OpenAI has introduced a new AI model called o1 under the secret initiative ‘Project Strawberry’. This model is designed to handle more complex reasoning, especially in fields like science, coding, and mathematics. It’s part of OpenAI’s efforts to develop AI systems that can think more deeply and solve tougher problems.
Model Development
The o1 model is built to process questions in a way that mimics human thinking. Its key abilities include:
- Looking at problems from different perspectives.
- Learning from mistakes, reviewing its answers to improve them.
During early tests, o1 showed impressive performance, solving complex problems in physics, chemistry, and biology at a level comparable to a PhD student. It excelled in math and coding, successfully answering 83% of the questions in a tough math competition, outperforming older AI models.
Cost-Effective Option: o1-Mini
OpenAI also launched a more affordable version called o1-Mini, which is 80% cheaper than the full o1 model. Despite the lower cost, o1-Mini still offers strong reasoning capabilities, making it a good choice for tasks like coding without requiring a large budget.
Safety Measures
OpenAI has introduced new safety protocols for the o1 model, ensuring it follows guidelines better than before. Key measures include:
- Improving the model’s safety score from 22 to 84.
- Collaborating with UK and US governments to thoroughly test its safety.
- Giving early access to safety experts to review and enhance the model.
Implications for Jobs and Research
The o1 model could impact industries like software development and data analysis, where it may shift job roles, pushing people to focus more on creative thinking and problem-solving. At the same time, AI like o1 could lead to new opportunities in AI safety and ethics.
For researchers, o1 can be a valuable tool, helping to solve complex problems faster, particularly in areas like healthcare and scientific research. It can also process large amounts of data efficiently, making research more streamlined.
How to Access OpenAI o1
The o1 model is available to ChatGPT Plus and Team users. Users can choose between the o1 and o1-Mini models using a model picker feature. Currently, users can send up to 30 messages to o1-preview and 50 messages to o1-Mini each day. In the future, OpenAI plans to increase these limits and automatically select the best model based on the user’s needs. ChatGPT Enterprise and Edu users will gain access next week.
