AI Progress and Recommendations

AI is moving beyond simple chatbots and entering a stage that surpasses human intellectual capability. Currently, AI performs tasks in software engineering and other fields—tasks that used to take seconds—over hours, and it is expected to soon tackle complex tasks that take days or weeks.

The cost per unit of intelligence is dropping sharply, by about 40x every year. OpenAI predicts that by 2026, AI will begin making minor discoveries, and from 2028 onward, it will have systems capable of making meaningful scientific discoveries.

These advances will drive innovation across various fields including healthcare, materials science, drug development, climate modeling, and personalized education. However, we must prepare for rapid economic transitions and changes to the social contract.

OpenAI recognizes the risks of superintelligent systems and treats Safety and Alignment as top priorities. To this end, it proposes the following:

• Common safety standards and sharing of research findings among frontier labs
• Building public oversight and accountability frameworks commensurate with AI capabilities