Will AGI Serve Us? The Human Factor

Imagine waking up one morning to find your smartwatch flashing with an urgent notification: "Potential health concern detected. Please consult your doctor within 48 hours." What if that simple alert could save your life?

That's the world we're stepping into, friends! This week, I'm taking you on a journey through three fascinating AI breakthroughs that are changing how technology touches our lives.

Picture this: In a bustling hospital, doctors receive alerts about patients who might develop kidney problems two whole days before symptoms appear. Meanwhile, across town, researchers slip on special glasses that can literally see how excited or calm you feel while using them. And halfway around the world, a creative team types a few words and watches as lifelike videos spring to life before their eyes.

The future isn't just knocking—it's walking right through our front door and making itself at home! But here's the twist in our story: these powerful tools are only as good as the human touch behind them. When DeepMind's kidney prediction system was trained mostly on men's health data, it didn't work as well for women—a reminder that even the smartest AI needs our human wisdom to truly help everyone.

So grab your favorite drink, get comfy, and let's explore how the latest tech wonders are bringing us closer to a world where AI doesn't just compute—it understands, adapts, and maybe even cares.

Welcome to this week's adventure in tech with heart!

Dive into these groundbreaking stories that show us not just what AI can do, but how it's becoming more human every day.

Humanizing AGI: The Power of HCI

The linked DeepMind publication highlights their development of a deep learning model capable of predicting acute kidney injury (AKI) up to 48 hours in advance, which they describe as a significant breakthrough in healthcare AI. The model aims to address one of the leading causes of preventable patient harm by identifying deterioration before it occurs. However, the research has faced criticism due to its reliance on skewed training data, with 93.6% of the dataset being male and only 6.38% female, leading to lower performance for women. Despite these limitations, the study underscores the importance of representative datasets in healthcare AI and suggests opportunities for public healthcare systems to leverage their data for similar advancements.

Meta's AR Glasses See Right Through You

Meta has released new AR glasses designed specifically for research purposes. These glasses are equipped with sensors that can measure the wearer's heart rate, eye movements, and other biometric data. The goal is to provide researchers with a platform to study how people interact with AR technology in real-world scenarios and to gather data that can be used to improve the design and functionality of future AR devices. While not intended for consumer use, these glasses represent a step forward in Meta's ongoing efforts to develop advanced AR hardware and understand its potential applications.

Alibaba's Wan 2.1: A New Rival to OpenAI's Sora

Alibaba Cloud's wan2.1 is an advanced AI video generation model that transforms text descriptions into high-quality videos featuring realistic movements. It supports both Chinese and English text inputs and uses technologies like VAE and DiT for superior video quality, achieving a VBench score of 84.7%. Users praise its ability to generate complex movements, making it suitable for various applications, including content creation and education. The platform offers easy integration with API access, detailed performance analytics, and scalability for enterprise needs.

Hand Picked Video

In this video, we’ll explore how Pikadditions by Pika Labs is revolutionizing video editing.

Top AI Products from this week

• MGX - MGX (MetaGPT X) is a multi-agent AI platform based on real software SOPs. Chat with the AI team leader, product manager, architect, engineer, and data analyst 24/7 to create websites, blogs, shops, analytics, games, or anything else you want to try.

• Teamble AI - Teamble AI helps give and get 10x better employee feedback throughout the year. Better 1 on 1's > better feedback > higher performance!

• Producta - If you are already using AI for coding, Producta will make things faster. Give it tickets directly in your Linear or Jira board. It will help you clarify them, and then open a pull request with the solution.

• Pika 2.2 with Pikaframes - Pika 2.2 is HERE, with 10s generations, 1080p resolution, and Pikaframes— key frame transitions anywhere from 1-10s. More transformation, more imagination

• Chikka - Curious about user feedback, employees’ idea, or friends’ opinions? Set up your own AI Voice Agent in 5 mins to engage warmly with real humans and ask great questions in natural conversations. After that, you are just one click away from actionable insights!

• Protoclone - Meet Protoclone: a faceless, anatomically precise android companion made for individuals and businesses alike. With 200+ degrees of freedom, 1,000+ Myofibers, and 500 sensors, it’s designed for seamless, human-like movement and interaction.

This week in AI

• Quantum-Resistant Chip - The QS7001 chip uses quantum-resistant encryption, reducing data transmission time to protect against potential quantum computing attacks.

• DeepSeek R2 AI Model - DeepSeek is fast-tracking its R2 AI model, initially set for May, aiming to improve coding and multilingual reasoning. Built cost-effectively, it challenges Western AI dominance.

• Murati's AI Startup - Ex-OpenAI CTO Mira Murati launched Thinking Machines Lab with OpenAI co-founder John Schulman. Focused on adaptable, capable AI systems, it aims to bridge gaps in AI understanding and customization.

• ElevenLabs Audiobook Platform - ElevenLabs now enables authors to create and publish audiobooks directly on its platform, streamlining production and distribution.

• Copilot's Free Access - Microsoft now offers free, unlimited access to Copilot's Voice and Think Deeper (powered by OpenAI's o1) features, enhancing user interaction and complex task handling.