The true pace of AI progress is being underestimated

—

© [2025] [𝘒𝘐-𝘎𝘦𝘯.] 𝘔𝘢𝘳𝘤𝘦𝘭 𝘖𝘩𝘳𝘦𝘯𝘴𝘤𝘩𝘢𝘭𝘭. 𝘉𝘢𝘴𝘪𝘦𝘳𝘦𝘯𝘥 𝘢𝘶𝘧 𝘞𝘦𝘳𝘬𝘦𝘯 𝘷𝘰𝘯 𝘈𝘯𝘥𝘳𝘦𝘢𝘴 𝘖𝘩𝘳𝘦𝘯𝘴𝘤𝘩𝘢𝘭𝘭.

Many people still see AI primarily as a chatbot or a more advanced search engine. According to OpenAI, however, the technology has already moved well beyond this: modern systems can now handle complex reasoning tasks that previously took experts hours, while public perception is struggling to keep pace with this development.

The company expects AI models to soon be able to take on tasks that would require a human days or even weeks to complete. According to OpenAI, the first autonomous discoveries by AI could become realistic from 2026, with larger breakthroughs potentially following from 2028. This is being enabled by a dramatic decline in the “price per unit of intelligence”, which has fallen by a factor of 40 each year in recent years.

Experts expect AI not only to make processes more efficient or help reduce costs, but increasingly to become part of companies’ value-creation processes and therefore an integral element of their business models.

What does this mean for the use of AI in professional research?

AI is rapidly evolving from an assistant into a genuine research partner. This makes it possible not only to analyse complex volumes of information more quickly, but also to identify new connections and insights automatically – potentially providing significant leverage for many forms of research.

At the same time, more complex tasks and reasoning modes increase costs for technology companies, as they require more tokens per task. This is likely to lead AI providers to raise prices in future and make more services paid-for. The alternative is to use cheaper, less capable models that require fewer resources.

By Maria Kleiner