End of Privacy?

2 Weeks back, Jan Leike, the head of AI Safety team at OpenAI, left the company and essentially OpenAI’s “Superalignment team” to join Anthropic, citing OpenAI deprioritised Safety. A lot of looming fears after this one.

But first what is “superalignment”?

It’s the process of aligning the AGI with human values so they don’t go rogue on us and end humanity. We’ve covered this and a lot of other concepts in our AGI Course.

Anyway, earlier this week, a retired U.S. Army general and former NSA Chief joins OpenAI's AI Safety team (and OpenAI’s board) focusing on utilizing artificial intelligence to improve cybersecurity by quickly identifying and addressing cyber threats.

Elon Musk said this was the beginning of invasion of privacy (even phone calls). The other side of the spectrum says there will finally be proper monitoring and regulatory enviornment at OpenAI.

Only time will tell.

More this week:

• Google announces a large language model for Personal Health devices

• Nvidia trained a model on Synthetic data

Is this the end or beginning of privacy as we know it.

The world feels it’s the former. Let’s dive in.

Stuff you should know

OpenAI’s safety team beefed!

Paul M. Nakasone, a retired US Army general and former director of the National Security Agency (NSA), has been appointed to OpenAI's board of directors. He will join the company's Safety and Security Committee, led by CEO Sam Altman. Nakasone was nominated to head the NSA by former President Donald Trump and served in the role from 2018 until his departure in February 2024. Prior to his departure, he advocated for the renewal of Section 702 of the Foreign Intelligence Surveillance Act, a surveillance program that was eventually reauthorized by Congress in April.

OpenAI stated that Nakasone will contribute to the company's safety and security initiatives, focusing on utilizing artificial intelligence to improve cybersecurity by quickly identifying and addressing cyber threats. Recent departures from OpenAI, including co-founder and chief scientist Ilya Sutskever and researcher Jan Leike, have raised concerns about the company's safety culture and processes. Nakasone's extensive experience in cybersecurity will be invaluable in helping OpenAI achieve its mission of ensuring artificial general intelligence benefits all of humanity.

Nvidia trained a model on Synthetic data?

Nvidia has released the Nemotron-4 340B model family, a groundbreaking family of open models that redefine synthetic data generation for training large language models (LLMs). This development marks a significant milestone in the AI industry, empowering businesses across various sectors to create powerful, domain-specific LLMs without the need for extensive and costly real-world datasets.

The Nemotron-4 340B family includes base, instruct, and reward models, which form a comprehensive pipeline for generating high-quality synthetic data. With a significant 9 trillion tokens used in training, a 4,000 context window, and support for over 50 natural languages and 40 programming languages, Nemotron-4 340B outshines its competitors, including Mistral’s Mixtral-8x22B, Anthropic’s Claude-Sonnet, Meta’s Llama3-70B, Qwen-2, and even rivals the performance of GPT-4.

Google launches an LLM for Personal Devices!

Google researchers have developed a novel large language model that aims to understand and reason about personal health data from wearable devices. The model, called PH-LLM, is designed to provide personalized insights and recommendations to users by contextualizing physiological data like sleep patterns and physical activity.

To evaluate PH-LLM, the researchers curated three benchmark datasets to test the model's expert domain knowledge, alignment with patient-reported outcomes, and ability to generate high-quality recommendations. The results show the model performs well, with performance on par with human experts in some areas. The researchers believe this work is a critical step towards developing truly personalized health assistants powered by AI.

Hand Picked Video

In this video, we'll look at why Google's Gemini will be the best LLM this year. Google's Gemini is poised to be the leading Large Language Model (LLM), due to its unparalleled context window size, versatility, multimodal capabilities, and commitment to innovation and safety.

Top AI Products from this week 

• AGI Course - An Introduction to AGI and the future of work.

• SubmitAITool - Submit AI Tool Discover the best AI tools. Submit an AI tool if you have a unique AI tool or want to share your favorite AI product with others.

• 1000 Happy Dads - We created a customized AI-assistant that lets you share memories (over voice or text) and writes a heartfelt letter + a fun shareable graphic that will automatically be sent to you over email.

• inncivio - inncivio is a learning platform, with personalized content to upskill teams. Unique curriculums are created using AI, on the go, thanks to the company’s own knowledge base.

• Drip - Drip is a self-reflection and journaling tool powered by AI. It helps you get in touch with your inner thoughts and feelings, and then helps you turn those thoughts and feelings into action.

• Inbox Zero AI - Apple Engineer Shayan Arman redesigns Email with a new way to manage your inbox.

• Epipheo AI - Epipheo AI is a powerful tool to create engaging videos. It uses Gen AI to generate dynamic visuals, compelling scripts, and professional voiceovers.

• Wunjo - Experience Wunjo 2.0: AI-powered face swap, lip sync, content restyling, remove objects, text and background.

This week in AI

• Microsoft's AI Chief Audits OpenAI's Code - Microsoft's AI chief audits OpenAI's code, highlighting a complex and competitive dynamic between the two tech giants. Microsoft's investment in OpenAI gives it a strategic advantage, and Suleyman's presence allows Microsoft to audit OpenAI's code, potentially leading to a shift in the AI landscape.

• Dream Machine text To Video Model - Luma AI's Dream Machine is a revolutionary text-to-video AI tool that has taken the world of video content creation by storm. This groundbreaking technology enables users to generate high-quality, realistic video clips from simple text prompts in just minutes, opening up new possibilities for rapid and accessible video content creation.

• Improved Version of GEN 3 - Runway has released Gen-3 Alpha, a new video generation AI model trained on a new infrastructure for large-scale multimodal training. This model is capable of generating highly detailed videos with complex scene changes, a wide range of cinematic options, and detailed art direction capabilities.

• Jace the AI Agent - Zeta Labs, a London-based startup founded by former Meta engineers, has launched Jace, an AI agent that can execute in-browser actions on command without user guidance. Jace is powered by a large language model (LLM) and can handle a wide range of tasks, from basic searches to complex tasks like setting up a recruitment pipeline on LinkedIn.