• Superhuman AI
  • Posts
  • Google and OpenAI share major research breakthroughs

Google and OpenAI share major research breakthroughs

ALSO: Threads comes to Europe, and Grime's AI toys

Read time: 2.5 minutes

Welcome back, Superhuman

It’s a day of major breakthroughs in AI research, as Google publishes a paper proving that AI can discover new solutions to longstanding problems. OpenAI also shared 2 key pieces of research on AI safety.

But it’s not all theory for today’s email. You’ll also learn how to generate music with Google’s newly-released AI music creation tool.

TODAY’S MENU

  • Google’s new research breakthrough extends the capabilities of AI models

  • How to generate music with simple prompts

  • OpenAI publishes 2 major papers on AI safety, identifies 7 key principles

  • Friday Laughs: ChatGPT in French

  • 5 new AI tools to boost your productivity

  • AI Generated Images: Famous movies misspelled

NEWS

Today in AI & Tech

Image Source: Reuters

  • At Long Last: Meta finally launches Threads in Europe.

  • Chipping In: Intel releases a host of new AI-focused computing chips.

  • DJ GPT: Spotify is reportedly testing AI-powered music playlists.

  • Grokking: Musician Grimes launches AI-powered toys.

RESEARCH

Google achieves major breakthrough by demonstrating AI models can discover new solutions

Generating emails and blog posts is cool, but finding answers to decades old pure math problems is a whole new frontier for AI models. A new piece of research from Google published yesterday did just that, expanding our understanding of what AI models are capable of.

The researchers developed a new method called FunSearch (short for searching in the function space) and applied it to the cap set problem — a longstanding problem in extremal combinatorics math — to discover new solutions to the problem that go “beyond the best known ones,” according to the authors.

If true, not only is this the first time that an AI model has been used to discover new solutions to open problems in science and math, the authors also demonstrated real-world application for this finding by applying it to the “bin-packing” problem, which can be used for applications like creating more efficient data centers and packing transportation trucks more efficiently.

You can find the research paper here.

TOGETHER WITH ZUPYAK

Use AI to create content that ranks on search engines and brings you customers in no time 🚀

Zupyak, the easiest SEO AI writer you will ever use, just made their product even better. You can now create SEO optimized blog posts with AI images in 30 seconds! Cheaper and faster than using freelancers and no SEO knowledge needed.

AI AT WORK

How to create your own music with AI in less than 5 minutes

Source: Google

Looking for a quick weekend project to pick up new AI skills? Google just released MusicFX, a new AI tool that can generate music with just a few words.

Here’s how you can create music with MusicFX in less than 5 minutes:

  • Go to MusicFX and sign up with your Google account. Currently available in the US.

  • Now, simply type in your prompt in the prompt box and click the ‘Generate’ button.

  • You can also download the track by clicking on the download button at the bottom.

AI SAFETY

OpenAI publishes 2 pieces of research on AI safety, identifies 7 key practices

The possibility of AI posing a threat to humanity is no longer the stuff of science fiction. Several major figures from Elon Musk to OpenAI CEO Sam Altman have been vocal on AI safety over the last year. And now the company behind ChatGPT and GPT-4 has published 2 new papers on AI safety.

The first presents a new approach to aligning AIs with human interests. The key challenge that they identify is that humans will have to supervise AI systems much smarter than them in the future.

The paper finds that naive human supervision could scale poorly to superhuman models without further work. However, the paper argues that weaker AI models could be used to control more capable models.

The second paper outlines 7 key practices for governing AI systems that include:

  • Evaluating the AI system’s suitability for the task

  • Constraining the AI by requiring human approval

  • Setting default behaviors for the system

  • Making the AI’s activities and thought process more legible

  • Implementing automatic monitoring

  • Create reliable attribution

  • Creating the ability to turn off the AI system

You can find the first paper here and the second paper here.

5 AI Tools to Supercharge Your Productivity

Scade AI: Develop products and services, streamline your business processes, marketing, sales and finance with AI – all effortlessly.

Snapby: Generates super-realistic images instantly by using simple prompts.

Section*: Set your AI business strategy with Section’s free library of AI and business frameworks, designed by leaders like Scott Galloway. Get free resources →

Viable: Turn user insights into actionable strategies effortlessly with Viable’s generative analysis.

TubeOnAI: Summarize and listen to any podcast in 30 seconds with AI.

* indicates a sponsored tool

FRIDAY LAUGHS

ChatGPT in French

AI-GENERATED IMAGES

Famous movies misspelled

Source: @Movie_Poster_Prompter on Midjourney

ADVERTISE WITH US

Acquire new customers and drive revenue by partnering with us

Superhuman is the world’s biggest AI newsletter for businesses and professionals with 500,000+ readers working at the world’s leading startups and enterprises. Companies like Amazon, Calendly, and Masterworks feature their products in Superhuman. Main ads are typically sold out 4 weeks in advance. You can book future ad spots here.

🧞 Your wish is my command

What did you think of today's email?

Your feedback helps me create better emails for you!

Login or Subscribe to participate in polls.

Reviews of the day

Thanks for reading.

Until next time!

p.s. if you want to sign up for this newsletter or share it with a friend or colleague, you can find us here