In partnership with

Edition #29: Open AI releases its best model yet- Chat GPT o1

Date: 23-Sept-2024

Hey AI enthusiast,

Welcome to the latest edition of the One More Thing in AI newsletter! We are diving deep into Open AI’s Chat GPT o1 release in this edition.

There are some cool examples of folks using it!

Best regards,

Renjit Philip

In this edition:

• Open AI’s new model Chat GPT o1 is out!

• Examples of the Chat GPT o1 model at work and play …

• Make your voice heard!

• AI Jobs

• References

• Share and Subscribe

Open AI’s new model Chat GPT o1 is out!

OpenAI has developed a new AI model called o1. This model is exceptionally good at solving tough problems. For example:

- Programming Questions: On Codeforces, a site for competitive programming, o1 scores better than 89% of participants.

- Math Competitions: In the AIME, a math contest that helps select students for the USA Math Olympiad, o1 ranks among the top 500 students in the entire country.

- Science Tests: On a challenging test called GPQA, which covers physics, biology, and chemistry, o1 does better than human PhD students.

Even though o1 isn't as easy to use as some other models yet, OpenAI is releasing an early version called OpenAI o1-preview. You can try it out now in ChatGPT, and some trusted users can access it through an API.

OpenAI trained o1 using a method called reinforcement learning. This helps the model learn to think carefully through problems step by step, which we call a "chain of thought." What's cool is that this training method doesn't need a lot of data. They've noticed that o1 gets better when they:

- Spend more time training it (called train-time compute).
- Allow it more time to think when solving problems (called test-time compute).

The way they are improving and scaling this model is different from how they usually train large language models. They're still exploring the best ways to make it even better.

Why does it matter?

This is reported to be an AI model that is different in design from Chat Gpt4 and one that Open AI has been working on for a long time. Could this be the improvement on the transformer model that we have been waiting for?

This model demonstrates reasoning abilities at a PhD level- what is next? Boggles the mind!

Also, spare a thought for all those startups that were working on developing agents using the older LLMs. This model makes them obsolete in one fell swoop.

Examples of the Chat GPT o1 model at work and play (mostly!)

1. Full Weather iPhone App with Animations in less than 10 minutes

2. Calculator on a single page in less than 5 minutes

Offshore talent that come with impressive resumes

“She’s contributing far greater than an executive assistant. She’s also a project manager, podcast producer, copywriter, social media manager, and more!” - Chris Hutchins, All The Hacks

Known for their executive assistants, marketing, finance, and ops talent — Oceans is the global talent partner for brands like True Classic, Pattern Brands, and Othership.

Get full-time, highly experienced talent for only $3000/month.

Learn more about Oceans

3. AISteroid Game w/ SciFi vibes

4. Flappy Bird Game

5. Create Art with ChatGPT o1 and Illustrator Scripts

Make your voice heard!

I would love to know more about what you like and don’t like, specifically in connection to this newsletter. Please take this 2-minute survey to help me produce more content that is useful for you.

AI Jobs

References

• Sam Altman on o1

• Open AI Blog -Learning to reason with LLMs

That wraps up another power-packed edition of the Newsletter. Want the insider advantage? Subscribe to our email edition to access content before anyone else, and keep it forever in your inbox.

Also, you get exclusive content that only subscribers get access to. Discover the exclusive benefits of being a subscriber!

Share and Subscribe

Forward to a friend

If you want to interact with me on social media, follow me on LinkedIn and Twitter.