AIR#26 - March 30, 2024

Good morning, AI enthusiasts! As you savor the first sip of your freshly brewed coffee, today's edition of AIR: The AI Recon is brimming with stories that are sure to ignite your imagination and fuel your curiosity. At the forefront of today's tales of innovation, we're diving into the world of voice cloning with OpenVoice's groundbreaking advancements, offering instant, multilingual, and emotionally adaptable voice cloning that's setting new standards in the field. It's not just about creating echoes of our voices; it's about reshaping the way we interact with technology, making it more personal and accessible than ever before.

But the excitement doesn't stop there. We're also exploring the latest leap in AI-driven reasoning with the introduction of Grok-1.5, boasting enhanced reasoning capabilities and an impressive 128K token context length. This development is not just a step forward; it's a giant leap for AI applications, from coding to complex problem-solving, showcasing the relentless pace of innovation in the AI landscape. It's a testament to the boundless potential of AI to transform our digital interactions, making them more intuitive and efficient.

And for those with a keen eye on the intersection of AI and business, Demis Hassabis's vision for Google's future shines bright in today's lineup. With a track record of turning games and protein folding into AI breakthroughs, Hassabis's efforts to keep Google at the cutting edge of AI innovation amidst fierce competition and challenges are nothing short of inspiring. It's a narrative of ambition, resilience, and the transformative power of AI that's shaping the future of technology. So, as you enjoy your morning coffee, let's delve into these stories and more, exploring the fascinating world of AI where innovation meets impact, one breakthrough at a time.

Business

🔥 Can Demis Hassabis's AI Mastery Propel Google into the Future?
Demis Hassabis, DeepMind's founder, aims to keep Google at AI's cutting edge with breakthroughs from games to protein folding, amidst challenges and competition.

OpenAI's Voice Engine: Revolutionizing Speech with AI
OpenAI's Voice Engine crafts realistic voices from a 15-sec sample, aiming for safe, beneficial AI uses while navigating misuse risks.

Building a GPT-Powered Company in 2024: A Practical Guide
Embrace the future: Dive into our practical guide for building a GPT-powered company in 2024, from basic adoption to pioneering custom AI solutions. 🚀🤖 #AIRecon

NYC AI Chatbot Advises Businesses to Illegally Discriminate, Misinforms on Laws
NYC AI chatbot misleads businesses into illegal practices, spreading false information on laws and rights. Urgent fix needed.

NYC AI Chatbot Misleads Businesses Into Illegal Practices
NYC's AI chatbot misadvises businesses, suggesting illegal practices, risking fines and legal trouble. Urgent fixes needed.

Join the PhoneScreen.AI Beta: Revolutionize Recruiting with AI-powered Phone Screens
Revolutionize hiring with PhoneScreen.AI: AI-powered screens mean no more resume dungeons, just efficient candidate ranking. Join the beta!

Engineering

🔥 [Paper] OpenVoice: Instant Voice Cloning with Multilingual and Emotional Flexibility
OpenVoice revolutionizes voice cloning: instant, multilingual, and emotionally adaptable, outperforming rivals at a fraction of the cost.

Announcing Grok-1.5: Enhanced Reasoning and 128K Token Context Length on 𝕏 Platform
Grok-1.5 launches with enhanced reasoning, 128K token context, excelling in math and coding benchmarks, soon on 𝕏 platform.

🔥 [GitHub] Arraymancer: Fast, Portable Deep Learning Library in Nim
Arraymancer: Nim's tensor library accelerates deep learning across CPU, GPU, and devices with OpenMP, Cuda, OpenCL. Fast, ergonomic, portable.

[Paper] Qwen1.5-MoE: Achieving 7B Model Performance with Only 2.7B Activated Parameters
Qwen1.5-MoE model matches 7B models' performance with only 2.7B parameters, slashing training costs by 75% and speeding up inference by 1.74x.

[Google Research] AutoBNN: Revolutionizing Time Series Forecasting with Bayesian Neural Networks
Google's AutoBNN revolutionizes time series forecasting by blending Bayesian neural networks' scalability with traditional models' interpretability.

[Github] TornadoVM Enhances Java GPU Capabilities with Tensor API v0.1 and ONNX RT Integration
TornadoVM's new Tensor API v0.1 boosts Java GPU with Tensor utilities & ONNX RT integration, enabling advanced AI capabilities.

[GitHub] Google DeepMind's LongFact: Benchmarking Long-Form Factuality in Large Language Models
Google DeepMind's LongFact project sets a new standard in evaluating long-form factuality of AI models, offering tools and benchmarks for improvement.

NVIDIA Grace Hopper Superchips Power HPE Cray EX254n Blade at GTC 2024
NVIDIA Grace Hopper Superchips revolutionize HPE Cray EX254n blade, merging AI & HPC power at GTC 2024.

OpenAI's Voice Engine: Transforming Text to Speech with AI
OpenAI's Voice Engine now turns text into speech mimicking any voice from a 15-sec sample, revolutionizing speech synthesis AI.

OpenAI's Voice Engine: Transforming Text to Your Voice in Seconds
OpenAI's new Voice Engine can mimic your voice from a 15-second clip, transforming text to speech that sounds just like you.

[Guide] MLMY: Machine Learning My Way by Evgeny Pogrebnyak
Evgeny Pogrebnyak's "MLMY: Machine Learning My Way" offers a fresh, organized dive into ML, from basics to deep learning, with practical links and updates.

Ray AI Framework's Security Negligence Exposed by Oligo Researchers
Ray AI's lack of security features exposes thousands of projects to exploits, challenging modern cybersecurity standards.

Graph Algorithms Boost Cyber Threat Detection in Machine Learning: A University of West Florida Study
University of West Florida study reveals graph algorithms significantly enhance machine learning in detecting cyber threats, offering new strategies for cybersecurity.

[GitHub] VoiceCraft: State-of-the-Art Voice Cloning and Editing
VoiceCraft on GitHub revolutionizes voice cloning and editing, offering zero-shot TTS and speech edits with just seconds of reference audio.

[GitHub] MyScaleDB: Open-Source SQL Vector Database Built on ClickHouse
MyScaleDB launches on GitHub: an open-source SQL vector database optimized for AI, built on ClickHouse for fast, scalable data management.

[Github] Lava: Comprehensive Framework for Neuromorphic Computing
Lava, GitHub's comprehensive neuromorphic computing framework, enhances deep learning and neural networks with cutting-edge updates.

Academic

[Paper] Automating Text Mining with TnT-LLM on Large Language Models
TnT-LLM revolutionizes text mining, automating label generation & classification with minimal human effort, enhancing accuracy & efficiency at scale.

[Paper] Benchmarking Long-form Factuality in LLMs with SAFE Method
New study shows SAFE method boosts LLMs' fact-checking ability, making AI cheaper and more reliable than human annotators.

[Paper] The Impact of GenAI Detection Tools on Academic Integrity and Inclusivity
GenAI text detectors falter against modified content, challenging academic integrity and fairness in education.