hurry up and check them out! The other two show the ecosystem is alive and kicking from a venture funding and development perspective. Enjoy!

Welcome to all our new subscribers here. New and longer readers: Who of your friends would appreciate this newsletter as much as you do? Care to forward it?

thanks! Now here are today’s top stories.

-Marshall Kirkpatrick, Editor

OpenAI Introduces Impressive Text-to-Video Model Called Sora

First impacted: Creative professionals, Media, Marketing
Time to impact: Short

OpenAI has introduced Sora, a text-to-video AI model capable of creating videos up to one minute long based on text inputs. It features the ability to execute multiple shots in a sequence while keeping the characters and visual style consistent. Despite Sora's ability to craft complex scenes with detailed characters and backgrounds, it struggles with simulating physical dynamics, understanding cause and effect, and grasping spatial details. Nonetheless, I strongly suggest viewing the videos in the link. Sora is currently in a limited release phase, as the OpenAI team aims to assess its potential impacts, especially on misinformation. [Sora: Creating video from text] Share this story by email

Google Launches Gemini 1.5 with a Whopping 1 Million Token Context Length

First impacted: AI developers, Data scientists
Time to impact: Short

Google has launched Gemini 1.5, an AI model that it claims is more efficient and yet performs similar to their largest model to date, Gemini 1.0 Ultra. Gemini 1.5 has been designed to process up to 1 million tokens, marking a new record for context window capacity in large-scale models. The additional processing power will allow users to upload hundreds of pages of text, entire code repos, and long videos. It achieves this using a new Mixture-of-Experts (MoE) architecture which enhances the efficiency of the model by dividing it into smaller "expert" neural networks that activate depending on the type of input given. The model is currently in limited preview but you can check out a demo in the link. [Our next-generation model: Gemini 1.5] Share this story by email

Magic.dev raises $117M to Build AI Coworker

First impacted: Software Developers, AI Engineers
Time to impact: Medium

Software development firm Magic.dev has secured $117M in funding in their pursuit of creating an AI software engineer that is a coworker, not just a copilot. Interestingly, their website states that they believe "Code generation is both a product and a path to AGI" which could explain the focus on developing AI software engineers first. [Your AI Pair Programmer | Magic] Share this story by email

LangChain Announces Series A and Launches LangSmith

First impacted: LLM Application Developers, Software Testers
Time to impact: Short

LangChain has launched LangSmith, a tool for developing, monitoring, and testing large language model (LLM) applications, according to a blog post from the company. The firm also announced a $25M Series A funding round led by Sequoia Capital, and highlighted that the tool has logged more than 40 million traces in January alone, post its beta release. (If you haven’t seen their Prompt Hub, check it out - it’s super cool!) [Announcing the General Availability of LangSmith and Our Series A Led By Sequoia Capital] Share this story by email

That’s it! More AI news tomorrow!