When Openai started to give private demonstrations of its new GPT-4 technology at the end of 2022, the skills even shocked the most experienced AI researchers. It could answer questions, write poetry and generate computer code in a way that seemed far ahead of its time.
More than two years later, OpenAi released its 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 the chatbot system that did no 'reason for the reasoning'.
After this release, the technology of OpenAi, just like a person, can spend a considerable amount of time thinking about a question before he answers, instead of giving an immediate answer.
GPT-4.5, which can be used to provide the most expensive version of Chatgpt with power, is probably not as much excitement as GPT-4, largely because AI research has been shifted in new directions. Yet the company said that the technology would “feel more natural” than his earlier chatbot technologies.
“What distinguishes the model is the ability to have warm, intuitive, naturally flowing conversations, and we think it has a stronger understanding of what users mean when they ask for something,” said Mia Glaese, vice -president of research at OpenAi.
In the fall, the company introduced technology called OpenAI O1, which was designed to reason by tasks with mathematics, coding and science. The new technology was part of a broader effort to build AI that can reason through complex tasks. Companies such as Google, Meta and Deepseek, a Chinese start-up, develop comparable technologies.
The aim is to build systems that can solve a problem carefully and logically through a series of individual steps, each built on the last, similar to how people reason. These technologies can be useful in particular for computer programmers that use AI systems to write code.
These reasoning systems are based on technologies such as GPT-4.5, which are called large language models or LLMs
LLMS teaches their skills by analyzing enormous amounts of text that have been broken down from the internet, including Wikipedia articles, books and chat logs. By locating patterns in all that text, they learned to generate text themselves.
To build reasoning systems, companies place LLMS through an extra process called reinforcement learning. Through this process – which can extend in weeks or months – a system can learn behavior through extensive trial and error.
By going through different mathematical problems, for example, it can learn which methods lead to the correct answer and which do not. If it repeats this process with a large number of problems, it can identify patterns.
OpenAi and others believe that this is the future of AI development. But in some respects they are forced in this direction because they have no internet data that is needed to train systems such as GPT-4.5.
Some reasoning systems perform better than regular LLMs on certain standardized tests. But standardized tests are not always a good judge how technologies will perform in real-world situations.
Experts point out that the new reasoning system cannot necessarily reason a person. And just like other chatbot technologies, they can still get things wrong and come up with things – a phenomenon called hallucination.
OpenAI said that GPT-4.5 would be available from Thursday for anyone who was subscribed to Chatgpt Pro, a service of $ 200 a month that offers access to all the newest tools from the company.
(The New York Times has sued OpenAi and his partner, Microsoft, in December for infringing the copyright of news content with regard to AI systems.)
