OpenAI's project Strawberry leaked

Company researchers believe they are close to AI with human-level reasoning

Welcome back!

Today is all about OpenAI (plus a quick announcement from Amazon). The company is continuing to push the boundaries of AI innovation despite facing an onslaught of legal troubles. Let’s go.      

In today’s Daily Update:

  • 🗞️ OpenAI is working on new reasoning tech code-named ‘Strawberry’        

  • 🤖 Amazon rolls out AI shopping assistant to all US users       

  • 📸 Whistleblowers accuse OpenAI of ‘illegally restrictive’ NDAs     

  • 🚨 AI Roundup: Four quick hits

Read time: 2.5 minutes

TOP STORY

🗞️ OpenAI is working on new reasoning tech code-named ‘Strawberry’

DALL-E 3

OpenAI researchers believe the company is nearing the completion of an AI system that can perform human-level reasoning. According to internal documents acquired by Reuters in May, the project is code-named “Strawberry.”

What you should know:

  • Strawberry is described as a system that enables AI models to plan ahead and navigate the internet autonomously. OpenAI refers to this as “deep research.”

  • The project could be related to the development of an AI search engine designed to compete with Google’s core product. 

  • Previously known as Q, project Strawberry reportedly sparked internal safety concerns that led to CEO Sam Altman’s brief firing last year.  

  • How Strawberry works is a tightly kept secret within OpenAI. There is no clear timeline for when it may become publicly available. 

Why it matters: A new study by researchers from MIT and Boston University revealed that leading language models like GPT-4 rely primarily on memorization rather than reasoning. Strawberry is reportedly similar to a method developed by Stanford researchers in 2022 called “STaR,” which enables AI models to reach higher intelligence levels by creating their own training data. This approach could be key to building language models that transcend human intelligence. 

AI TOOL OF THE DAY

🤖 Amazon rolls out AI shopping assistant to all US users

Source: Amazon

Amazon is rolling out its Rufus shopping assistant to all U.S. customers in the Amazon mobile app. 

What it does:

  • Rufus answers questions about products by drawing information from Amazon’s product listing details, reviews and the web. 

  • Ex. “Is this coffee maker easy to clean and maintain?” or “is this mascara a clean beauty product?”

  • Amazon says Rufus helps customers find reviews, get product recommendations and quickly compare options. 

  • The chatbot is also able to track customers’ orders. 

Why it matters: Rufus is quietly asserting itself as a strong addition to Amazon’s mobile app. It’s kind of like being able to talk to a knowledgeable employee in a real store, although Rufus is available 24/7 and knows everything about all of Amazon’s products. This should help Amazon customers quickly make informed purchases on a platform that features hundreds of millions of products. 

BUSINESS SPOTLIGHT

📸 Whistleblowers accuse OpenAI of ‘illegally restrictive’ NDAs

DALL-E 3

Anonymous whistleblowers have accused OpenAI of placing illegal restrictions on employees’ non-disclosure agreements in a letter to Securities and Exchange Commission (SEC) chair Gary Gensler.  

The details:

  • Lawyers representing the whistleblowers filed separate complaints asking the SEC to investigate OpenAI’s severance and NDAs. 

  • They say the company’s NDAs “prohibited and discouraged” employees and investors from communicating security violations with the SEC.

  • OpenAI’s employee exit agreement is also being criticized for stripping former employees of their vested equity if they fail to comply with the agreements. 

  • Last month OpenAI CEO Sam Altman apologized for his company’s exit policy and claimed it is already in the process of fixing its practices.

Why it matters: Given the wide range of security concerns shared by former OpenAI employees in recent months, there’s probably merit to this lawsuit. Transparency with regulators and the public is key to the safe development of advanced AI, especially with projects like Strawberry looming in the near future.

MORE TRENDING NEWS

 🚨 AI Roundup: Four quick hits

DALL-E 3

  • Apple stock hits record high as analysts predict $5 trillion valuation amid AI iPhone hype. 

  • Google launches new AI-powered video creation app called “Google Vids.”

  • YouTube Music is testing an AI-generated radio feature and a song recognition tool. 

  • The European Union officially publishes the full text of its AI Act. The new regulations will come into law on August 1. 

THAT’S A WRAP

Want to continue the conversation? Connect with me on LinkedIn and I’m happy to discuss any of today’s news. Thanks for reading The Daily Update!

(P.S. If you want to share this newsletter with a friend or colleague you can find it here.)