The article discusses the development of AI chatbots, like ChatGPT and DeepSeek, that now incorporate reasoning capabilities. Unlike earlier versions that provided instant answers, these newer systems take time to process and break down complex problems before providing a solution. This involves extra computational work, potentially employing trial-and-error or step-by-step approaches.
OpenAI's updated ChatGPT is highlighted as a key example. Initially designed for tasks in math, science, and programming, it demonstrated improved performance on industry-standard AI progress tests. Competitors such as Google, Anthropic, and China's DeepSeek have also developed similar technologies.
The article raises questions regarding the true nature of AI reasoning. Is it comparable to human reasoning, or simply a more advanced form of computation? The core definition provided is that AI reasoning involves extra computational work after a question is posed, allowing for a more deliberate and thorough response than instant answers provided by older systems.