Even the best experts in A.I. found themselves aghast by the capabilities shown by OpenAI in its revolutionary GPT-4 technology late last year. It could remarkably answer questions, write poems, and generate computer code.
After two years of waiting, OpenAI introduces its next model to the world new GPT-4.5. This new generation of models differs critically, as OpenAI stated that GPT-4.5 will be the last chatbot under its portfolio that does not support ‘chain-of-thought reasoning.’
With this release, the artificial intelligence technologies of OpenAI could begin to simulate human-like contemplation, where it could take time before an answer to a question is given rather than rushing to speculate.
GPT-4.5, which powers the premium version of ChatGPT, may not cause as much excitement as its predecessor, all the evolving context under A.I. research. However, the company states that this technology will give a ‘more natural’ experience than previous chatbot iterations.
“What distinguishes this model is its capacity for engaging in warm, intuitive, and fluid conversations. We believe it has a deeper understanding of user intent when they make requests,” explained Mia Glaese, OpenAI’s vice president of research.
This was what the organization introduced during the fall: OpenAI o1, a technology meant to reason through tasks of math, coding, or even science. This is part of a larger vision to have A.I. thinking through the limits of logic applied to many problems. Other competitive companies such as Google, Meta, and the Chinese startup DeepSeek are also competing in such offerings.
To better develop such systems, issues should be approached methodically and logically through a chain of steps, similar to how human reasoning occurs. Such technologies could prove invaluable for computer programmers who would arm A.I. with the ability to execute typical programming tasks.
These reasoning systems are built around technologies such as GPT-4.5, which are actually classified under large language models or L.L.M.s.