OPENAI, the company behind the popular chatbot chatbot chatgpt, presented its Last model of artificial intelligence, Openai O1Designed to excel in complex reasoning tasks in science, coding and mathematics. This new model aims to fight against some of the limits observed in previous models by improving the way in which AI “thinks” before generating answers.
In a press release, Openai explained that the O1 model was formed to spend more time deliberating problems, imitating the way humans reason. “We have trained these models to spend more time thinking about problems before responding, just like a person,” said society. This approach, which Optai calls “reasoning”, allows the model to refine its reflection process, to consider several strategies and to recognize errors before presenting an answer. The objective is to improve precision, especially in more complex tasks.
The new O1 series is OPENAI's response to the current challenge of AI models generating incorrect or misleading information, often called “hallucinations”. This was a persistent problem with previous models like GPT-4, which, although powerful, fought with complex reasoning tasks and sometimes brought incorrect responses with great confidence.
According to Openai, O1 has already demonstrated impressive capacities in internal tests. It works at a level comparable to doctoral students on difficult problems in physics, chemistry and biology. In a qualification examination for the prestigious International Mathematics Olympiad, O1 correctly resolved 83% of the problems, a significant improvement compared to the success rate of 13% of GPT-4.
The potential applications of the model are large. OPENAI has highlighted its use in health care to annotate cell sequencing data and in physics to generate complex mathematical formulas used in quantum perspective. This improved reasoning could benefit the fields which require workflows in several stages or detailed analytical work, such as scientific research, software development and engineering.
Openai highlighted the importance of security because it deploys this new generation of AI models. The company claims that the O1 model is more robust when it comes to joining the security guidelines, in particular in the scenarios where users are trying to bypass its safety measures, known as “jailbreaking”. In the tests, the O1 model surpassed the GPT-4 to resist Jailbreak attempts, marking 84 on a difficult security test compared to the GPT-4 score of 22.
The company has also concluded agreements with AI security institutes in the United States and the United Kingdom, granting early access to the evaluation and test model. OPENAI works with the two governments to ensure that the development and deployment of AI technologies follow strict security standards, in particular given the potential risks of AI in fields such as disinformation and biases.
Despite this progress, the demand for growing energy posed by large models of AI as Openai O1 becomes an important concern. These models require large amounts of computing power to generate complex responses, which in turn leads to energy consumption.
Openai advances pave the way to companies like Qudata to improve the experiences of daily users. With Qudata AI Assistant, available for free, users can Take personalized and interactive cats with a personalized chatbot Based on the Chatppt. Whether you are looking for information on a specific subject or just want to have an occasional conversation, the Qudata AI chatbot can provide intelligent responses and devoted to a wide range of subjects.
The publication of Openai O1 is only the beginning of what the company describes as a new era of the AI focused on reasoning. This is the first model in the O1 series, and Openai plans to publish updates and regular improvements. Although the current model lacks certain features that have made the Chatppt so popular – such as web browsing and file downloads – this should evolve quickly.
