AI Newsletter: Open AI loves strawberry; Is Google AI reading your drive PDFs?
Plus: A new AI chip claims to be 20 times faster than Nvidia
Hey! Here goes the latest next, big and what from the world of AI.
#BIG idea of the day
AI Health Coaching: Here is what Sam Altman and Arianna Huffington are up to
Thrive AI Health, a new company by Sam Altman and Arianna Huffington, aims to create a personalized AI health coach. This AI coach would analyze a user's health data (biometric and behavioral) and offer recommendations for improving health habits.
The Thrive AI Health Coach will be trained on the latest peer-reviewed science, biometric, lab and other medical data, as well as users' personal preferences and goals around the five key daily behaviors.
The envisioned platform could be integrated into various existing apps or platforms (workplace tools, calendars etc.).
Read the rest on why this is a big deal and a new path for Gen AI apps.
#NEXT: AI News
- Google's Gemini AI was caught scanning PDFs in Google Drive without permission — and you can’t even disable it!
- Apple, Nvidia, and Anthropic used thousands of swiped YouTube videos (from Mr Beast, PewDiePie, Khan Academy, MIT, Harvard, The Wall Street Journal, NPR etc) using their subtitles to train AI, according to an investigation by Proof News and Wired, raising concerns about potential copyright infringement.
- Finally, Android users get the Claude App! Anthropic is bringing its AI Chatbot, the primary competitor to OpenAI’s ChatGPT, to over three billion devices.
- Andrej Karpathy, former head of AI at Tesla and researcher at OpenAI, has launched a new “AI Native” startup called Eureka Labs, which will use AI to assist in education.
- A new AI Chip “Sohu” is reportedly 20 times faster at running transformers like ChatGPT than Nvidia's flagship H100, claim its Harvard dropout founders.
- A new AI model, SpreadsheetLLM, by Microsoft researchers, tackles the challenge of using AI with spreadsheets, a major step forward in enterprise AI.
#NEXT: AI products
- Blaze : Use AI to find high-intent leads + trigger relevant outreach
- Jobright AI : Your AI job search copilot
- AutoReels.AI : Automate faceless videos for TikTok, YTB & more
- Riffo.AI: Say goodbye to your messy filenames
#WHAT: OpenAI’s Project Strawberry
OpenAI’s Project Strawberry might unlock human-like reasoning in AI models. OpenAI is developing this secretive model to enable AI to plan, navigate the internet autonomously, and conduct deep research. This is part of the recently published five levels(see below) of AI that Open AI sees in the near future, with human like reasoning being at Level 2.
Sam Altman sees reasoning ability as essential for AI progress, and Strawberry is key to overcoming current AI limitations.
GPT-Mini
Interestingly, Open AI is developing something called “GPT-Mini,” which is available in LMSYS Chatbot Arena’s ‘battle’ section. This section is a website where anonymous models compete and are evaluated by user votes.
- The ‘upcoming-gpt-mini’ model is identified as ChatGPT by OpenAI
- If you remember, OpenAI tested unreleased models like ‘im-a-good-gp2-chatbot’ at the same arena before GPT-4o’s launch.
- The appearance of these mysterious models hints at a new mini-GPT soon, maybe a native LLM for Mobile devices.
Quote of the day
