AI is on track to become much more than just one assistant. According to a report by Nature.com, Openi chief scientist Jakub Pachoki suggested that in the near future, the AI ​​models will be able to conduct original scientific research on their own – a jump beyond humans with the help of humans with current knowledge.

Pachoki, who took over as the chief scientist of OpenAII in 2024, says that the models are gradually moving towards being able to work with minimal human inputs. While devices such as Chatgpt still depend on signals and continuous direction, Openai has already developed a system that can work freely for a short period and produce useful results. He says that it can develop rapidly if more computing power is applied to scientific challenges with open ends.

He pointed to deep research, a system designed by Openai, as an initial example. Even with limited calculations, the equipment can process and analyze large versions of uncontrolled information for several minutes in advance. With more advanced models and more resources, Pachoki says that AI may soon contribute to areas such as automated software development, hardware engineering and even novel scientific discoveries.

These abilities are the process of learning reinforcement in the heart – a process that teaches models through reaction and recurrence. Pachoki explains that while the AI ​​models initially learn by absorbing the dataset on a large scale during pre-training, real progress occurs when learning reinforcement helps them develop strategy to solve complex tasks.

He says that recent reforms in logic models are largely inspired by how reinforcement learning is being used-only to polish the model, but also to help develop its way of thinking and decision making. This is a change from earlier models that mimic the output of human-like output depending on the pattern only in data.

However, Pachoki also admits that the AI ​​models do not wonder how humans do. They do not remember how or when they learned, and they lack time and experience understanding. Nevertheless, their ability to simulate logical stages and solve problems makes them useful tools for tasks that require structured thinking.

Since Openai works on a more powerful model and soon prepares to release an open weight version, focusing on carrying forward the limits of achieving AI in Pachoki science and technology. While the debate around the Artificial General Intelligence (AGI) continues, they believe that the real test will be how well these systems can work freely and can deal with problems that we have not yet solved.