- The Rundown AI
- Posts
- OpenAI's new AI code named 'Strawberry'
OpenAI's new AI code named 'Strawberry'
PLUS: Mysterious AI models in the LMSYS arena
Rowan Cheung
July 15, 2024
Sign Up | Advertise | Tools | AI University
Welcome, AI enthusiasts.
The world has been anxiously waiting for OpenAI’s next move — and after a period of quiet, the rumor mill just rumbled back to life.
With a new Q* rebrand, a gpt-mini model, and more whistleblower complaints, the best reality show in AI may be ramping up for its new season. Let’s get into it…
In today’s AI rundown:
OpenAI’s Q* gets a ‘Strawberry’ evolution
Mysterious AI models appear in LMSYS arena
Turn any text into an interactive learning game
Whistleblowers file new OpenAI complaint
5 new AI tools & 4 new AI jobs
More AI & tech news
Read time: 4 minutes
LATEST DEVELOPMENTS
OPENAI
Image source: Midjourney
The Rundown: OpenAI is reportedly developing a secretive new AI model codenamed ‘Strawberry’ (formerly Q*), designed to dramatically improve AI reasoning capabilities and enable autonomous internet research.
The details:
Strawberry is an evolution of OpenAI's previously rumored Q* project, which was touted as a significant breakthrough in AI capabilities.
Q* had reportedly sparked internal concerns and was rumored to have contributed to Sam Altman's brief firing in November 2023 (what Ilya saw).
The new model aims to navigate the internet autonomously to conduct what OpenAI calls "deep research."
The exact workings of Strawberry remain a closely guarded secret, even within OpenAI — with no clear timeline for when it might become publicly available.
Why it matters: The Internet has been waiting for new OpenAI activity as competitors catch up to GPT-4o — and after a bit of a lull, the rumor mill is churning again. With Strawberry, an AGI tier list, new models in the arena, and internal displays of human-reasoning capabilities, the AI giant may soon be ready for its next major move.
TOGETHER WITH HUMAN X
The Rundown: Sick of boring AI conferences? From the team behind Money20/20, Shoptalk, and HLTH comes the most consequential gathering in AI — HumanX.
Attend HumanX on March 10-13, 2025 at the Fountainbleu in Las Vegas to:
Gain practical knowledge and actionable strategies to implement AI in your work
Hear from the brightest AI minds in the industry with leaders from Snowflake, Groq, Salesforce, Applied Intuition and more
Explore the latest advancements and real-world applications of AI that are transforming your industry
Apply for a pass now — use the discount code “HX25n_therundown” for $150 off this week!
LMSYS
Image source: Lmsys screenshot
The Rundown: Three mysterious new models have appeared in the LMSYS Chatbot Arena — with ‘upcoming-gpt-mini’, ‘column-u’, and ‘column-r’ available to test randomly against other language models.
The details:
The new models are available in the LMSYS Chatbot Arena’s ‘battle’ section, which puts anonymous models against each other to gauge outputs via user vote.
The ‘upcoming-gpt-mini’ model identifies itself as ChatGPT and lists its creator as OpenAI, while column-u and column-r refuse to reveal any identifying details.
OpenAI has previously tested unreleased models in LMSYS, with ‘im-a-good-gp2-chatbot’ and ‘im-also-a-good-gpt2-chatbot’ appearing prior to GPT-4o’s launch.
Why it matters: Does OpenAI have a small, potentially on-device model coming? The last time we saw mysterious LLMs appear in the Battle arena was before the company’s last major model release — and if the names are any indication, we could have a new mini-GPT in the very near future.
AI TRAINING
The Rundown: Claude 3.5 Sonnet’s new Artifacts feature lets you transform any text or paper into an engaging, interactive learning quiz game to help with practicing for exams, employee onboarding, training, and so much more.
Step-by-step:
Head over to Claude AI.
Choose and copy the text you want to turn into a learning game.
Paste the text into Claude 3.5 Sonnet and ask it to create an interactive learning game in the form of a quiz with explanations.
Review the generated game and ask Claude to make any necessary adjustments.
Note: We ran a one-hour workshop covering various use cases with Artifacts, which you can access in The Rundown’s AI University.
PRESENTED BY CODESIGNAL
The Rundown: CodeSignal just released a new learning path designed to help you master advanced skills in Natural Language Processing (NLP) —the foundation of Large Language Models like ChatGPT.
In this learning path, you’ll learn how to:
Transform text data into features ML models can understand
Build text classification algorithms
Train and optimize NLP models
OPENAI
Image source: The Washington Post
The Rundown: Whistleblowers just filed a complaint with the SEC alleging that OpenAI used overly restrictive non-disclosure agreements to prevent employees from reporting concerns to regulators, violating federal whistleblower protections.
The details:
The agreements allegedly prohibited employees from communicating securities violations to the SEC, also requiring them to waive rights to whistleblower incentives.
The complaint also claims OpenAI's NDAs violated laws by forcing employees to sign these restrictive contracts to obtain employment or severance.
OpenAI CEO Sam Altman previously apologized for exit agreements that could strip former employees of vested equity for violating NDAs.
OpenAI said in a statement that the company’s whistleblower policy “protects employees’ rights to make protected disclosures.”
Why it matters: We just detailed how OpenAI’s busy week may be hinting at some major new moves… But will these skeletons in the closet spoil the party? This isn’t the first group to blow the whistle on internal issues, and while Altman and OpenAI have said changes have been made — it apparently hasn’t been enough.
NEW TOOLS & JOBS
💃 MOVE AI Move Live - AI-powered motion capture for real-time 3D experiences
🔍 Telescope- AI-driven lead generation for targeted B2B lists
🤖 Threado AI - Automated AI support for customer-facing teams
💸 Centra - Streamlined payroll and benefits for startups
🔨 Tegon - Open-sourced issue tracking tool for engineering teams
🕸️ Mistral AI - Web Crawling & Indexing Engineer
📦 Databricks - Principal Delivery Manage
📊 Waymo - Senior Product Manager, Simulation and Evaluation
💻 Captions - Software Engineer - Backend
QUICK HITS
Meta’s Llama-3 405B model is set to release on July 23 and will be multimodal, according to a new report from The Information.
Amazon announced expanded access to its Rufus AI-powered shopping assistant for all U.S. customers, offering personalized product recommendations and enhanced responses to shopping queries.
Samsung revealed plans to release an upgraded version of the Bixby voice assistant later this year powered by the company’s own LLM, as part of a broader push to integrate AI across its device lineup.
HR software unicorn Lattice (founded by Sam Altman’s brother Jack) has backtracked on a controversial plan to give AI ‘workers’ employee status, following intense criticism from employees and tech leaders.
Japanese investment giant Softbank acquired struggling British AI chipmaking firm GraphCore, hoping to revitalize the former Nvidia rival and bolster its AI hardware portfolio.
U.S. Rep. Jennifer Wexton debuted an AI-generated version of her voice, allowing her to continue addressing Congress despite speech limitations caused by a rare neurological condition.
THAT’S A WRAP
SPONSOR US
Get your product in front of over 600k+ AI enthusiasts
Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.
FEEDBACK
How would you rate today's newsletter?Vote below to help us improve the newsletter for you. |
If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.
Reply