AIR#58 - May 01, 2024

Good morning, AI aficionados! Today’s AIR: The AI Recon edition is like opening a treasure chest of the digital age, brimming with gems that sparkle with innovation, controversy, and a dash of the unexpected. From the hush-hush whispers of Apple’s covert operations in Zurich, poaching AI wizards from Google, to the legal battleground where Mercury News and its allies draw swords against Microsoft and OpenAI over the contentious use of copyrighted content. These stories aren’t just headlines; they’re the pulsating heart of ongoing debates about the ethics, ownership, and future direction of AI technology.

As we navigate through the thicket of legal tussles, there’s a glimmer of revolutionary promise on the horizon with AI’s new frontier in gene editing. The CRISPR technology, now supercharged by AI, is on the brink of transforming medical science, promising a future where genetic disorders are no longer a life sentence but a curable anomaly. This leap towards precision medicine is not just a win for science; it’s a beacon of hope for millions, illustrating AI’s potential to heal and redefine life’s blueprint.

But let’s not forget the playground of the programmers and the dreamers, where AI’s influence morphs from the profound to the practical. Android Studio’s Gemini and Tesla’s grand plan to turn parked cars into a mobile AWS for AI processing are testaments to AI’s relentless march into every corner of our lives, from how we code to how our cars might soon moonlight as data centers. As you sip on your morning brew, let these stories unfold a day filled with curiosity, caution, and the ceaseless quest for innovation that defines our journey through the AI landscape. Let’s dive into the wonders and challenges that lie ahead, shall we?

Business

Apple Steals AI Experts from Google to Launch Secret European Lab
Apple poaches AI experts from Google for a secret AI lab in Zurich, eyeing advanced AI models for mobile devices.

Mercury News and Others Sue Microsoft, OpenAI for Copyright Infringement
Mercury News and others sue Microsoft, OpenAI for using copyrighted articles to train AI, demanding fair compensation and legal action.

Tecton's Guide to Scaling and Serving AI Embeddings in Production
Tecton's new AI embedding solution tackles production challenges, enhancing predictive & generative AI applications with efficient, scalable infrastructure.

Newspapers Sue OpenAI and Microsoft Over AI Training Data Use
Major newspapers sue OpenAI & Microsoft for using their content to train AI without permission.

India's Election AI Boom: $50 Million on Synthetic Campaigns
India's election sees $50M boom in AI content as parties hire synthetic media firms to craft campaigns, setting ethical lines against deepfakes.

Tesla Plans to Turn Its Cars into Mobile AWS for AI Processing
Tesla eyes turning parked cars into a mobile AWS for AI processing, leveraging idle vehicle compute power for profit.

Engineering

Symbl.ai's Nebula LLM Outperforms Claude 3, GPT-4, and Llama 3 in Emotional Intelligence Benchmark
Symbl.ai's Nebula LLM tops Claude 3, GPT-4, & Llama 3 in emotional intelligence, leading the way in human-like AI interactions.

Writing Malware with AI: No Dev Team Required
AI now writes malware & anti-malware, no coding skills needed. Experiment reveals ease & risks of generative AI in cybersecurity.

[VSCode Extension] Traycer: Real-Time AI-Powered Code Analysis
Traycer, a new VSCode extension, offers live AI-powered code analysis, enhancing software quality and developer productivity with real-time insights.

Android Studio Introduces Free AI-Powered Coding Assistant Gemini
Android Studio's Gemini, the AI coding assistant, is revolutionizing app development with smart code completion, error fixes, and tailored advice.

[Github] SugarAI's CopilotOne: Integrate Siri-Like AI Assistants into React Apps
SugarAI launches CopilotOne on GitHub, enabling React apps to integrate Siri-like AI assistants for voice and text actions, enhancing user interaction.

[Github] Bumpgen: AI-Powered NPM Package Upgrader by xeol-io
Bumpgen by xeol-io automates NPM package upgrades with AI, fixing breakages in TypeScript projects. Say goodbye to manual version bumps!

[Github] GigaxGames/gigax: Local LLM-Powered NPCs for Game Development
GigaxGames launches gigax, enabling game devs to create LLM-powered NPCs that run locally, enhancing gameplay with dynamic actions and memory.

Amazon Q: The Ultimate AI-Powered Development and Data Assistant
Amazon unveils Amazon Q: AI to boost software dev and data analysis, promising productivity leaps and enhanced privacy.

AI's Reluctance to Depict Jesus in a Servant Role: A Curious Exploration
Exploring why AI struggles to depict Jesus in a servant role, revealing biases and technical limitations in image generation.

[Paper] RAGCache: Boosting Efficiency in Retrieval-Augmented Generation with Dynamic Knowledge Caching
RAGCache revolutionizes AI by dynamically caching knowledge, slashing computation costs and doubling throughput for language models.

[Open Source] LMSYS and UC Berkeley's Chatbot Arena: A Community-Driven LLM Evaluation Platform
LMSYS & UC Berkeley launch Chatbot Arena, a live platform for community-driven LLM evaluation, enhancing model understanding through open-source collaboration.

🔥 [Book] Alice's Adventures in a Differentiable Wonderland by Simone Scardapane: A Primer on Differentiable Programming
Dive into "Alice's Adventures in a Differentiable Wonderland" for a primer on the magic of differentiable programming, bridging theory and code.

🔥 [GitHub] Run Llama3 with 1M+ Token Context Locally
Run Llama3 8B locally, now with 1M+ token context for deeper AI insights. Powered by Gradient and Crusoe Energy, pushing LLM boundaries.

Academic

AI-Designed CRISPR Tools Revolutionize Gene Editing
AI revolutionizes CRISPR gene editing, designing new tools for precision medicine with machine learning, promising less errors and wider applications.

Understanding Language Models as n-Dimensional Objects in Space by Alex Levy
Alex Levy explores LLMs as n-dimensional objects in space, suggesting understanding and generative abilities stem from navigating their complex surfaces.

Dean Ball Critiques California's New AI Bill as Science Fiction Turned Law
Dean Ball criticizes California's AI bill as overly ambitious, likening its sci-fi-esque restrictions to potentially stifling tech innovation.

[Paper] Crafting the Largest Japanese Web Corpus for Superior LLM Training
New study unveils the largest Japanese web corpus, enhancing LLMs with unparalleled text quality and size, outperforming previous datasets.

Rethinking AI Evaluation: The Case for Pareto Curves Over Leaderboards
Out with AI leaderboards, in with Pareto curves for cost-effective accuracy evaluation. Say goodbye to misleading metrics.

[Paper] Evaluating LLMs with a Panel of Models: A Cost-Effective, Bias-Reducing Approach
New study shows evaluating LLMs with a diverse panel of smaller models is cheaper, reduces bias, and outperforms single large-model methods.