OpenAI o1 is the first ‘reasoning’ ChatGPT model

AI companies aim to develop artificial intelligence that can think like humans. OpenAI o1 takes humanity closer to that goal by becoming the first “reasoning” AI model. 

This model can handle more complex tasks than the current flagship GPT-4o model. These include problems involving STEM subjects like physics, chemistry, and biology. 

READ: Meta and OpenAI to launch AI models with ‘reasoning’ skills

More importantly, it can recognize its mistakes and improve its responses to challenging situations. ChatGPT Plus subscribers may access OpenAI o1 via the model selector. 

How does the OpenAI o1 model work?

If you’ve been following AI trends, you’d be glad to know that o1 is the rumored Strawberry model in development. However, The Verge says OpenAI doesn’t provide clear details regarding its creation.

OpenAI research lead Jerry Tworek says o1 “has been trained using a completely new optimization algorithm and a new training datasheet specifically tailored for it.” 

Unlike previous models, OpenAI taught the latest model to solve problems using reinforcement learning, which teaches the system through rewards and penalties.

This technique seems to show that the AI model is becoming more “human” as reinforcement learning is similar to Reinforcement Theory.

Simply Psychology says psychologist BF Skinner (Burrhus Frederic Skinner) developed the theory, which involves shaping behavior through consequences. 

OpenAI o1’s “chain of thought” process further proves its improving human-like capabilities. It enables the AI to go through problems step-by-step like humans. 

The new model sets itself apart from GPT-4o by better solving complex problems like math. OpenAI chief research officer Bob McGrew told The Verge:

“The model is definitely better at solving the AP math test than I am, and I was a math minor in college.” 

Tworek adds, “There are ways in which it feels more human than prior models.” The model has a limited time to process queries, so it may say something like, “Oh I’m running out of time, let me get to an answer quickly.” 

OpenAI o1 also has a smaller version, o1-mini, which is a faster, cheaper reasoning model suited for coding. Interesting Engineering says it is ideal for applications that require reasoning without broad-world knowledge. 

You may access the o1 and o1-mini by subscribing to ChatGPT Plus for $20 monthly. Then, log in and select them via the model selector.