OpenAI’s high-order inference model has arrived.
OpenAI clarifies: No, “GPT Next” is not a new model.
Originally dubbed “Project Strawberry,” the model was released to the public on Thursday under the name “o1.” As expected, o1 was able to “think” (process) before answering the prompt, unlike previous large language models that used predictive text to form responses. “We have developed a new family of artificial intelligence models designed to take more time to think before reacting,” OpenAI said in a statement.
Tweet may have been deleted
Tweet may have been deleted
In a demo shared on the OpenAI website, the video shows a user asking o1 to “write a grammar question” in ChatGPT [sic] Correct sentences without using any letters more than once. bed”.
In another video, research leader Jerry Tworek asks o1 to solve a logic puzzle about the ages of princes and princesses. The model decodes the problem and answers the age equation step by step by understanding the different variables and determining which equations are needed to solve the problem.
Mix and match speed of light
As the demo shows, o1 is best suited for solving complex mathematical problems in science, physics, and coding. The announcement states: “Healthcare researchers can use o1 to annotate cell sequencing data, physicists can use o1 to generate the complex mathematical formulas required for quantum optics, and developers in all fields can use o1 to build and execute multi-step work. process.
This also means that the o1 preview version has certain limitations. It cannot browse the web or receive uploaded files and images. OpenAI says GPT-4o remains the best model for this type of task.
In the safety section, OpenAI said that under previous authorizations from President Biden’s Artificial Intelligence Executive Order, U.S. and British security institutes have been allowed to conduct early o1 testing.
OpenAI’s o1 model can be previewed by Developer AI and ChatGPT Plus users. When you go to the ChatGPT application, you can select o1 in the drop-down menu.
