When Openai began to give private demonstrations of its new GPT-4 technology at the end of 2022, its skills Shocked even the most experienced AI researchers. He could answer questions, write poetry and generate computer code so as to be far from his time.
More than two years later, Openai published his successor: GPT-4.5. The new technology means the end of an era. Openai said that GPT-4.5 would be the latest version of its chatbot system which has not made “chain reasoning”.
After this version, Openai technology can, like a human, spend a lot of time thinking about a question before answering, rather than providing an instant answer.
The GPT-4.5, which can be used to feed the most expensive version of Chatgpt, is unlikely to generate as much excitement as the GPT-4, largely because research on AI has moved to new directions. However, the company said that the technology “would feel more natural” than its previous Chatbot technologies.
“What distinguishes the model is its ability to engage in warm, intuitive and naturally flowing conversations, and we think that it has a stronger understanding of what users mean when they ask for something,” said Mia Glaese, vice-president of research in Openai.
In the fall, the company Introduced technology called Openai O1which was designed to reason through tasks involving mathematics, coding and science. The new technology was part of a wider effort to build the AI that can reason through complex tasks. Companies like Google, Meta and In depthA Chinese start-up, develops similar technologies.
The objective is to build systems that can carefully and logically solve a problem through a series of discreet steps, each building on the last one, similar to the way humans words. These technologies could be particularly useful for IT programmers who use AI systems to write code.
These reasoning systems are based on technologies like GPT-4.5, which are called large language models, or llms
The LLM learn their skills by analyzing enormous amounts of text covered on the Internet, including articles, books and Wikipedia discussion newspapers. By identifying reasons in all this text, they learned to generate text by themselves.
To build reasoning systems, companies put LLM through an additional process called learning to strengthen. Thanks to this process – which can extend over weeks or months – a system can learn behavior thanks to large trials and errors.
By working on various mathematical problems, for example, he can learn which methods lead to the right answer and those who do not. If he repeats this process with a large number of problems, he can identify the models.
Openai and others believe that it is the future of the development of AI. But in some respects, they were forced in this direction because they have On internet data necessary to form systems like GPT-4.5.
Some reasoning systems surpass ordinary LLMs on certain standardized tests. But standardized tests are not always a good judge of how technologies will work in real situations.
Experts point out that the new reasoning system cannot necessarily reason like a human. And like other chatbot technologies, they can always be mistaken and invent things – a phenomenon called hallucination.
OPENAI said that from Thursday, GPT-4.5 would be available for anyone subscribed to Chatgpt Pro, a $ 200 service per month that gives access to all the latest business tools.
(New York Times heard OPENAI and its partner, Microsoft, in December for the violation of copyright of news content linked to AI systems.)