close_game
close_game

OpenAI's new o1 model can solve 83% of International Mathematics Olympiad problems

Sep 14, 2024 03:42 PM IST

OpenAI's new o1 model can be used for scientific research in physics, chemistry, and engineering, where rigorous reasoning and complex problem-solving are prime

OpenAI will launch o1, its latest artificial intelligence (AI) model in two weeks. This will mark the debut of a new class of reasoning AI models and comes amid the speculation of the release of "Strawberry" AI.

A symbol for the OpenAI virtual assistant on a smartphone, arranged in Riga, Latvia, on Friday, Aug. 16, 2024. (Andrey Rudakov/Bloomberg)
A symbol for the OpenAI virtual assistant on a smartphone, arranged in Riga, Latvia, on Friday, Aug. 16, 2024. (Andrey Rudakov/Bloomberg)

The AI company will also be releasing o1-mini, which is a lighter and more cost-effective version of the new model. This one will be ideal for tasks including coding and problem-solving.

Also Read: WhatsApp Business new verified badge and AI tools: All you need to know

What can o1 do?

o1 can solve intricate, multi step problems like math and coding types. It mimics human-like reasoning and explaining its thought process as well along the way.

It also promises improved accuracy and a big reduction in hallucinations (Which is when an AI model generates false or misleading information).

o1 can be used as a powerful tool for scientific research in physics, chemistry, and engineering, where rigorous reasoning and complex problem-solving are important.

How different is o1 from OpenAI's previous models?

o1 uses reinforcement learning, compared to the pattern-mimicking training method used by older AI models. Reinforcement learning is when a system learns through rewards or penalties.

Also Read: Income tax notices sent to TCS employees are a 'discrepancy,' ITRs will be reprocessed: Report

It also uses a "chain of thought" process, which mimics human cognition by breaking down problems into logical, sequential steps. OpenAI Chief Research Officer Bob McGrew says that o1 outshines its predecessors in math-related tasks.

o1 was able to tackle 83% of the problems in the International Mathematics Olympiad while GPT-4o in comparison could only solve 13% of problems correctly.

When will o1 be available?

All ChatGPT Plus and Team users already will have access to o1 preview and o1 mini, while Enterprise and Edu users will get access next week.

OpenAi plans to make o1-mini available to free users as well, but no date has been announced for this.

o1-preview costs $15 per million input tokens and $60 per million output tokens for developers wanting to integrate it into their applications. This is thrice the price of GPT-4o

What are the pitfalls of o1?

o1 is more expensive and slower to use than its predecessors, while also not being optimised for web browsing and processing files and images.

What can o1 possibly be a step towards to in the future?

For OpenAI, o1 is a step towards a future where AI can be an autonomous agent capable of making decisions, taking action on the behalf of users, and solving real-world problems, revolutionising various industries from healthcare to engineering.

Also Read: India hikes edible oil import tax on palm, soyabean, and sunflower oil to support farmers

Stay updated with the...
See more
Stay updated with the latest Business News on Petrol Price, Gold Rate, Income Tax Calculator along with Breaking News Events and Latest News Updates on Hindustan Times.
SHARE THIS ARTICLE ON
SHARE
Story Saved
Live Score
OPEN APP
Saved Articles
Following
My Reads
Sign out
New Delhi 0C
Wednesday, October 09, 2024
Start 14 Days Free Trial Subscribe Now
Follow Us On