When Openai began to present special demonstrations for the new GPT-4 technology in late 2022, its skills Even most of the most experienced artificial intelligence researchers shocked. It can answer questions, write hair and create a computer code in ways that looked before their time.
After more than two years, Openai released: GPT-4.5. New technology indicates the end of an era. Openai said that the GPT-4.5 will be the latest version of its Chatbot system, which did not do “great thinking”.
After this version, the Openai technology, like humans, may spend a great time to think about a question before answering, rather than providing an immediate response.
GPT-4.5, which can be used to run the most expensive version of Chatgpt, is unlikely to be born with excitement like GPT-4, in large part of it because artificial intelligence research has turned in new directions. However, the company said that technology “will feel more natural” than previous Chatbot techniques.
“What distinguishes the model from each other is his ability to engage in warm, intuitive conversations and naturally flow, and we believe that it has a stronger understanding of what users mean when they ask for something,” said Mia Galaiz, Openai’s deputy chief of research.
In the fall, the company The technology called Openai O1 was presentedIt is designed to cause tasks that involve mathematics, coding and science. The new technology was part of a wider effort to build artificial intelligence that could think through complex tasks. Companies like Google, Meta and DibsicIt is a Chinese starting company, which is developing similar technologies.
The goal is to create systems that can solve a problem carefully and logically through a series of separate steps, each one building on the latter, similar to the cause of humans. These technologies can be especially useful for computer programmers who use artificial intelligence systems to write code.
These thinking systems depend on technologies such as GPT-4.5, which are called large language models, or LLMS
LLMS learns their skills by analyzing huge quantities of text that has been executed online, including Wikipedia articles, books and chat records. By defining patterns in all this text, they learned to create a text on their own.
To build thinking systems, LLMS companies through an additional process called reinforcement learning. Through this process – which can extend over weeks or months – the system can learn behavior through a comprehensive experience and error.
By working through various mathematics problems, for example, he can learn methods that lead to the correct answer that do not do so. If you repeat this process with a large number of problems, it can determine the patterns.
Openai and others believe this is the future of developing artificial intelligence. But in some respects, they were forced in this direction because they have Influence internet data The necessary to train systems such as GPT-4.5.
Some of the systems of causes outperform the regular LLMS in certain unified tests. But uniform tests are not always a good judge on how technologies perform in real world positions.
Experts point out that the new thinking system cannot necessarily be a cause like a person. And like other Chatbot technologies, they can still get a mistake and make things – a phenomenon called hallucination.
Openai said that starting on Thursday, GPT-4.5 will be available to anyone who subscribed to Chatgpt Pro, a $ 200 service per month that provides access to the latest tools of the company.
(New York Times prosecution Openai and its partner, Microsoft, in December due to the violation of copyright for news content related to artificial intelligence systems.)
