AI Researchers Can Now Conduct Research Completely Autonomously

Source: Daily AI Research Report #61

Lead

SakanaAI has launched AI Scientist v2, a system that can carry out scientific research at workshop level fully autonomously. The breakthrough represents a quantum leap in automated research.

Content

Artificial intelligence has reached a new milestone: researchers at SakanaAI have developed a system that can carry out complete scientific research without human intervention.

AI Scientist v2 uses what they call "agentic tree search" - a method in which the AI system systematically explores various research possibilities, generates hypotheses, designs experiments, and draws conclusions based on the results.

How does it work?

The system combines several AI agents that work together:

1. A "hypothesis generator" that proposes research questions
2. An "experiment designer" that plans how the hypotheses should be tested
3. A "data analyzer" that interprets the results
4. A "paper writer" that documents the findings

This approach resembles how human research teams work, but operates in a fraction of the time.

Implications for research

Experts believe this could revolutionise scientific research:

• Speed: Hypotheses can be tested and falsified in hours instead of months
• Scaling: Thousands of lines of research can be explored in parallel
• Objectivity: Reduces human bias in the research process

The GitHub repository already has over 3,500 stars and is generating great interest in the research community.

Ethical questions

The development also raises important questions:

• Who gets the credit for AI-generated discoveries?
• How do we ensure quality control of autonomous research results?
• Will this replace human researchers or free them up for more creative work?

SakanaAI emphasises that the system is intended as a tool to accelerate research, not to replace human creativity and intuition.

Links

• GitHub: https://github.com/SakanaAI/AI-Scientist-v2
• Source report: /nexus/research/data/2026-03-29-report.md

Meta

• Keywords: AI, research, automation, machine learning, scientific discovery
• Author: Dr. Alban (AI assistant)
• Created: 2026-03-29 02:00 UTC
• Requires editorial review: Yes