Recently, a new trend has been emerging in the scientific community — AI scientists. In simple terms, this type of artificial intelligence can not only propose hypotheses and conduct experiments but also independently write research papers. Although I have always believed that AI's best role is as a tool for humans rather than a replacement, this type of AI scientist seems to show considerable potential.
Image source: Picture generated by AI, authorized service provider Midjourney
We often hear about scientists utilizing AI, such as the 2019 study from the Massachusetts Institute of Technology. The research team trained AI to successfully discover a new antibiotic called Halicin, which can combat drug-resistant bacteria like MRSA. This AI's success has filled people with anticipation for the prospects of AI scientists.
Currently, Japan's Sakana AI company is funding laboratories at the University of British Columbia and Oxford University, aiming to develop an AI scientist capable of conducting scientific research entirely independently. This AI will study existing scientific literature, propose hypotheses, conduct experiments, and eventually write a complete research paper.
To reduce errors, the UBC research team has established a step-by-step execution process. Initially, the AI scientist will propose several hypotheses from existing literature and score these ideas based on "interest, innovation, and feasibility." Next, the AI selects a hypothesis and ensures through a literature database that the idea is novel. Then, the AI uses a coding assistant program to conduct experiments, records research notes, judges whether follow-up experiments are needed, and finally writes the research paper.
However, such AI scientists are not perfect and still face some issues. For example, AI may exhibit biases, tending to select data that has been extensively researched. Although in some cases its "innovative thinking" can help solve complex problems, this lack of intuition and experience may also lead to some clumsy conclusions.
Additionally, the development of AI scientists has raised some ethical issues, such as who should be responsible for the achievements of AI scientists, or who should bear the responsibility for their errors, plagiarism, etc.
In the world of science, the application of AI has led to breakthroughs in many fields. Whether it's finding new antibiotics or predicting weather patterns, AI is gradually becoming an indispensable assistant to scientists. I believe that the greatest achievement of AI scientists will be their collaboration with human scientists, jointly leading the future of scientific research.
Key Points:
1. 🤖 AI scientists can independently propose hypotheses, conduct experiments, and write research papers.
2. 🔍 Although AI scientists show innovative potential, they still face issues with biases and inaccurate conclusions.
3. 🧠 Collaboration with human scientists is considered key to the greatest achievements of AI scientists.