Of course, the idea of fully automating scientific research has long been a dream. A team of researchers has lately taken a big step to achieve the goal. It has come up with AI Scientist. It is developed by Sakana AI and researchers from Canada and the UK. The AI Scientist is designed to handle the entire research process. It is designed to read existing literature and form hypotheses to conduct experiments and even write research papers.
AI Scientist is built on a large language model and performs tasks by searching through existing research. It makes small adjustments to algorithms and tests them to see if they improve. It even conducts a kind of automated peer review to evaluate its own results.
However, the AI Scientist has some limitations. It is currently capable of only conducting research in the field of machine learning. It cannot perform laboratory work and this is a significant part of many scientific disciplines. Gerbrand Ceder, a materials scientist, points out that AI can generate hypotheses, but there is still a long way to go before robots can carry out physical experiments.
The results produced by AI Scientist are incremental and not groundbreaking. Some researchers are not happy with output quality and states that the system lacks creativity. Science now is something more than just analyzing data and writing papers. It involves discussions, debates and informal exchanges of ideas too.
However, it is true that AI Scientist represents an important step in the ongoing effort to automate certain aspects of scientific work. Even though it might not replace human researchers, it can of course handle several repetitive tasks and this would be great help to scientists.
