OpenAI's secret "project strawberry"

While awake, the human brain generates enough electrical power to light a small light bulb.

Love or hate this intro? Hit reply and tell us why!

What’s in store:

• OpenAI has a new project called “Strawberry”.

• Rabbit R1 makes the headlines … again!

• Around the web: Conspiracy theories about Trump shooting trend on X, promoted by Elon Musk.

• Master AI with Zero to Mastery.

Read Time: 6 minutes

OPENAI

OpenAI's secret project "Strawberry"

OpenAI are working on a new AI initiative called “Strawberry”.

This project aims to upgrade the reasoning capabilities of AI models, pushing beyond current limitations.

Details of Strawberry remain mostly under wraps.

However, it is designed to allow AI to not just generate answers but to autonomously navigate the internet for in-depth research.

This marks a significant step towards AI models capable of advanced planning and problem-solving.

Key points from an internal document reveal that Strawberry focuses on “post-training” techniques, refining AI models after initial training on large datasets.

This method could potentially enable AI to tackle complex tasks over extended periods, something that has proven challenging for existing models.

Here’s a snippet about the project:

• Strawberry aims to improve AI’s ability to solve multi-step problems.

• The project uses specialised post-training techniques to upgrade model performance.

• Focus on enabling AI to handle complex, extended tasks autonomously.

Can it make AI smarter?

How it works is a closely guarded secret within OpenAI.

The project uses models that aim to enable AI to perform deep research by navigating the web reliably.

This is something current AI models have struggled with, as they often fail with common sense problems and logical fallacies.

OpenAI’s goal is to significantly improve the reasoning capabilities of its AI models.

This involves a specialised way of processing the AI model after its initial training.

Researchers believe that upgrading reasoning is key to achieving human or super-human intelligence in AI.

OpenAI CEO Sam Altman has stated that progress in AI will focus on reasoning ability.

Other tech giants like Google, Meta, and Microsoft are also exploring ways to improve AI reasoning.

There is debate among researchers about whether large language models can truly incorporate long-term planning and humanlike reasoning.

Strawberry is part of OpenAI’s plan to overcome these challenges.

The project involves fine-tuning AI models after their initial training to hone their performance.

This method is similar to the "Self-Taught Reasoner" (STaR) developed at Stanford, which aims to elevate AI intelligence by iteratively creating its own training data.

The goal is to have AI models conduct autonomous research and potentially even take on tasks typically done by software and machine learning engineers.

Can it beat me at tic-tac-toe?

Vote for live results and see results + opinions from yesterday at the bottom of the email.

AI COURSES

Step-By-Step to a $100,000+ / Year Career Using AI (in 2024!)

Zero To Mastery has taught tech skills to over 1,000,000. Graduates have gotten hired at companies like Tesla, NVIDIA, Apple, Google, NASA, and many more.

And their recently launched AI courses take you step-by-step from complete beginner to AI Developer that can get hired this year.

• Learn To Code + Master Prompt Engineering

• Build AI Apps + Work With Large Language Models

• Complete Career Toolkit to Land Interviews & Get Hired

But the best part? You won't be learning alone.

You'll get to join the private ZTM Discord with 1,000s of other students, alumni, mentors, and even your instructors. It's the most active and highly-rated discord of any online learning platform (Google it if you don't believe us 😉).

ZTM is offering Mindstream readers an exclusive 7-day free trial so that you have no excuses to try it!

And if you sign up before July 18th, you'll get an extra 15% off an annual subscription (that's less than $1/day!).

So make today the day you take a chance on YOU. Join thousands of students learning 2024's most in-demand skill (it'll be the best investment you can make).

On This Day

July 15th 2006 - Twitter, then known as Twttr, now known as X, is launched, becoming one of the largest social media platforms globally.

Trending Tools

Wondershare Virbo: This AI avatar generator allows users to create spokesperson videos in minutes, is available on multiple platforms and supports over 120 languages.

Vocol AI: A voice collaboration platform powered by GPT technology, Vocol AI can quickly convert speech into text with multi-language translation.

B2B Rocket AI Agents: This tool automates B2B sales processes and lead generation, aiming to boost revenue by creating a digital workforce that operates on autopilot.

Emma: Build a custom AI assistant using ChatGPT. Emma allows you to connect to your organisation’s resources or upload files, helping with various inquiries.

venturefy: Utilises AI to codify and verify corporate relationships, helping businesses to identify new opportunities and build trust quickly.

RABBIT R1

Rabbit R1 stored all chats without option to delete

Rabbit has released a security update for its R1 AI assistant to fix a major flaw: previously, users couldn’t erase chat logs stored on the device.

The new update includes a Factory Reset option to wipe all user data, which wasn’t possible before.

The update also addresses another issue where the R1’s pairing data could see journals - the usage logs of the device.

This has now been fixed, so pairing data can no longer read or log journal entries.

The company assures that no abuse of this data has been detected.

What you should know:

• The new update allows full deletion of user data with Factory Reset.

• Fix prevents pairing data from accessing user journals.

• Rabbit promises improved security measures following an internal API key leak.

From flaws to fixes

Rabbit is also working on improving its security practices after an internal leak of API keys by an employee, who has since been terminated and is under investigation.

The company is reviewing its logging practices to prevent future issues.

We were as surprised as you to see R1 still making headlines.

Mindstream Picks

Valve, known for keeping operations secret, has 181 out of 336 staff focused on game development, as revealed by antitrust lawsuit documents.

SpaceX can’t recover 20 Starlink satellites left in very low orbit after a Falcon 9 upper stage malfunction on July 11.

Cryptocurrency tokens related to former US President Trump surged by double digits after he survived an assassination attempt, with MAGA up 26% and TRUMP2024 up to 40%.

Intel and Alphabet stocks are top picks for 2024, with Intel up 13% and Alphabet up 10%, driven by new chip releases and strong AI investments.

Don’t Miss - Conspiracy theories about the Trump shooting surfaced on X, with topics like “#falseflag” trending. Elon Musk, who supports “Free speech” on social media, posted his endorsement of Trump following the incident. Other platforms focused on verified news over conspiracies.

AI Art

Our Image of the Day

Daily Image Prompt

Submit your artwork to Mindstream →

Yesterday’s Poll

“Would you use a robot ice cream parlour?”

Yes, my childhood dream has come true! - 52% ✅

No, I'd avoid it. - 48%

Your Views:

“It will be a completely renewed tradition of childhood and adulthood!” - mahflowlife

“When taking an ice cream here in Italy, we exchange a couple of jokes and smiles for the day, a good recommendation on which flavour I should try that day and this makes my day. Isn't what you're suggesting a danger to humanity? People will die without normal human interaction.” - svetlanayan

Submit your opinions in our daily poll to be featured!