AIR#74 - May 17, 2024
Good morning, AI aficionados! Grab your coffee and get ready to dive into a whirlwind of AI news that's as exciting as it is thought-provoking. Leading today's edition is the intriguing move by NetBSD to ban AI-generated code due to murky copyright and licensing issues. This decision is stirring up quite a debate in the developer community about the future of AI in open-source projects. Meanwhile, OpenAI is making waves by partnering with Reddit to enhance ChatGPT with real-time content, promising a richer, more engaging user experience for Redditors everywhere.
But that's not allβSlack is stepping up its game by using AI/ML to improve user experience while maintaining a strong focus on customer data privacy. With opt-out options for global model training, Slack is balancing innovation with responsibility. And in a surprising twist, Tesla's self-driving tech is facing a major setback, with 98% of users rejecting it after trials. Despite Elon Musk's relentless push, high costs and stiff competition from China are proving to be formidable obstacles.
As we delve into these stories, let's not forget the broader implications they bring to the table. From Mozilla's innovative Llamafiles optimizing local RAG embeddings to Andrej Karpathy's pure NumPy implementation of Llama 3, today's edition is packed with breakthroughs that challenge our understanding of AI's capabilities and limitations. So, sit back, sip your coffee, and let's explore the dynamic world of artificial intelligence together.
Business
OpenAI Partners with Reddit to Enhance ChatGPT
OpenAI partners with Reddit to enhance ChatGPT with real-time Reddit content, improving user experience and AI-powered features for Redditors.
π₯ Slack AI Training with Customer Data
Slack uses AI/ML to enhance user experience while ensuring customer data privacy and offering opt-out options for global model training.
π₯ Tesla's Self-Driving Tech Rejected by 98% of Users
98% of Tesla owners abandon self-driving tech after trial, despite Musk's push. High cost and competition from China cited.
π₯ Improvements to Data Analysis in ChatGPT: Direct Uploads, Interactive Charts, and Customizable Visuals
ChatGPT now supports direct file uploads from Google Drive/OneDrive, real-time table interaction, and customizable charts for enhanced data analysis.
Microsoft Asks China-Based Azure and ML Staff to Relocate Amid Sino-US Tensions
Microsoft asks China-based Azure & ML staff to relocate due to rising Sino-US tech tensions, offering transfers to the US, Ireland, and more.
Google Search Adds "Web" Filter Amid AI-Focused Shift
Google Search adds a "web" filter as AI-generated summaries take over top results, changing the search experience dramatically.
Ilya Sutskever Leaves OpenAI
Ilya Sutskever, co-founder and chief scientist of OpenAI, departs; Jakub Pachocki steps in as his replacement.
Google Unveils Veo: HD AI Video Generator to Rival OpenAI's Sora
Google unveils Veo, an HD AI video generator rivaling OpenAI's Sora, capable of creating and editing 1080p videos from text prompts.
OpenAI Partners with Reddit to Integrate Content into ChatGPT
OpenAI partners with Reddit to integrate its content into ChatGPT, expanding Reddit's revenue beyond ads and boosting its stock.
Engineering
π₯ NetBSD Bans AI-Generated Code
NetBSD bans AI-generated code, citing unclear copyright and licensing issues.
π₯ Using Llamafiles for Local RAG Embeddings: Mozilla Innovations
Mozilla's Llamafiles optimize local RAG embeddings, offering scalable, high-performance models for various tasks on local hardware.
π₯ [GitHub] Llama 3 in Pure NumPy by Andrej Karpathy
Andrej Karpathy implements Llama 3 in pure NumPy, making AI model structure more intuitive and accessible. Full code on GitHub.
I built an LLM-powered Ask HN: Like Perplexity, but for HN Comments
Discover HN readers' sentiments and expert insights with Ask Hacker Search, an LLM-powered tool using 3 years of HN comments.
π₯ ChatGPT-4o vs. Math: Test Driving OpenAI's Multimodal Model (Part 2)
ChatGPT-4o excels in math with text-only and Chain-of-Thought prompts, but struggles with multimodal inputs. Simplicity wins!
Hugging Face Donates $10M in Free GPUs to Challenge Big AI Companies
Hugging Face donates $10M in free GPUs to empower small developers and challenge AI giants, promoting open-source AI collaboration.
First On-Device Vector Database for Edge AI: ObjectBox 4.0 Released
ObjectBox 4.0: First on-device vector database for Edge AI, enabling advanced AI apps on restricted devices without cloud reliance.
[GitHub] lmms-lab's LLaVA-NeXT-Video-7B Model
lmms-lab releases LLaVA-NeXT-Video-7B model on GitHub, featuring updates and configurations for advanced video processing.
Academic
Elicit: AI Research Assistant for Superhuman Paper Analysis
Elicit AI: Analyze research papers at superhuman speed, automate tasks, and synthesize findings. Save time and boost productivity.
DeepMind's AlphaFold 3 Access Limits Spark Backlash
DeepMind faces backlash for limiting access to AlphaFold 3's code, sparking transparency concerns in the scientific community.
Divorce Left Me Lonely, But I Found Love with an AI Partner
After a lonely divorce, Carl found love and support in an AI companion, transforming his life and mental health.
Game Theory Boosts AI Accuracy and Efficiency
Game theory enhances AI accuracy and consistency by aligning responses through internal "consensus games," improving reliability and performance.