. . AI: Google to Turn Pencil Drawings Into Playable Gables (2.26.24)

Friends, today's edition delves into the latest developments from the open source community to companies like Google. Once again we see how Google is being challenged but also doing leading tech research. Zoom out a little further and one thing is clear, a lot of companies are looking to expand their reach throughout the AI value chain as they search for their footing and a defensible business strategy.

Welcome to our new readers and thank you to those helping spread the word. If you find this email useful, we’d love it if you’d share it with a friend.

Now here are today’s stories.

-Marshall Kirkpatrick, Editor

Google Team Launches Genie, 2D AI Model with Big Promise for Robotics

First impacted: AI researchers, Game developers
Time to impact: Medium

Google's Open Endedness Team introduces Genie, an advanced AI with 11 billion parameters, capable of autonomously learning and turning images (paintings or children's drawings for example) into interactive 2D environments. This model stands out for its ability to learn without supervision and its potential beyond 2D applications, including training on robotics data to approach real-world models. Plus it's really cool to imagine drawing a world in pencil and then making it playable on your computer. [via @_rockt] Explore more of our coverage of: Google DeepMind, Autonomous Learning AI, 2D Environment Modeling. Share this story by email

Mistral Debuts Advanced Language Model, Mistral Large, as a Close Competitor to GPT-4

First impacted: AI developers, AI researchers
Time to impact: Short

Mistral has launched Mistral Large, a new language model, which the company says is only surpassed in performance by GPT-4. According to a blog post from Mistral, the model can converse in five languages, execute commands and has a 32k token context. Notably, Microsoft formed a partnership with Mistral and the new model is available on the Azure platform, as well as self-deployment. [Au Large] Explore more of our coverage of: Mistral Large, Language Models, Azure Partnership. Share this story by email

Gartner Predicts Search Engine Volume Will Drop 25% by 2026 due to AI

First impacted: Advertisers, Content Creators, Marketers
Time to impact: Long

Gartner, Inc. says that by 2026, traditional search engine use will decrease by 25% due to the rise of AI chatbots and generative AI (GenAI) systems. They argue that as GenAI reduces content creation costs, the focus will shift towards content quality and verification, potentially reshaping technology and marketing strategies. [Gartner Predicts Search Engine Volume Will Drop 25% by 2026, Due to AI Chatbots and Other Virtual Agents] Explore more of our coverage of: AI Chatbots, Generative AI, Technology Marketing. Share this story by email

OpenHermesPreferences v0.1 Launches with 1 Million AI Preferences Dataset

First impacted: AI researchers, Model Tuners, Data scientists
Time to impact: Short

The team behind OpenHermesPreferences v0.1 has launched a dataset of around one million AI preferences garnered from different LLMs. According to the creators, this dataset integrates responses from teknium/OpenHermes-2.5, Mixtral-8x7B-Instruct-v0.1, and Nous-Hermes-2-Yi-34B and uses Pairwise Reward Model (PairRM) to score and rank responses, facilitating the training of models or the alignment of LLMs via Direct Preference Optimization techniques. This is a great tool for the opensource community to finetune models. [argilla/OpenHermesPreferences · Datasets at Hugging Face] Explore more of our coverage of: AI Preferences Dataset, LLM Alignment, Direct Preference Optimization. Share this story by email

Qualcomm Launches AI Hub for Developers

First impacted: Mobile App Developers, AI Model Developers
Time to impact: Short

Qualcomm Technologies has launched the Qualcomm AI Hub, a collection of AI models designed for developers and optimized to take advantage of their hardware acceleration by using all compute cores. This reminds us of how chipmaker Nvidia announced an AI factory in October 2023; hardware companies are moving up and down the AI value chain to minimize the roadblocks of building and incorporating AI models in their ecosystems. [Home - Qualcomm® AI Hub] Explore more of our coverage of: Qualcomm AI Hub, AI Development, Snapdragon Platforms. Share this story by email