TLDR AI 2024-04-03

State of AI Market Map 🗺️, Xbox AI Chatbot 🤖, Enhanced Multimodal LLM 🖼️

🚀
Headlines & Launches

10th Annual State of AI Market Map (30 minute read)

An in-depth analysis of the state of companies and solutions in AI. There are well over two thousand companies working on all sorts of AI problems. This post also explores open source, wrappers, funding, and much more.

Microsoft Is Working On An Xbox AI Chatbot (2 minute read)

Microsoft is testing a new AI-powered Xbox chatbot designed to automate support tasks, expanding its testing pool and integrating the bot with Xbox's support documents to answer queries and process game refunds.

Sam Altman No Longer Owns The OpenAI Startup Fund (1 minute read)

OpenAI CEO Sam Altman has handed over control of OpenAI's corporate venture fund, the OpenAI Startup Fund, to Ian Hathaway, who joined OpenAI in 2021 and played a key role in managing the Startup Fund.
🧠
Research & Innovation

Industrial Documents OCR dataset (2 minute read)

A huge dataset of 26m pages and 18B tokens of extremely high-quality OCR’d dataset of industrial PDF documents.

Many shot jailbreaking (15 minute read)

Anthropic developed a technique to jailbreak long context models. It has shared these findings with other organizations and implemented mitigations. This post outlines the technique and some of the things it did to defend against the technique.

Efficient Video Understanding (38 minute read)

Researchers have developed R2-Tuning for understanding videos by identifying specific moments based on verbal prompts.
👨‍💻
Engineering & Resources

Open source software agent matches Devin performance (GitHub Repo)

SWE-agent gets 12% on 100% of SWE bench. It uses GPT-4 to help it write software and solve PRs.

Chug dataset loader (GitHub Repo)

Powerful and efficient multimodal dataset loader focused on text and image tasks.

Enhanced Multimodal LLM (3 minute read)

The Draw-and-Understand project's SPHINX-V is a multimodal large language model designed to deepen human-AI interaction through visual prompts.
🎁
Miscellaneous

Building AI Models Is Faster and Cheaper Than You Probably Think (2 minute read)

YC companies are defying the notion that building AI models requires immense resources by training or fine-tuning their own foundation models with YC's support, achieving remarkable feats like generating professional music and designing novel proteins within just three months. These 25 startups, leveraging YC's funding and technical resources, have developed innovative AI solutions across various fields, demonstrating that with creativity and strategic insights, significant advancements in AI are accessible to smaller teams.

Synthetic data at scale for pre-training models (17 minute read)

The HuggingFace team shows how to seed, synthesize, filter, and scale synthetic data creation for language model pre-training.

Generative SF: Marketplaces in AI Edition (4 minute read)

In this article, Instacart and Faire leaders discuss how Generative AI is enhancing their B2B and B2C marketplaces, improving internal workflows, and personalized customer interactions. They highlight AI's role in accelerating mission-critical tasks, boosting productivity, and refining operational efficiency. Both companies emphasize that while AI tools are transformative, the core business principles remain unchanged.
⚡️
Quick Links

Replit Launches New Product In Race For AI Coding Assistants (2 minute read)

Replit is launching Replit Teams, a new tool that allows developers to collaborate in real-time on software projects with an AI agent that automatically fixes coding errors.

AutoQuant notebook (Colab Notebook)

This notebook can export HuggingFace models into 5 different quantization formats: GGUF, GPTQ, EXL2, AWQ, and HQQ.

Samsung Might ‘Redefine’ Bixby With Galaxy AI After All (1 minute read)

Samsung is considering an upgrade for its voice assistant Bixby, incorporating generative AI technology similar to what powers its Galaxy AI features on the Galaxy S24 series.
The most important AI, ML, and data science news in a free daily email.
Join 500,000 readers for